Route planning and predictive traffic analysis using

Machine Learning
A Project Report

Submitted
in the partial fulfilment of the requirements
for the award of degree of
Bachelor of Technology
in

COMPUTER SCIENCE & ENGINEERING

(JNTUA, Anantapuramu)
by
J Vidyavathi 202N1A0522
G Ramesh 202N1A0520
T Revanth Rakesh 202N1A0539
S Md Sohail 202N1A0534

Under the guidance of

Mrs. G. Alekhya
Assistant Professor, Dept. of CSE
BITS, Kurnool

Department of Computer Science & Engineering

BRINDAVAN
INSTITUTE OF TECHNOLOGY & SCIENCE (BITS-KNL)
(Approved by AICTE & Affiliated to JNTUA) N.H.
7, PEDDATEKUR, KURNOOL-518218.
May- 2024
 BRINDAVAN
INSTITUTE OF TECHNOLOGY & SCIENCE (BITS-KNL) N.H.
7, PEDDATEKUR, KURNOOL-518218.

CERTIFICATE

This is to certify that the Project Work Report entitled “Route planning and
predictive traffic analysis - using Machine Learning” is the Bonafede record of
the Project Work carried out under my Guidance by

J Vidyavathi 202N1A0522
G Ramesh 202N1A0520
T Revanth Rakesh 202N1A0539
S Md Sohail 202N1A0534

in partial fulfilment of the requirements for the award of Bachelor of Technology in

Computer Science & Engineering and is submitted to the Department of Computer
Science & Engineering, Brindavan Institute of Technology & Science, N.H.-7,
Peddatekur, Kurnool.

Head of the Dept. Guide

Mr. S. Srihari Reddy Mrs. G. Alekhya

Assistant Professor, Dept. of CSE Assistant Professor, Dept. of CSE
Brindavan Brindavan
Institute of Technology & Science, Institute of Technology &
N.H. 7, Peddatekur, Kurnool. science, N.H. 7, Peddatekur,
Kurnool.

External Examiner
 ACKNOWLEDGMENTS

Our hard work never shines if we do not convey our heartfelt gratitude to
those people from whom we got considerable support and encouragement during this
Project Work.

We would like to express our gratitude to our Guide Mrs. G. Alekhya,

Assistant Professor, Department of Computer Science & Engineering,
BRINDAVAN INSTITUTE OF TECHNOLOGY & SCIENCE, for her constant
support and guidance throughout the Project Work.

We would like to thank Mr. S. Srihari Reddy, Asst. prof & Head of the
Department, Computer Science & Engineering, BRINDAVAN INSTITUTE OF
TECHNOLOGY & SCIENCE, for his valuable suggestions from time to time
during this Project Work.

We also express our special thanks to Prof. N. Siva Prasad Reddy, Director
(Academics), Dr. S. Ramesh Reddy, Principal, and Prof. P. Girish Reddy, Vice
Principal, BRINDAVAN INSTITUTE OF TECHNOLOGY & SCIENCE, for their
support and encouragement to complete our Project Work.
We would like to thank Mr. Abdul Yunus Basha, Assistant Professor of
Department of Computer Science & Engineering, Project Coordinator for his support
throughout the project.

We are happy to express our sincere thanks to all teaching and non-teaching
staff, Department of Computer Science & Engineering. for their help. Finally, we are
thankful to all those who helped directly or indirectly in making endeavour a success.

Project Associates
 ABSTRACT

The ever-increasing traffic congestion in urban areas has become a significant

concern worldwide. Accurate prediction of traffic flow is crucial for effective traffic
management and planning. In this project, we propose a machine learning-based
approach to predict traffic flow based on vehicle counts and geographic information.
Initially, the system takes inputs of bike, car, and truck counts from various locations.
These counts are then used to classify the traffic into three categories: normal, high,
and low traffic. This initial phase provides a baseline for understanding the current
traffic situation in different areas. In the next phase of the project, we plan to enhance
the system by incorporating a mapping feature. Users will input the source and
destination of vehicles, and the system will calculate the distance between the two
points. This additional feature will provide a more comprehensive understanding of
the traffic situation, taking into account not only the current vehicle count but also the
distance between different locations. By combining vehicle count data with
geographic information, our system aims to provide more accurate and insightful
traffic flow predictions. The machine learning algorithms used in this project will be
trained on historical traffic data to improve prediction accuracy. Overall, our project
aims to contribute to better traffic management and planning by providing a reliable
tool for predicting traffic flow in urban areas. With further development, this system
could potentially be integrated into existing traffic management systems to provide
real-time traffic flow predictions and optimize traffic flow in urban areas. Feel free to
adjust it according to your project's specific details and requirements.

Key words: Traffic congestion, Machine learning, Traffic flow prediction, Vehicle
counts, Geographic information, Classification, Normal traffic, High traffic, Low
traffic, Mapping feature, Source and destination, Distance calculation

i
 TABLE OF CONTENTS
S.NO LIST OF CONTENTS PAGE NO

Abstract i

List of Figures ii

List of Results iii

1. Introduction
1.1 Background 1
1.2 Objectives 3
1.3 Scope 5
1.4 Overview of Traffic Flow prediction 5

2. Literature Review
2.1 Previous work in traffic flow prediction 6
2.2 Key finding from previous studies 6
2.3 Maps and challenges in existing approaches 7

3. Data Collection
3.1 Data Source 10
3.2 Data processing 11
3.1.1 Data cleaning techniques for traffic 12
prediction
3.1.2 Handling Missing Values in Traffic data 13
3.1.3 Data Nomination 14
3.3 Feature Engineering 15

4. Functionality of project
4.1 Existing system 16
4.2 Proposed System 17
4.3 Over of proposal Model 18
 4.4 Block Diagram 19

5. System Design
5.1 UML Diagram 20
5.2 Block Diagram 21
5.3 Architecture 22
5.4 Model Training 24
5.5 Model Evaluation 25

6. Functional Requirement
6.1 Software Requirements 26
6.1.1 Backend Technology 26
6.1.2 Frontend technology 28
6.2 Machine Learning model 29
6.2.2 Type of machine learning model 33
6.3 System Component 34
6.4 Mapping System 35

7. Implementations
7.1 Code Explanation 38
7.2 Results 49

8. Conclusion and learning outcomes

8.1 Conclusion 54
8.2 Future Enhancement 55
8.3 Learning Outcomes 56

9. References 58
 LIST OF FIGURES

Fig.3.1 Traffic conditions 15

Fig.5.1 UMl diagram 19

Fig.5.2 Block diagram 20

Fig.5.3 Architecture 21

Fig.6.1 Overview of ML 30

Fig.6.2 Types of ML 31

Fig.6.3 Route Mapping 35

ii
 LIST OF RESULTS

Fig .7.1 User Interface 50

Fig.7.2 Authenticate the traffic conditions 51

Fig.7.3 Indicate the login page of the candidate 52

Fig.7.4 Mapping system to input the source and destination 52

Fig.7.5 Route planning and distance calculation 53

iii
CHAPTER 1
Introduction
1.1Background

Traffic congestion is a significant problem in urban areas, leading to wasted time,

increased fuel consumption, and environmental pollution. Accurate prediction of traffic flow is
crucial for effective traffic management and planning. Traditional methods of traffic flow
prediction often rely on historical traffic data and statistical models, which may not capture the
complex and dynamic nature of traffic patterns accurately.
With the advancements in machine learning (ML) algorithms and the availability of
large-scale traffic data, ML-based approaches have emerged as a promising solution for traffic
flow prediction. ML models can effectively learn complex patterns and relationships from large
datasets, allowing for more accurate and reliable traffic flow predictions.
By leveraging ML techniques, traffic flow prediction systems can consider various
factors such as vehicle counts, weather conditions, road infrastructure, and historical traffic
patterns to provide more accurate and real-time predictions. These systems can help
transportation authorities and urban planners better understand traffic dynamics, optimize traffic
flow, and alleviate congestion in urban areas. Traffic congestion is a significant issue in urban
areas worldwide, causing substantial economic losses, increased travel times, fuel consumption,
and environmental pollution. According to a study by INRIX, a global leader in transportation
analytics, congestion costs the United States alone nearly $87 billion annually in lost productivity
and wasted fuel. Furthermore, the average American commuter spends 54 hours stuck in traffic
each year.

Traditional traffic management systems typically rely on historical data and static models
to predict traffic patterns and manage congestion. While these systems have been effective to
some extent, they often struggle to adapt to dynamic traffic conditions, resulting in suboptimal
traffic flow management. Moreover, the increasing complexity of urban transportation networks,
along with the growing volume of vehicles on the road, has made traditional traffic management
approaches less effective.

In recent years, there has been growing interest in leveraging machine learning (ML)
techniques to predict traffic flow more accurately. ML models have demonstrated promising

1
results in various fields, including image recognition, natural language processing, and now,
transportation management. By analysing large volumes of traffic data, ML algorithms can
identify complex patterns and relationships that may not be apparent to traditional traffic
management systems.

ML-based traffic flow prediction systems have the potential to revolutionize urban
transportation management by providing real-time insights into traffic conditions and predicting
future traffic flow with greater accuracy. These systems can enable transportation authorities to
implement proactive measures to manage congestion, such as optimizing traffic signal timings,
identifying alternative routes, and adjusting toll prices dynamically.

Key benefits of ML-based traffic flow prediction systems include:

1. Improved Accuracy: ML models can analyse large and diverse datasets to identify
complex patterns and relationships, leading to more accurate traffic flow predictions.

2. Real-time Insights: ML models can provide real-time insights into traffic

conditions, allowing transportation authorities to respond quickly to changing traffic patterns and
manage congestion more effectively.

