WOLAITA SODO UNIVERSITY

GRADUATE STUDIES DIRECTORATE

Research Proposal on
Time series crime prediction analysis using machine learning case of
Wolaita Sodo city police department.

By:
Wondimagegn Desta
SCHOOL: School of Informatics
DEPARTMENT: Information Technology
PROGRAM: Regular
ADVISOR: Siraj Sibhatu (Ph.D.)

Date: April/5/ 2024

Wolaita Sodo, Ethiopi
Approval sheet

Submitted by:
Wondimagegn Desta _____________ __________________
Name of student Signature Date
Approved by:
1. Siraj Sibhatu (PhD) _____________ __________________
Name of Advisor Signature Date
2. ____________________ _____________ __________________
Name of Examiner Signature Date
3. _____________________ _____________ __________________
Name of Chairman, DGC Signature Date
4. _________________ _____________ __________________
Name of PG Coordinator Signature Date
5. _____________________ _____________ __________________
Name of Director, GSD Signature Date

i
Table of Contents
Approval sheet ............................................................................................................................................... i