3. Adaptability: ML models can adapt to changing traffic conditions and learn from
new data, making them more effective than static models at managing dynamic traffic
environments.

4. Cost-effectiveness: By reducing congestion and improving traffic flow, ML-based

traffic management systems can help reduce fuel consumption, lower emissions, and save
commuters time and money.

In this project, we aim to develop an ML-based traffic flow prediction system to

accurately predict traffic conditions and help transportation authority’s manage congestion more
effectively in urban areas. By leveraging state-of-the-art ML techniques and analysing real-time

2
traffic data, we aim to develop a scalable and adaptable traffic flow prediction system that can be
deployed in various urban environments

1.2 Objectives
Traffic flow prediction using machine learning (ML) aims to forecast the future traffic
conditions based on historical traffic data. The objective is to develop accurate models that can
predict traffic flow patterns, such as congestion, based on factors like time of day, day of the
week, weather conditions, and special events. By accurately predicting traffic flow,
transportation authorities can optimize traffic management, improve road safety, reduce
congestion, and enhance overall transportation efficiency. ML models can utilize various
algorithms such as regression, time series analysis, and deep learning to analyse large datasets
and make real-time predictions for better traffic management.
The primary objective of this project is to develop a machine learning (ML) based traffic
flow prediction system to accurately predict traffic conditions and assist transportation
authorities in managing congestion effectively in urban areas. To achieve this overarching
objective, the project has several specific objectives.

The first objective is to gather real-time and historical traffic data from various sources,
including traffic sensors, GPS data, and traffic cameras. This involves establishing data
collection pipelines to continuously gather data from these sources. Once the data is collected, it
needs to be pre-processed to remove noise, handle missing values, and standardize data formats
for further analysis. Data preprocessing techniques such as data cleaning, normalization, and
feature scaling will be employed to ensure that the data is suitable for training ML models.

The second objective is to identify and engineer relevant features that influence traffic
flow. These features may include historical traffic volume, weather conditions, road topology,
time of day, day of the week, and special events. Feature engineering techniques such as feature
selection, dimensionality reduction, and feature transformation will be employed to extract
meaningful features from the raw data. The goal is to develop techniques for feature extraction
that capture the underlying patterns and relationships in the data, which can then be used to train
ML models for traffic flow prediction.

3
 The third objective is to evaluate and select appropriate ML algorithms for traffic flow
prediction. Various ML algorithms such as regression models, decision trees, random forests, and
neural networks will be considered for model development. Factors such as prediction accuracy,
scalability, and computational efficiency will be taken into account when selecting the ML
algorithms. Once the ML algorithms are selected, the next step is to develop and train ML
models using the pre-processed data and selected features. The trained ML models will be used
to predict traffic flow for specific road segments and time intervals.

The fourth objective is to assess the performance of the trained ML models using
appropriate evaluation metrics. Metrics such as mean absolute error (MAE), root mean square
error (RMSE), and coefficient of determination (R-squared) will be used to evaluate the
prediction accuracy of the models. The performance of the ML models will be compared with
baseline models and traditional traffic flow prediction methods to evaluate their effectiveness.

The fifth objective is to integrate the trained ML models into a scalable and adaptable
traffic flow prediction system. This involves developing a software framework that can ingest
real-time traffic data, apply the trained ML models for prediction, and provide actionable insights
to transportation authorities. A user-friendly interface will be developed to allow transportation
authorities to access real-time traffic predictions and insights. The traffic flow prediction system
will be deployed in urban areas to assist transportation authorities in managing congestion
effectively.

The final objective is to evaluate and validate the performance of the deployed traffic
flow prediction system in real-world urban environments. The performance of the system will be
validated by comparing the predicted traffic flow with actual traffic flow data collected from
urban areas. Feedback will be collected from transportation authorities and users to evaluate the
usability, effectiveness, and impact of the system on managing congestion.

By achieving these objectives, the project aims to develop an effective and scalable ML
based traffic flow prediction system that can assist transportation authorities in proactively
managing congestion and improving traffic flow in urban areas.

4
1.3 Scope
The scope of this project is to develop a traffic flow prediction system using machine
learning techniques. The system aims to predict traffic flow patterns, including congestion, based
on historical traffic data, weather conditions, and other relevant factors. The project will involve
data collection, pre-processing, feature selection, and model training. Various machine learning
algorithms such as regression, decision trees, and neural networks will be explored and compared
to determine the most accurate prediction model. Additionally, the system will be designed to be
scalable and adaptable, allowing for real-time traffic flow prediction and integration with
existing traffic management systems. The final deliverable will be a robust and reliable traffic
flow prediction system that can assist transportation authorities in optimizing traffic flow,
reducing congestion, and improving overall road safety and efficiency.

1.4 Overview of Traffic Flow Prediction

Traffic flow prediction using machine learning involves using historical traffic data to
forecast future traffic conditions. The process typically includes data collection from various
sources such as traffic cameras, loop detectors, and GPS devices. Features such as traffic volume,
speed, time of day, weather conditions, and special events are analysed to train machine learning
models. Commonly used machine learning algorithms for traffic prediction include decision
trees, random forests, support vector machines, and neural networks. These models are trained on
historical data and then used to make predictions about future traffic conditions. By accurately
predicting traffic flow, cities can better manage traffic congestion, optimize traffic signal
timings, and provide real-time traffic information to commuters, ultimately improving overall
transportation efficiency and reducing travel time for commuters. Additionally, these predictions
can be used to develop intelligent transportation systems that can adapt to changing traffic
conditions in real-time.

5
CHAPTER 2
Literature
Review
2. Literature Review
Literature of machine learning is wide (Grunberg et al., 2003, Bradley and Fried, 1999,
Meek, 2001 and Walker, 2000). The following is a brief description of the various machine
learning articles. The paper by Shimmered Hermes (1993), describes an interactive note-taking
system for pen-based computers with two distinctive features. The system is an example of a
learning apprentice software-agent. A machine learning component characterizes the syntax and
semantics of the user’s information.
In another work, Soderland and Lehnert (1994) present a novel approach that uses
machine learning to acquire knowledge for some of the higher-level IE processing. It was found
that performance equals that of a partially trainable discourse module requiring manual
customization for each domain.
The work by Price et al. (2003) is another addition to the machine learning literature. The
authors propose and study a formal model of implicit imitation that can accelerate reinforcement
learning dramatically in certain cases.
Still on machine learning, the important study due to Gordon (2000) should be noted.
The study develops agents that are adaptive and predictable and timely.
This is to improve the efficiency of re-verification after learning, so that agent has a
sufficiently rapid response time.1.2 Motivation Traffic clog is one such reason because of which
a lot of people lose their important and significant amount of time. A successful implementation
of project will gradually reduce the traffic.

2.1 Previous work on traffic flow prediction

Time series analysis methods such as Autoregressive Integrated Moving Average
(ARIMA) have been widely used for traffic flow prediction. These methods rely on historical
traffic data to forecast future traffic patterns. Statistical methods like linear regression,
exponential smoothing, and seasonal decomposition have been employed for traffic flow
prediction. These methods analyse historical traffic data and identify patterns to make prediction.
Simulation models, including microsimulation and mesoscopic simulation, have been used to
predict traffic flow. These models simulate the movement of individual vehicles or groups of
vehicles to predict traffic conditions. Regression models such as Linear Regression, Support

6
Vector Regression (SVR), Decision Tree Regression, Random Forest Regression, and Gradient
Boosting Regression have been utilized for traffic flow prediction.
These models learn patterns from historical traffic data and make predictions based on
input features such as time of day, day of week, weather conditions, and road characteristics.
Time series forecasting models like ARIMA, Seasonal ARIMA (SARIMA), and Prophet have
been employed for traffic flow prediction. These models capture temporal dependencies in the
data and forecast future traffic patterns based on historical data. Deep learning models such as
Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) Networks,
Convolutional Neural Networks (CNN), and hybrid models have been used for traffic flow
prediction. These models are capable of capturing complex patterns in the data and making
accurate predictions based on large volumes of historical traffic data. Several studies have
compared the performance of different traffic flow prediction models. These studies have
evaluated the accuracy, robustness, and computational efficiency of various prediction models
under different traffic conditions and data scenarios. Comparative studies have highlighted the
strengths and limitations of different traffic flow prediction methods. While some methods may
perform well under certain conditions, they may fail to generalize to other traffic scenarios or
suffer from computational inefficiency. Previous studies have identified several factors that
affect traffic flow prediction, including weather conditions, special events or holidays, road
conditions, and other external factors. Time series analysis methods such as Autoregressive
Integrated Moving Average (ARIMA) have been widely used for traffic flow prediction.

2.2 Key findings from Previous studies

Traditional traffic flow prediction methods have limitations in capturing complex
patterns and dependencies in the data. These methods may fail to provide accurate predictions
under certain traffic conditions or data scenarios. Machine learning approaches for traffic flow
prediction may suffer from overfitting, data scarcity, and computational complexity. These
challenges need to be addressed to develop robust and efficient prediction models.

The quality and availability of data can significantly impact the performance of traffic
flow prediction models. Data pre-processing and cleaning techniques are required to handle
noisy and incomplete data and improve prediction accuracy. Scalability and real-time prediction
are critical challenges for traffic flow prediction systems. Prediction models need to be scalable
to handle large volumes of data and capable of making real-time predictions to support dynamic
7
traffic management and control. This detailed section provides an in-depth overview of previous
work on traffic flow prediction, including traditional approaches, machine learning methods, key
findings, and challenges faced by researchers in this field.
Previous studies have identified several factors that affect traffic flow prediction,
including weather conditions, special events or holidays, road conditions, and other external
factors. These factors need to be considered when developing traffic flow prediction models to
improve prediction accuracy.
Researchers have utilized various data sources for traffic flow prediction, including
historical traffic data, real-time traffic data, and open data sources and APIs.The availability and
quality of data have a significant impact on the performance of traffic flow prediction models.
Previous studies have used various evaluation metrics to assess the performance of traffic
flow prediction models, including Mean Absolute Error (MAE), Root Mean Squared Error
(RMSE), Mean Absolute Percentage Error (MAPE), and R-squared (R²) value. These metrics
provide insights into the accuracy and reliability of prediction models under different traffic
conditions and data scenarios.
2.3 Gaps and challenges in Existing Approaches
Traffic flow prediction methods have limitations in capturing complex patterns and
dependencies in the data. These methods may fail to provide accurate predictions under certain
traffic conditions or data scenarios. Machine learning approaches for traffic flow prediction may
suffer from overfitting, data scarcity, and computational complexity. These challenges need to be
addressed to develop robust and efficient prediction models.
The quality and availability of data can significantly impact the performance of traffic
flow prediction models. Data pre-processing and cleaning techniques are required to handle
noisy and incomplete data and improve prediction accuracy. Scalability and real-time prediction
are critical challenges for traffic flow prediction systems. Prediction models need to be scalable
to handle large volumes of data and capable of making real-time predictions to support dynamic
traffic management and control.
Traffic flow is influenced by a multitude of factors, including road infrastructure, weather
conditions, special events, and human behaviour. Capturing and modelling these complex
patterns and dependencies in the data poses a significant challenge for traditional prediction
methods. Machine learning approaches offer flexibility in capturing nonlinear relationships and
8
interactions among various features, but they may still struggle to accurately predict traffic flow
under certain conditions.
Machine learning models trained on limited or biased data may suffer from overfitting,
where the model learns to memorize the training data rather than generalize to unseen data.
Overfitting can lead to poor performance and unreliable predictions, especially in dynamic and
evolving traffic environments. Regularization techniques, cross-validation, and ensemble
methods can help mitigate overfitting and improve model generalization.
Obtaining high-quality and comprehensive traffic data can be challenging, particularly in
regions with limited infrastructure or resources. Data scarcity can hinder the development and
evaluation of traffic flow prediction models, leading to suboptimal performance and
generalization. Data augmentation techniques, transfer learning, and synthetic data generation
methods can help address data scarcity issues and improve model robustness.
Developing and training machine learning models for traffic flow prediction often
involves handling large volumes of data and complex algorithms, resulting in significant
computational complexity and resource requirement

9
 CHAPTER 3
Data Collection
3.1 Data source

1. Loop Detectors: Loop detectors are sensors embedded in roadways that detect the
presence and movement of vehicles. They provide real-time data on traffic volume, speed, and
occupancy. ML models can analyse this data to predict traffic patterns and identify congestion.
2. GPS and Mobile Apps: GPS data from vehicles and mobile apps like Google
Maps, Waze, and HERE provides valuable information on vehicle locations, speed, and travel
routes. This data can be anonymized and aggregated to predict traffic flow and congestion using
ML algorithms.
3. Weather Data: Weather conditions such as rain, snow, fog, and wind can
significantly impact traffic flow. Weather data from meteorological stations and weather APIs
provide information on temperature, precipitation, visibility, and wind speed. ML models can
incorporate weather data to improve the accuracy of traffic flow predictions.
4. Historical Traffic Data: Historical traffic data contains information on traffic
patterns, congestion, and incidents over time. ML models can analyse this data to identify trends
and patterns, allowing them to make accurate predictions about future traffic flow.
5. Public Transportation Data: Data from public transportation systems, such as bus
and train schedules, can also be valuable for predicting traffic flow. ML models can analyse this
data along with traffic data to understand how public transportation affects traffic patterns.
6. Social Media and News Feeds: Social media platforms and news feeds provide
real-time information about accidents, road closures, and other incidents that can impact traffic
flow. ML models can analyse this unstructured data to identify relevant events and incorporate
them into traffic flow predictions.
By integrating data from these diverse sources, ML models can accurately predict traffic
flow, helping drivers and transportation authority’s make informed decisions and alleviate
congestion on roadways.

10
3.2 Data Processing
Data pre-processing is an essential step in traffic flow prediction using machine
learning (ML) models. This process involves cleaning, transforming, and preparing raw data
collected from various sources such as sensors, cameras, and GPS devices for analysis.
Raw data is collected from various sources, including traffic sensors, cameras, GPS
devices, and other monitoring systems. This data includes information such as vehicle counts,
speed, occupancy, and other relevant traffic parameters.
The collected data often contains missing values, outliers, and noise, which need to be
addressed. Data cleaning involves identifying and handling missing values through imputation or
removal, detecting and filtering out outliers, and removing noise from the data.
Relevant features are extracted from the cleaned data to capture meaningful information
for traffic flow prediction. These features may include time of day, day of week, weather
conditions, road type, historical traffic patterns, and special events. Feature engineering
techniques such as one-hot encoding, label encoding, and feature scaling are applied to prepare
the features for model training.
The processed data is divided into training, validation, and test sets. The training set is
used to train the machine learning model, the validation set is used to tune hyperparameters and
evaluate model performance during training, and the test set is used to evaluate the final model
performance.
Before training the machine learning model, the features are scaled to ensure that they are
on the same scale and have similar ranges. Common feature scaling techniques include min-max
scaling and standardization (z-score normalization).
By processing the raw traffic data through these steps, it becomes suitable for training
machine learning models to predict traffic flow accurately.
One common normalization technique is Min-Max scaling, which scales the features to a
fixed range, usually between 0 and 1. This technique is beneficial when the features have varying
ranges and helps in preserving the relationships between the original values.
Another normalization technique is Standardization (Z-score normalization), which
transforms the features to have a mean of 0 and a standard deviation of 1. This technique
assumes that the data follows a Gaussian distribution and is less affected by outliers compared to
Min-Max scaling.

11
3.2.1 Data cleaning techniques for traffic prediction
In the realm of traffic prediction, data cleaning serves as the cornerstone for generating
reliable and accurate forecasts. The process begins with addressing missing values, a common
challenge in real-world datasets. Whether due to sensor malfunctions or incomplete data
collection, missing values can significantly distort predictions if left unattended. Employing
techniques such as mean imputation, interpolation, or advanced machine learning algorithms,
analysts fill in these gaps with estimated values based on surrounding data points, thereby
ensuring the continuity and integrity of the dataset.

Simultaneously, outlier detection and removal are imperative to maintain the robustness
of predictive models. Outliers, anomalies in the dataset that deviate substantially from the norm,
can skew predictions and compromise model performance. Through statistical methods like
Zscore analysis or interquartile range, analysts identify and mitigate outliers, either by correcting
erroneous data points or excluding them altogether, guided by domain expertise.

Furthermore, data smoothing techniques are applied to attenuate noise inherent in traffic
data, especially prevalent in time-series datasets. Techniques such as moving averages or
Savitzky-Golay filtering are employed to extract underlying trends and patterns, facilitating
clearer insights into traffic dynamics and aiding in more accurate prediction modelling.
Feature engineering emerges as another crucial aspect of data cleaning, where analysts
derive new features from existing ones or external sources to enrich the dataset. Incorporating
variables such as time of day, day of week, or historical traffic patterns enhances the predictive
capabilities of the model, capturing nuanced variations in traffic behaviour.

Normalization and scaling are indispensable for ensuring uniformity across features,
mitigating biases that may arise from differing scales or units. By standardizing feature values
through techniques like Min-Max scaling or standardization, analysts prevent certain features
from unduly influencing model outcomes, promoting fair and balanced predictions.

Redundant or irrelevant features are identified and pruned from the dataset through
techniques such as correlation analysis or principal component analysis, streamlining the dataset
for more efficient modelling.
12
3.2.2 Handling missing values in traffic Data
Handling missing values in traffic data is crucial for maintaining the integrity of traffic
prediction models. Initially, it's essential to identify missing values, discerning their patterns and
extent across features. Understanding the missingness mechanisms aids in selecting appropriate
imputation techniques. Common methods include mean, median, or mode imputation,
forward/backward fill, linear interpolation, and machine learning-based approaches. However,
the choice often relies on domain knowledge and the nature of the dataset. Evaluating imputation
performance using metrics like MAE or RMSE ensures the reliability of imputed values.
Sensitivity analysis further assesses the robustness of imputation methods under various
scenarios, enhancing confidence in the cleaned dataset. Through these systematic approaches,
missing values are effectively managed, preserving the quality and accuracy of traffic prediction
models.
In the realm of traffic data analysis, the presence of missing values poses a significant
challenge that demands meticulous attention. Once identified, missing values require careful
handling to prevent bias and maintain the integrity of subsequent analyses. This process typically
involves understanding the underlying mechanisms of missingness, selecting appropriate
imputation techniques, and evaluating their efficacy. Mean, median, or mode imputation offers
simplicity, while more sophisticated methods like linear interpolation or machine learning-based
approaches provide nuanced estimations, especially when data exhibit complex patterns.
However, the effectiveness of these techniques often hinges on domain knowledge and the
specific characteristics of the dataset. Rigorous evaluation, including sensitivity analysis, ensures
the reliability and robustness of the imputed values, thereby fortifying the foundation for accurate
traffic prediction models. By navigating through these steps with precision and insight, data
analysts can mitigate the impact of missing values, empowering their analyses to yield insights
that drive informed decision-making in traffic management and planning.

3.2.3 Data normalisation and scaling for traffic analysis

In traffic analysis, normalization and scaling are indispensable pre-processing steps aimed
at ensuring that features are comparable and conducive to effective machine learning model
training. These techniques are pivotal for handling diverse datasets containing features with
varying scales, units, and distributions. Normalization adjusts feature values to a common scale,

13
typically between 0 and 1, facilitating fair comparisons among different features. This is
particularly beneficial for traffic-related features such as traffic volume, speed, and vehicle
count, which may originate from different sources or measurement systems, thus requiring
harmonization for meaningful analysis.
Scaling, on the other hand, modifies the range of feature values without necessarily
confining them to a specific interval. One commonly used scaling method is Min-Max scaling,
which rescales feature values to a predetermined range, often between 0 and 1. This technique is
calculated by subtracting the minimum value of the feature from each value and then dividing by
the difference between the maximum and minimum values. Min-Max scaling is advantageous
when the distribution of the data is known and bounded, providing a standardized representation
of features across the dataset.
Alternatively, standardization, or Z-score normalization, transforms feature values to have
a mean of 0 and a standard deviation of 1. This technique is robust to outliers and suitable for
datasets where the distribution of the data is unknown or not bounded. Standardization ensures
that features with larger variances do not dominate the learning process and enables machine
learning algorithms to converge more effectively.
In the context of traffic analysis, both normalization and scaling are applied to various
features such as traffic flow, speed limits, road length, and time intervals. These preprocessing
techniques harmonize the disparate features, enabling a fair comparison and interpretation of
their contributions to the predictive model. By standardizing the input data, normalization and
scaling enhance the performance of machine learning algorithms, such as regression, clustering,
and neural networks, by improving convergence speed and preventing numerical instabilities.
Furthermore, normalization and scaling contribute to the interpretability and
generalization of the resulting models, ensuring that they can be effectively applied to new,
unseen data. Overall, normalization and scaling play a critical role in preparing traffic data for
analysis and modeling, paving the way for more accurate and reliable traffic management
solutions that can address complex transportation challenges effectively.

14
3.3 Feature Engineering

Fig: 3.1 Traffic conditions

Traffic Flow Prediction using Machine Learning," Here focus on feature engineering to
enhance the accuracy of our predictions. We begin by extracting essential features from the
vehicle count data, such as total vehicle count, vehicle type distribution, and time of day.
Additionally, we incorporate geographic features derived from our mapping system, including
distance between source and destination, road type, and traffic history along the route. By
carefully selecting and engineering these features, we aim to improve the performance of our
machine learning models in predicting traffic flow accurately under various conditions.

15
 CHAPTER 4
Functionality of
project
4.1 Existing System
Your project aims to develop a traffic flow prediction system using machine learning
techniques. The system will take inputs such as the count of bikes, cars, and trucks to determine
the traffic intensity. Based on this count, the system will categorize the traffic into three levels:
normal traffic, high traffic, and low traffic.
In the next phase of your project, you plan to incorporate a mapping system. Users will
input the source and destination of vehicles, and the system will provide the distance between
these locations. This integration will enhance the functionality of the system, providing users
with more comprehensive information about traffic conditions and travel distances.
The existing system involves data collection on vehicle counts, which serves as the basis
for traffic intensity prediction. Machine learning algorithms will analyse this data to classify
traffic into different levels. Additionally, the integration of a mapping system will provide users
with more detailed information about travel distances between locations.
The system's user interface will be intuitive, allowing users to easily input vehicle counts
and locations. The machine learning model will then process this data and provide real-time
predictions about traffic conditions. The mapping system will enhance user experience by
providing accurate distance information between destinations.
Overall, your project aims to provide a comprehensive solution for traffic flow prediction,
incorporating both vehicles count analysis and a mapping system to offer users valuable insights
into traffic conditions and travel distances.
Develop an intuitive user interface (UI) that allows users to input their source and
destination locations easily. Provide clear instructions and options for users to specify their travel
preferences and purposes (e.g., commuting, leisure, etc.). Ensure that the UI is user-friendly and
accessible across different devices and platforms.
Once the user inputs the source and destination locations, the system should utilize the
mapping API to calculate the distance between these locations. This distance information will be
valuable for users to plan their routes and estimate travel times accurately.
Implement a feedback mechanism where users can provide input on the accuracy of
traffic predictions and the usefulness of distance information. This feedback loop can help
improve the system over time by incorporating user preferences and addressing any shortcomings
in prediction accuracy

16
4.2 Proposed System
In our final year B. Tech project, we are developing a system for predicting traffic flow
using machine learning (ML) techniques. The system aims to assist commuters by providing
real-time information about traffic conditions and suggesting optimal routes.
The system utilizes ML algorithms to analyse the count of bikes, cars, and trucks on a
given route. Based on this data, it categorizes the traffic into three levels: normal traffic, high
traffic, and low traffic. This categorization helps commuters make informed decisions about their
travel routes and times.
In the next phase of our project, we plan to enhance the system by integrating a mapping
feature. Users will be able to input their source and destination locations, and the system will
provide the distance between the two points. By combining traffic flow prediction with mapping
capabilities, our system will offer a comprehensive solution for commuters.
The proposed system aims to address the following objectives:
1. Real-time Traffic Prediction: Using ML algorithms to analyse vehicle counts and
predict traffic conditions accurately.
2. Traffic Categorization: Categorizing traffic into normal, high, and low traffic to
help users plan their journeys efficiently.
3. Mapping System Integration: Integrating a mapping feature to provide users with
the distance between their source and destination locations.
4. Optimal Route Suggestions: Suggesting the most optimal routes based on
predicted traffic conditions and distance between locations.
Our system will be a valuable tool for commuters, helping them save time and make more
informed decisions about their travel routes.
By combining traffic flow prediction with mapping capabilities, we aim to create a
comprehensive solution that addresses the needs of modern commuters effectively.

17
4.3 Overview of proposed model
Your proposed system for traffic flow prediction using machine learning for your BTech
final project sounds promising and addresses a critical real-world problem.
The system aims to predict traffic flow by analysing the counts of bikes, cars, and trucks.
By inputting these counts, the system will categorize the traffic into three levels: normal, high, or
low traffic. This classification will provide users with an overview of the current traffic situation,
allowing them to plan their routes more efficiently.
In the next phase of the project, you plan to enhance the system by incorporating a
mapping feature. Users will be able to input their source and destination, and the system will
calculate the distance between the two points.
This addition will enable users to not only understand the current traffic situation but also
plan their routes based on the predicted traffic flow and the distance to their destination.
The proposed system offers several advantages. Firstly, it provides users with real-time
information about traffic conditions, allowing them to make informed decisions about their travel
routes. Secondly, by incorporating a mapping feature, users can plan their journeys more
effectively, considering both the predicted traffic flow and the distance to their destination.
Lastly, the use of machine learning algorithms ensures that the system can continuously learn and
improve its predictions based on the incoming data, resulting in more accurate traffic flow
predictions over time.
Traffic flow prediction via machine learning entails preprocessing raw traffic data,
extracting relevant features like time and weather, and training models on historical data.
Techniques such as regression, decision trees, and neural networks are employed to forecast
traffic flow levels. Real-time data integration enhances predictions, while evaluation metrics like
MAE and RMSE gauge model performance. Deep learning and graph-based models capture
complex patterns and spatial relationships, while transfer learning adapts models to different
geographical areas. Hybrid approaches combining machine learning with traditional models and
data fusion techniques enrich prediction accuracy. Ethical considerations ensure fairness,
transparency, and accountability in system development and deployment, promoting equitable
outcomes for stakeholders.

18
 Overall, this project aims to develop a comprehensive system for traffic flow prediction
and route planning, utilizing machine learning techniques to provide users with accurate and real-
time traffic information.

19
CHAPTER 5
System Design
5.1 UML diagram

Data pre-processing

Feature Extraction

Machine Learning Model

Distance Calculator

Traffic Level
Classifier

Result display

Fig 5.1 UML Diagram

• Represents the structure of the system, including input, pre-processing, feature

extraction, machine learning model, mapping system, traffic level classifier, and result
display modules.

20
5.2 Block diagram

Fig 5.2Block Diagram for traffic flow prediction

Incorporate a mapping system to input source and destination and calculate distances.
Choose a platform for deploying the application, considering factors like scalability, cost,
and ease of use. Collect real-time or historical data on vehicle counts at various locations. Clean
and pre-process the data, handling missing values, outliers, etc. Split the data into training and
testing sets to evaluate the model's performance. Develop a user-friendly interface where users
can input vehicle counts, select source and destination, and view traffic predictions and
distances. Train a model to predict traffic flow based on the vehicle counts and other features.
This could be a classification model for normal, high, and low traffic. Evaluate the performance
of your model using appropriate metrics. Incorporate a mapping system to input source and
destination and calculate distances. Choose a platform for deploying the application, considering
factors like scalability, cost, and ease of use. Collect real-time or historical data on vehicle counts
at various locations. Clean and pre-process the data, handling missing values, outliers, etc. Split
the data into training and testing sets to evaluate the model's performance. Develop a user
friendly interface where users can input vehicle counts, select source and destination, and view
traffic predictions and distances. Train a model to predict traffic flow based on the vehicle counts
and other features. This could be a classification model for normal, high, and low traffic.

21
Evaluate the performance of your model using appropriate metrics. Incorporate a mapping
system to input source and destination and calculate distances. Choose a platform for deploying
the application, considering factors like scalability, cost, and ease of use.

5.3 Architecture

Fig:5.3 Architecture for traffic flow prediction

22
Deployment: Choose a platform for deploying the application, considering factors like scalability, cost,
and ease of use.
Data Collection: Collect real-time or historical data on vehicle counts at various locations.
Data Pre-processing: Clean and pre-process the data, handling missing values, outliers, etc.
Split the Data: Split the data into training and testing sets to evaluate the model's performance.
User Interface (UI): Develop a user-friendly interface where users can input vehicle counts,
select source and destination, and view traffic predictions and distances
Machine Learning Model: Train a model to predict traffic flow based on the vehicle counts and other
features. This could be a classification model for normal, high, and low traffic.
Model Evaluation: Evaluate the performance of your model using appropriate metrics.
Integration of Mapping System: Incorporate a mapping system to input source and destination
and calculate distances.
Deployment: Choose a platform for deploying the application, considering factors like scalability, cost,
and ease of use.
Data Collection: Collect real-time or historical data on vehicle counts at various locations
Data Pre-processing: Clean and pre-process the data, handling missing values, outliers, etc.
Split the Data: Split the data into training and testing sets to evaluate the model's performance.
User Interface (UI): Develop a user-friendly interface where users can input vehicle counts,
select source and destination, and view traffic predictions and distances
Machine Learning Model: Train a model to predict traffic flow based on the vehicle counts and other
features. This could be a classification model for normal, high, and low traffic.
Model Evaluation: Evaluate the performance of your model using appropriate metrics.
Integration of Mapping System: Incorporate a mapping system to input source and destination
and calculate distances.
Deployment: Choose a platform for deploying the application, considering factors like scalability, cost,
and ease of use.

23
5.4 Model Training
In our B. Tech final project, titled "Traffic Flow Prediction Using Machine Learning," we
aim to predict traffic flow based on vehicle counts and introduce a mapping system to further
enhance our predictions.
The first phase of our project involves gathering data on the count of bikes, cars, and
trucks passing through specific locations. Using this data, our model will classify the traffic into
three categories: normal traffic, high traffic, and low traffic. By analysing the count of vehicles,
the model will learn patterns and correlations to make accurate predictions about traffic flow.
For the second phase of our project, we plan to integrate a mapping system. Users will
input the source and destination of the vehicles, and our system will calculate the distance
between these points. This additional data will provide more context for our traffic flow
predictions, allowing the model to consider factors such as distance travelled, road conditions,
and time of day.
To train our model, we will use supervised machine learning techniques. We will start by
collecting a large dataset containing information about vehicle counts, source-destination pairs,
and corresponding traffic flow conditions. This dataset will serve as the training data for our
model.
For feature engineering, we will extract relevant features such as vehicle counts, distance
between source and destination, and time of the day. These features will be fed into our machine
learning model, which will be trained using algorithms such as Decision Trees, Random Forest,
or Gradient Boosting.
Once trained, our model will be able to accurately predict traffic flow based on vehicle
counts and source-destination distances. We will evaluate the performance of our model using
metrics such as accuracy, precision, recall, and F1-score, ensuring that it meets the required
standards for real-world applications.
By combining vehicle count data with a mapping system, our project aims to provide
more accurate and comprehensive traffic flow predictions, which can be invaluable for urban
planning, transportation management, and traffic optimization.

24
5.5 Model Evaluation
Your project aims to predict traffic flow using machine learning, incorporating vehicle
count data and a mapping system. To evaluate the model, you can employ several metrics to
assess its performance accurately.
Firstly, for the vehicle count-based prediction, you can use metrics such as accuracy,
precision, recall, and F1-score to evaluate the model's performance in classifying traffic into
"normal," "high," and "low" categories. Accuracy measures the overall correctness of the model's
predictions, while precision measures the model's ability to correctly identify high traffic when it
is actually high. Recall measures the model's ability to find all the relevant cases within high
traffic, and the F1-score balances precision and recall.
Furthermore, when you integrate the mapping system to predict traffic flow based on
source and destination inputs, you can evaluate the model's performance using additional metrics
such as Mean Absolute Error (MAE) or Root Mean Squared Error (RMSE) to measure the
difference between predicted distances and actual distances. Lower MAE and RMSE values
indicate better performance.
To ensure robust evaluation, you should also consider using techniques such as cross
validation, which helps assess how the model performs on unseen data. Additionally, you could
employ techniques like hyperparameter tuning to optimize the model's performance.
By employing a comprehensive evaluation strategy, including both vehicle count-based
predictions and distance predictions from the mapping system, you can ensure that your model
accurately predicts traffic flow, providing valuable insights for traffic management and planning.
To comprehensively evaluate the machine learning model for predicting traffic flow,
metrics like accuracy, precision, recall, and F1-score assess vehicle count-based predictions.
Integrating a mapping system adds evaluation metrics like MAE and RMSE for distance
predictions. Techniques such as cross-validation ensure performance on unseen data, while
hyperparameter tuning optimizes model effectiveness. This approach ensures robust evaluation
and accurate traffic flow prediction, crucial for effective traffic management and planning

25
CHAPTER 6
Functional
Requirements
6.1 Software Requirements
6.1.1 Back-end technology
Back-end technology refers to the server-side of an application, which includes the
server, application, and database. It's the part of the application that users don't see, but it's
responsible for making sure that the front-end, or user interface, functions correctly.
There are several key components to back-end technology:
• Server: The server is the computer or system that hosts the application. It
receives requests from the front-end, processes them, and sends back the appropriate
response. Servers can range from simple single-purpose machines to complex networks
of servers, depending on the needs of the application.

• Application: The application is the software that runs on the server and
processes requests. It's responsible for things like authentication, authorization, and
business logic. Applications can be built using a variety of programming languages and
frameworks, such as Java, Python, Ruby on Rails, or Node.js.

• Database: The database is where the application stores and retrieves data.
It's where user information, application data, and other important information is stored.
There are many different types of databases, including relational databases like MySQL
and PostgreSQL, and NoSQL databases like MongoDB and Cassandra.

Back-end technology is crucial for the functioning of any web or mobile application.
Without it, the application wouldn't be able to authenticate users, retrieve data, or perform other
essential functions. It's also responsible for ensuring that the application is secure, scalable, and
efficient.

In our project, we used [insert specific back-end technologies used in your project]. These
technologies were chosen because they are well-suited to the needs of our application and
because they are widely used and well-supported within the development community.

26
Flask:
Flask is a lightweight and flexible micro web framework for Python, commonly used in
back-end development to build web applications, APIs, and web services. Its simplicity,
flexibility, and extensibility make it an excellent choice for a wide range of projects, including
route planning and traffic flow prediction using machine learning (ML).

In the context of route planning and traffic flow prediction, Flask serves as the backbone
of the application, handling requests from clients and communicating with the machine learning
models and data sources. Here's how Flask is typically used in such applications:
• API Endpoints: Flask allows developers to define API endpoints that
clients can interact with. These endpoints serve as the interface between the client-side
application (such as a web or mobile app) and the server-side application. For route
planning and traffic flow prediction, endpoints can be created to receive requests for
route information and traffic predictions.
• Integration with Machine Learning Models: Flask facilitates the
integration of machine learning models into the application. The models can be trained to
predict traffic flow based on historical traffic data, weather conditions, time of day, and
other relevant factors. Flask allows these models to be loaded into memory and used to
make predictions in real-time.
• Data Processing and Response Handling: When a client makes a request
for route information or traffic predictions, Flask handles the request, processes any
necessary data (such as user input or current traffic conditions), and sends a response
back to the client. This response may include route recommendations, estimated travel
times, traffic congestion information, and other relevant details.
• Scalability and Performance: Flask is lightweight and scalable, making it
well-suited for applications that require high performance and reliability. Flask
applications can easily be deployed on cloud platforms like AWS, Google Cloud, or
Azure, allowing them to scale to handle large numbers of requests.

27
Streamlit:
Streamlit is an open-source Python library that makes it easy to create web applications
for machine learning and data science projects. It allows developers to quickly build interactive
and customizable web interfaces for their machine learning models and data analysis tools,
without requiring knowledge of web development languages such as HTML, CSS, or JavaScript.

Streamlit is designed to be simple and easy to use, with a clean and intuitive API that
allows developers to create powerful web applications with just a few lines of Python code. It
provides a wide range of built-in widgets and components that can be used to create interactive
user interfaces, including sliders, buttons, text inputs, dropdown menus, and more. Developers
can use these widgets to control the input parameters of their machine learning models and
display the results in real-time.

In the context of traffic flow prediction using machine learning, Streamlit can be used to
create a web application that takes in input data such as traffic conditions, weather forecasts, time
of day, and historical traffic data, and uses a machine learning model to predict traffic flow and
congestion. Users can interact with the application through a simple and intuitive web interface,
adjusting input parameters and seeing the results update in real-time.

streamlit provides built-in support for popular machine learning libraries such as Scikit-
learn, TensorFlow, and PyTorch, making it easy to integrate machine learning models into web
applications. It also provides support for data visualization libraries such as Matplotlib and
Plotly, allowing developers to create interactive plots and charts to visualize the results of their
machine learning models.

Overall, Stream lit is a powerful tool for building web applications for machine learning
and data science projects. Its simplicity, ease of use, and built-in support for machine learning
and data visualization make it an excellent choice for building interactive and customizable web
interfaces for traffic flow prediction and other machine learning applications.

28
6.1.2Front end Technology HTML:
HTML (Hypertext Mark-up Language) is the standard mark-up language used to create
the structure and content of web pages. It provides the basic building blocks for web
development, defining the elements that make up a web page and how they are structured and
displayed in a web browser. Here's a detailed overview of HTML in front-end technology:

• Structure: HTML documents are structured using a system of elements, each

defined by a pair of tags. Tags are enclosed in angle brackets < >, with the
opening tag indicating the beginning of an element and the closing tag indicating
the end. For example, <p> defines a paragraph, <h1> defines a top-level heading,
and <img> defines an image.
• -0Elements and Attributes: HTML elements are used to define the structure and
content of a web page. Each element can have one or more attributes that provide
additional information about the element. For example, the <img> element has an
src attribute that specifies the URL of the image to be displayed.
• Common HTML Elements: HTML includes a wide range of elements for
structuring content, including headings <h1> to <h6>, paragraphs <p>, lists <ul>,
<ol>, and <li>, links <a>, images <img>, tables <table>, forms <form>, and
more. These elements can be combined and nested to create complex layouts and
designs.
• Semantic HTML: Semantic HTML refers to using HTML elements that clearly
describe their meaning and purpose. For example, using <header>, <nav>,
<main>, <section>, <article>, and <footer> elements to define the different parts
of a web page. Semantic HTML not only helps developers understand the
structure of a web page but also improves accessibility and SEO.
• HTML5: HTML5 is the latest version of HTML, introducing new elements,
attributes, and APIs for building modern web applications. HTML5 includes
elements for multimedia <audio> and <video>, graphics <canvas> and <svg>, as
well as new form input types and attributes.

29
6.2 Machine learning model
A machine learning model for traffic flow prediction involves the development and
training of algorithms that can learn from historical traffic data to make predictions about future
traffic conditions. These models use various machine learning techniques to analyse patterns and
relationships in the data and generate predictions based on input features such as time of day, day
of week, weather conditions, road characteristics, and historical traffic volume.

Fig 6.1: Overview of Machine Learning

Unsupervised learning, on the other hand, identifies patterns and structures in unlabelled
data, clustering similar instances or reducing dimensionality. Reinforcement learning trains
agents to interact with an environment, learning optimal actions through trial and error to
maximize cumulative rewards. Other techniques include semi-supervised learning, deep learning
with neural networks for complex pattern recognition, and transfer learning, which leverages
knowledge from one task to enhance performance in another. These approaches collectively
underpin innovations across industries, from healthcare to finance and beyond

Machine learning (ML) is a subset of artificial intelligence (AI) that provides systems
with the ability to automatically learn and improve from experience without being explicitly

30
programmed. In essence, machine learning algorithms detect patterns in data and make
intelligent decisions based on those patterns. The process begins with the algorithm learning
from historical data, identifying patterns, and then making predictions or decisions based on new
data. There are three main types of machine learning: supervised learning, unsupervised learning,
and reinforcement learning.

Fig 6.2: Types of Machine Learning

Supervised learning involves training a model on a labelled dataset, where each training
example is paired with the correct label. The model learns to make predictions by finding
patterns in the features of the data that correlate with the target label. For example, in traffic flow
prediction, the features might include historical traffic data such as time of day, day of the week,
weather conditions, and special events, while the target label is the expected traffic flow.
Unsupervised learning involves training a model on an unlabelled dataset and letting it
find patterns on its own. This type of learning is often used for clustering similar data points
together or for dimensionality reduction. In the context of traffic flow prediction, unsupervised
learning could be used to identify patterns in the data that are not immediately apparent, such as
grouping together similar traffic patterns during rush hour Reinforcement learning involves
training a model to make sequences of decisions. The model learns to achieve a goal in uncertain,

31
potentially complex environment by receiving feedback in the form of rewards or penalties. In
traffic flow prediction, reinforcement learning could be used to optimize traffic flow by adjusting
traffic signal timings based on real-time traffic conditions. Machine learning algorithms can be
further categorized into regression and classification algorithms.
Regression algorithms are used when the target variable is continuous. They predict a
continuous value based on input features. In traffic flow prediction, regression algorithms could
be used to predict the number of vehicles on the road at a given time. Classification algorithms
are used when the target variable is categorical. They predict which category or class the input
data belongs to. In the context of traffic flow prediction, classification algorithms could be used
to predict whether traffic will be light, moderate, or heavy based on the input features.
To train a machine learning model, the data is typically split into training and testing sets.
The training set is used to train the model, while the testing set is used to evaluate its
performance. Once the model is trained, it can be used to make predictions on new, unseen data.
In the case of traffic flow prediction, the trained model can take in current data on factors such as
time of day, weather conditions, and special events, and use that information to predict traffic
flow for the near future.
Before training a machine learning model, feature engineering is often performed to select
and pre-process relevant input features. In the context of traffic flow prediction, this may involve
extracting features such as time of day, day of the week, weather conditions, road infrastructure,
historical traffic patterns, and special events. These features provide valuable information for the
model to learn patterns and make accurate predictions.
Traffic flow prediction often requires considering temporal dynamics, as traffic patterns
can vary based on time of day, day of the week, and seasonality. Machine learning models for
traffic flow prediction may incorporate temporal features and use time-series forecasting
techniques to capture these dynamics and make accurate predictions.
In some cases, it is important for machine learning models to be interpretable, especially
in domains like traffic flow prediction where decisions based on model predictions may have
real-world implications. Techniques such as feature importance analysis, SHAP (Shapley
Additive Explanations) values, or model-agnostic interpretability methods can help explain the
factors influencing model predictions and enhance trust and transparency.

32
6.2.1 Types of Machine Learning Models:
There are several types of machine learning models that can be used for traffic flow
prediction, including regression models, time series forecasting models, and deep learning
models.
Regression Models:
Regression models such as Linear Regression, Support Vector Regression (SVR),
Decision Tree Regression, Random Forest Regression, and Gradient Boosting Regression are
commonly used for traffic flow prediction. These models learn patterns from historical traffic
data and make predictions based on input features.

Time Series Forecasting Models:

Time series forecasting models like Autoregressive Integrated Moving Average
(ARIMA), Seasonal ARIMA (SARIMA), and Prophet can also be used for traffic flow
prediction. These models capture temporal dependencies in the data and forecast future traffic
patterns based on historical data.
How Machine Learning Models are Used in Traffic Flow Prediction:
Machine learning models for traffic flow prediction use historical traffic data to learn
patterns and relationships between various input features and traffic flow. The data is first pre-
processed to handle missing values, outliers, and noise, and relevant features are extracted from
the data. These features are then used to train the machine learning model using a portion of the
historical data. Once trained, the model can make predictions about future traffic flow based on
new input data.
For example, a machine learning model trained on historical traffic data may learn that
traffic tends to be heavier during rush hours on weekdays, especially during bad weather
conditions. By analysing historical traffic patterns, the model can make predictions about future
traffic flow given the current time of day, day of week, and weather conditions. These
predictions can be used by traffic management systems to optimize traffic flow, reduce
congestion, and improve overall traffic efficiency.

33
6.3 System component
1. Traffic Flow Prediction Model:
Data Collection: Gather real-time data on vehicle counts using sensors or cameras installed at
different locations.
Data Pre-processing: Clean and pre-process the collected data to remove noise and
inconsistencies, and prepare it for training the machine learning model.
Training Model: Train the machine learning model using historical traffic data, including vehicle
counts and corresponding traffic flow conditions (normal, high, low).
Traffic Flow Prediction: Predict traffic flow (normal, high, low) based on the input vehicle
counts using the trained machine learning model.
2. Mapping System Integration:
Mapping API Integration: Integrate mapping APIs such as Google Maps or OpenStreetMap to
provide distance information between the source and destination of vehicles.
User Interface: Develop a user-friendly interface for users to input the source and destination of
vehicles.
Distance Calculation: Utilize the mapping API to calculate the distance between the source and
destination provided by the user.
Real-time Traffic Information: Display real-time traffic information along with the predicted
traffic flow based on the input vehicle counts and calculated distance.
Additionally, your system may include a database to store historical traffic data, a server to host
the machine learning model and handle user requests, and a front-end interface for users to
interact with the system easily.
By combining machine learning for traffic flow prediction and mapping system integration for
distance calculation, your project aims to provide accurate and real-time information about traffic
conditions to help users plan their routes efficiently.

34
6.4 Mapping System

Mapping System Module:

The Mapping System module is designed to assist users in planning their travel routes
effectively by providing them with accurate distance calculations between their source and
destination locations.

Fig 6.4: Route Mapping

35
Features:
1. Location Input:
Users are provided with input fields where they can enter their source and destination
locations. The input fields are designed to accept text input, allowing users to easily specify their
desired locations.

2. Distance Calculation:
Once users have entered their source and destination locations, they can click on the
"Calculate Distance" button. The system utilizes advanced mapping algorithms to calculate the
shortest distance between the two points. Distance calculation considers various factors such as
road conditions, traffic, and available routes.

3. Interactive Map Display:

Upon clicking the "Calculate Distance" button, the system displays the calculated
distance between the source and destination locations. The distance is displayed in both
kilometres and miles for user convenience. Additionally, an interactive map is displayed,
showing the recommended route between the two points. Users can zoom in and out of the map,
as well as pan to explore the route in more detail. The map interface also highlights important
landmarks and points of interest along the route.

User Interface Design:

The user interface of the Mapping System module is designed to be intuitive and user
friendly, ensuring a seamless experience for users:
1. Input Interface:
Users are presented with clear and easy-to-use input fields for entering their source and
destination locations. Text prompts provide guidance to users, helping them input their locations
accurately.
2. Distance Calculation:
Once users have entered their locations, they can easily initiate the distance calculation
process by clicking the "Calculate Distance" button. The system provides real-time feedback,
indicating that the distance calculation is in progress.

36
3. Interactive Map Display:
After the distance calculation is complete, the system displays the calculated distance
between the source and destination locations. An interactive map is presented alongside the
distance information, allowing users to visualize their travel route.
The map interface is dynamic and responsive, enabling users to explore the route in detail and
plan their journey effectively.
Benefits:

1.Efficient Route Planning: By providing users with accurate distance calculations and a
recommended travel route, the Mapping System module helps users plan their journeys more
efficiently.
2.Real-Time Updates: The system considers real-time factors such as traffic conditions and road
closures, ensuring that users receive the most up-to-date route information.
3.Improved User Experience: With its intuitive user interface and interactive map display, the
Mapping System module offers users a seamless and enjoyable experience, making travel
planning easy and convenient.
4.This should give a more detailed explanation of how the Mapping System module functions
and its benefits to the user. Let me know if you need further modifications!

37
 CHAPTER 7
Implementation
7.1 Code Explanation

Libraries:
These lines of code import the necessary libraries, such as pickle and Flask, load a
machine learning model (model.pkl), and set up a Flask web application. The home () function
defines the route for the home page of the web application.

38
Form submission for prediction :

T, D, W, C, B, Bu, Tr, to: These variables hold the values submitted via the form, such as
time, date, week, number of cars, bikes, buses, trucks, and total count, respectively. These values
will likely be used for prediction or further processing.
This code essentially handles form submissions from the /predict route, extracting the form
data and preparing it for further processing or prediction.

Interpreting Traffic Predictions: Converting Numerical Results into Human-Readable

Descriptions.

This code snippet appears to be part of a classification or prediction process where result
is a list or array containing the predicted traffic level for a particular day. Here's an explanation
of what each part of the code does:
• result [0]: Accesses the first element of the result list, which presumably
contains the predicted traffic level.
• The code then checks the value of result [0] to determine the predicted traffic
level:
• If result [0] is equal to 0, it indicates "Low Traffic Today".
• If result [0] is equal to 1, it indicates "Normal Traffic Today".
• If result [0] is equal to 2, it indicates "High Traffic Today".
39
 • If result [0] does not match any of the above cases, it falls into the else block,
indicating "Heavy Traffic Today".
• The corresponding traffic level string is assigned to the variable traffic based
on the value of result [0].
This code essentially translates the numerical prediction result into human-readable
traffic level descriptions, making it easier for users or systems to understand and interpret the
predicted traffic level for the day.

Result:

Finally, the if _name_ == '_main_': block ensures that the Flask app is only run if the
script is executed directly (not imported as a module). It starts the Flask application with
debugging enabled.
Overall, this code sets up a simple Flask web application that takes input from a form,
passes it to a machine learning model for prediction, and then displays the predicted traffic level
on a result page.

This code is a simple Flask application that serves as a basic registration and login system
with MongoDB as the database backend. Let's break it down step by step:

40
 1. Importing Libraries:
- Flask: A web framework for Python.
- request: To handle HTTP requests.
- jsonify: To serialize data into JSON format.
- render template: To render HTML templates.
- session: To manage user sessions.
- redirect: To redirect users to different URLs.
- Mongo Client: MongoDB client for Python.
- bison: To handle BSON data format.
- CORS: Cross-Origin Resource Sharing extension for Flask, allowing
cross origin requests.

2. Setting up MongoDB:
- Connects to a MongoDB database named "Route" on a remote server
using Mongo Client.

3. Initializing the Flask App:

Creates a Flask application instance.

Sets a secret key for session management.

Enables CORS for handling cross-origin requests.

4. Routing:
`/`: Renders the homepage. If a user is logged in, it fetches orders from the database where
`assign` field is empty and renders the homepage template with those orders.
- `/log`: Renders the login page.
- `/reg`: Renders the registration page.

5. Registration (`/register` route):

- Handles both GET and POST requests.
- Retrieves user registration data from the request form.

41
 - Checks if the username already exists in the database. If it does, it renders the
registration page with a message indicating that the username already exists.
- Checks if the email already exists in the database. If it does, it renders the
registration page with a message indicating that the email already exists.
- Checks if the passwords match. If they don't, it renders the registration page
with a message indicating that the passwords do not match.
- Checks if the password length is at least 8 characters. If it's not, it renders the
registration page with a message indicating that the password must be greater than 7
characters.
- If all checks pass, it inserts the user data into the database and renders the
registration page with a message indicating successful registration.

6. Rendering Templates:
- Uses Jinja2 templating engine to render HTML templates.
- Passes data to templates for dynamic rendering (e.g., passing orders
to the homepage template).

7. Error Handling:
- The code does not have extensive error handling. It primarily checks
for duplicate usernames and emails, mismatched passwords, and password length.
Other errors are not explicitly handled.

8. Session Management:
- Utilizes Flask's session management for user authentication.
However, the code doesn't have a login mechanism implemented yet. It would
typically involve setting a session variable upon successful login and checking it to
determine whether a user is logged in or not.

In summary, this code sets up a basic Flask application with registration functionality,
utilizing MongoDB as the database backend. It lacks a login mechanism and more robust
error handling, but it provides a foundation for building a simple web application with user
registration.
42
 This code extends the functionality of the previous Flask application to include user login
and an API endpoint for processing data related to route assignments. Let's delve deeper into its
functionality:
1. User Login (`/login` route):
Handles both GET and POST requests.

Checks if a user is already logged in by accessing the session data. If so, it redirects the
user to the homepage.
If no user is logged in, it retrieves the username and password from the request form. It
queries the MongoDB database to find a user with the provided username. If the user exists and
the password matches, it sets the user's session and redirects them to the homepage.
If the user does not exist or the password is incorrect, it renders the login page with a
message indicating the issue.
Route Assignment (`/assign` route):
Handles POST requests only and is decorated with `@cross origin () ` to enable cross
origin requests.
43
 Retrieves data from the form submitted by the client, which typically contains order IDs.
Iterates over the form data to retrieve information about each order from the MongoDB
database.
Constructs a list of locations (latitude and longitude pairs) based on the order data.

Constructs a request payload containing the locations data in a specific format expected by the Open
Route Service API.
Sends a POST request to the Open Route Service API endpoint

(`https://api.openrouteservice.org/v2/matrix/driving-car`) to calculate travel durations

between the provided locations.
The API request includes necessary headers such as authorization and content type.

Once the API call is made, it receives a response containing the travel duration matrix.
The code extracts the durations from the API response and stores them in the `matrix`
dictionary for further processing.

2.Error Handling:

The login route handles basic error scenarios such as incorrect username or password.
However, it could be improved with more specific error messages for better user feedback.
The route assignment endpoint does not include extensive error handling. Errors from the
API request or response are not explicitly managed.

3.Data Processing:
The code primarily focuses on handling HTTP requests, interacting with the MongoDB
database, and making API calls. Data processing involves retrieving order information from the
database and constructing payloads for API requests.

4.External API Integration:

Utilizes the OpenRouteService API to calculate travel durations between given locations.
It requires an API key for authorization, which is included in the request headers.

44
 In summary, this code enhances the previous Flask application by adding user login
functionality and an API endpoint for route assignment processing. It demonstrate show to
handle user authentication, interact with a MongoDB database, and integrate with external APIs
for additional functionality such as route optimization. However, it could benefit from more
robust error handling and security measures, such as input validation and error responses.

The provided code implements the Traveling Salesman Problem (TSP) solution to find the best
route among a set of locations. Here's a breakdown of the TSP implementation and how it fits
into the Flask application:

1. TSP Function Definition (`tsp`):

- The `tsp` function defines a recursive helper function `tsp_helper` to solve the TSP
problem.
- It takes a `matrix` parameter representing the distance matrix between locations. -
The `matrix` is a dictionary where keys represent locations and values represent distances to
other locations.
- The TSP algorithm aims to find the shortest route that visits each location exactly
once and returns to the starting location.
2. Recursive TSP Solver(`tsp_helper`):
45
 - The `tsp_helper` function recursively explores all possible permutations of
routes.
- It maintains `cur` as the current location, `remaining` as the set of locations
yet to be visited, `route` as the current route, and `distance` as the cumulative distance
travelled.
- At each step, it chooses the next location (`next node`) from the remaining
locations and explores all possible routes.
- When no locations remain (`not remaining`), it checks if the current route is
the best solution found so far. If so, it updates `best route` and `min_distance`.

3. Finding the Best Route:

- The `tsp_helper` function is invoked initially with the starting node
(`nodes[0]`), all remaining nodes (`all_nodes - {nodes[0]}`), the current route
`[0]`, and zero distance.
- As the function progresses recursively, it explores all possible routes
and updates `best_route` and `min_distance` whenever a better route is found.

4. Indexing and Constructing the Route:

- After finding the best route, the code constructs the actual route by indexing
into the `routes` list using the indices from the `best_route`.
- The resulting `res` list contains the locations in the optimal order of traversal.

5. Rendering the Template:

- Finally, the code renders the `index.html` template, passing the optimal route
(`res`) as context data.
- The HTML template can then display the optimal route to the user.

6. Output and Display:

- The optimal route is printed to the console for debugging purposes.
- The route is also displayed to the user through the rendered HTML template,
enhancing the user experience.

46
 In summary, this section of the Flask application integrates a TSP solver to
determine the best route among a set of locations. It demonstrates the application of
algorithms in real-world scenarios and enriches the functionality of the web application
by providing users with optimized routes.

This Flask application focuses on handling order-related functionalities, including creating,

deleting, and displaying orders. Let's discuss each route and its functionality:
1. Order Creation (`/order` route):
- This route is accessed when a user submits a new order
through a form or an API call.
- It retrieves the food item (`food`) and location (`lat`, `lng`)
from the request arguments.
- The route then inserts a new document into the MongoDB
collection `orders` containing information about the order, such as the food
item, latitude, longitude, and assignment status.
- After inserting the order, it redirects the user to the homepage
(`/`) to display the updated list of orders.

2. Order Deletion (`/delO` route):

- This route handles requests to delete an existing order.
- It receives the ID of the order to be deleted (`id`) from the
request arguments.
- Using the `delete one` method, it removes the order document
from the `orders` collection based on the provided ID.
- After deleting the order, it redirects the user to the homepage
(`/`) to reflect the changes in the order list.

3. Order Creation Form (`/createO` route):

- This route renders the `order.html` template, allowing users to
create new orders through a form.

47
 - The template typically includes input fields for the food item
and location, along with a submit button.
- Upon submitting the form, the order creation process is
initiated, redirecting the user to the homepage.

4. Main Application Execution (`if __name__ == '__main__’: `):

- This conditional block ensures that the Flask application is
executed only when the script is run directly, not when it's imported as
a module in another script.
- It runs the Flask application on port 5500, allowing users to
access the web application through a web browser.
In summary, this section of the Flask application manages the CRUD (Create,
Read, Update, Delete) operations related to orders. It provides functionalities for users to
create new orders, delete existing orders, and view the order creation form. By integrating
these functionalities with the MongoDB database, it enables users to interact with the
application dynamically, creating a seamless experience for managing orders.

7.2 Results

User Interface and Traffic conditions

• Designing a simple and intuitive UI where users can enter vehicle counts, select
their source and destination, and receive traffic predictions. Displaying the
predictions (normal, low, or high traffic) in a user friendly format. Incorporating
feedback mechanisms to improve model accuracy and user experience.

Login of candidate
• Visit the login page.
• Fill the username and password.
• directed into the Routing page.

48
Mapping system to input source and destination of vehicle
• Enter the source and destination locations for your vehicle, Receive the optimal route
and distance.
• Implementing mapping services (like Google Maps or OpenStreetMap) for distance
calculations and routing.
• Developing a feature to input starting and ending points and calculate the distance.

Route planning and Distance estimation

• Next will be After logging in with your username and password, you will be direct to
the routing page where you can view the route to your destination along with the
estimated distance.

49
Fig 7.1: User Interface

50
Fig 7.2: Authenticate the traffic conditions (low, normal, high)

51
 Fig 7.3: Indicate the login page of the candidate.

Fig 7.4: Mapping system to input the source and destination of vehicle

52
Fig 7.5 Route planning and distance estimation.

53
CHAPTER 8

Conclusion
8.1 Conclusion

1. It has been concluded that we develop the traffic flow prediction

system by using a machine learning algorithm by using regression model, the
prediction is done.
2. The public gets the benefits such as the current situation of the traffic
flow, they can also check what will be the flow of traffic on the right after one hour of
the situation and they can also know how the roads are as they can know mean of the
vehicles passing through a particular junction that is 4 here.
3. The forecasting or the prediction can help people or the users in
judging the road traffic easier beforehand and even they can decide which way to go
using their navigator and also this will prediction be also helpful.
4. With accurate traffic flow predictions, users can plan their routes more
efficiently, avoiding congested areas and minimizing travel time. This can lead to fuel
savings and reduced environmental impact by optimizing the use of transportation
resources.
5. By integrating real-time data into the traffic flow prediction system,
users can receive instant updates on changing traffic conditions. This allows for
dynamic route adjustments while en route, ensuring a smoother and more hassle-free
travel experience.
6. Emergency services can benefit from traffic flow predictions by
efficiently navigating through traffic to reach their destinations faster. This can be
crucial in situations where every minute counts, such as medical emergencies or
disaster response.
7. Traffic flow data collected over time can provide valuable insights for
city planners and policymakers. By analysing traffic patterns, authorities can make
informed decisions regarding road expansions, infrastructure development, and urban
planning to improve overall traffic management and reduce congestion.
8. Businesses that rely on transportation for goods delivery or service
provision can benefit from traffic flow predictions by optimizing their logistics
operations. This can lead to cost savings, improved delivery times, and enhanced
customer satisfaction

54
8.2 Future enhancement
1. Future Enhancements1. In the future, the system are often further
improved using more factors that affect traffic management using other methods like
deep learning, artificial neural network, and even big data.
2. The users can then use this technique to seek out which route would be
easiest to achieve on destination. The system can help in suggesting the users with
their choice of search and also it can help to find the simplest choice where traffic isn't
in any crowded environment.
3. These days, traffic prediction is extremely necessary for pretty much a
part of the state and also worldwide. So, this method of prediction would be helpful in
predicting the traffic before and beforehand. For better congestion prediction, the
grade and accuracy are prominent in traffic prediction.
4. Future traffic prediction systems could integrate multiple modes of
transportation, including not just cars but also public transit, bicycles, and pedestrians.
This holistic approach to mobility management would provide users with
comprehensive route options tailored to their specific needs and preferences,
promoting sustainable and efficient urban mobility.
5. Leveraging real-time traffic data, future systems could implement
dynamic pricing and incentive schemes to manage congestion more effectively. By
adjusting toll rates, parking fees, and incentives for alternative transportation modes
based on current traffic conditions, these systems can encourage behaviour changes
that help alleviate congestion and improve overall traffic flow.
6. Incorporating crowdsourced data from mobile apps, social media, and
connected vehicles can enhance the accuracy and timeliness of traffic predictions. By
tapping into the collective intelligence of users, these systems can gather real-time
traffic flow, providing more reliable and up-to-date predictions.
7. Future enhancements should prioritize user-centric design principles to
ensure that traffic prediction systems are intuitive, accessible, and inclusive for all
users. This may involve developing user-friendly interfaces, providing personalized
recommendations based on individual preferences and constraints, and fostering
community engagement and feedback mechanisms to empower users to contribute to
the improvement of the system.
55
8.3 Learning outcome
1.Learns about how really Traffic is generated on busy roads.
2.Larn about how to avoid being stuck in traffic.
3.Learn about how to control traffic.
Understanding Traffic Generation: Gain insights into the factors contributing
to traffic congestion on busy roads, including traffic volume, road capacity, traffic
signal timings, road geometry, and driver behaviour. Learn how these factors interact
to influence traffic flow dynamics and congestion patterns.
Strategies for Avoiding Traffic: Explore various strategies and techniques for
avoiding getting stuck in traffic, including alternative route planning, real-time traffic
updates through navigation apps, adjusting travel times to avoid peak hours, and
utilizing public transportation options. Understand the importance of proactive
planning and decision-making to minimize travel delays.
Traffic Control Techniques: Learn about traffic control methods and
technologies used to manage and regulate traffic flow, including traffic signals,
signage, lane markings, traffic calming measures, ramp metering, and intelligent
transportation systems (ITS). Understand the principles of traffic engineering and how
these techniques are applied to optimize traffic flow and improve safety on roadways.
Effective Communication: Develop skills in effective communication and
collaboration with transportation authorities, urban planners, and other stakeholders to
implement traffic control measures and infrastructure improvements aimed at
reducing congestion and enhancing mobility. Understand the importance of data
driven decision-making and stakeholder engagement in developing sustainable traffic
management solutions.
Analytical Skills: Enhance analytical skills in analysing traffic data,
identifying traffic patterns and trends, and diagnosing traffic congestion problems.
identify areas of improvement and formulate effective traffic management strategies.
Public Policy Implications: Gain awareness of the broader public policy
implications of traffic congestion, including economic impacts, environmental
consequences, equity considerations, and social welfare issues. Explore policy
interventions and regulatory measures aimed at addressing traffic congestion and
promoting sustainable transportation practices.
56
 Technological Innovations: Stay abreast of technological innovations and
advancements in transportation technology, including connected and autonomous
vehicles (CAVs), smart traffic management systems, and predictive analytics tools.
Understand how these technologies are shaping the future of transportation and
influencing traffic management strategies.
Continuous Learning and Adaptation: Recognize the dynamic nature of traffic
congestion and the need for continuous learning and adaptation in developing
effective traffic management solutions. Develop a mindset of curiosity, innovation,
and lifelong learning to stay informed about emerging trends and best practices in
traffic engineering and management.

57
CHAPTER 9
Reference
9. References
1.F. Tang, Y. Kawamoto, N. Kato, and J. Liu, “Future intelligent and secure
vehicular network toward 6G: machine-learning approaches,” Proceedings another
IEEE, vol. 108, no. 2, pp. 292 – 307, 2020 View at: Publisher Site | Google Scholar0

2. Y. Ni, L. Cai, J. He et al., “Toward reliable and scalable internet of vehicles

performance analysis and resource management,” Proceedings of the IEEE, vol. 108,
no. 2, pp. 324 – 340, 2020. View at: Publisher Site | Google Scholar

3.H. Zhou, W. Xu, J. Chen, and W. Wang, “Evolutionary V2X technologies

toward the internet of vehicles: challenges and opportunities,” Proceedings other
IEEE, vol. 108, no. 2, pp. 308 – 323, 2020.View at: Publisher Site | Google Scholar.

4.Dijkstra, E.W. "A note on two problems in connexion with graphs."

Numerische Mathematic 1, no. 1 (1959): 269–271. DOI: 10.1007/BF01386390
Bellman, R. "On a routing problem." Quarterly of Applied Mathematics 16, no. 1
(1958): 87–90. DOI: 10.1090/qam/1027

5.Zheng, Y., Liu, L., Wang, L., and Liu, Y. "Forecasting Urban Traffic Flow:
A Deep Learning Approach." IEEE Transactions on Intelligent Transportation
Systems 19, no. 3 (2018): 720-729. DOI: 10.1109/TITS.2017.2694959

6.Zhang, Z., Cui, S., and Zhu, Q. "Predicting Short-Term Traffic Flow with
LSTM Recurrent Neural Networks." Transportation Research Part C: Emerging
Technologies 54 (2015): 187-197. DOI: 10.1016/j.trc.2015.03.022
7.U.S. Department of Transportation. "Intelligent Transportation Systems
Data." Website
OpenStreetMap. "OpenStreetMap Data." Website

58
 8.Liao, W., Zheng, Z., and Xie, X. "A Novel Evaluation Metric for Short Term
Traffic Flow Prediction." IEEE Transactions on Intelligent Transportation Systems
20, no. 3 (2019): 1086-1096. DOI: 10.1109/TITS.2018.2830760

59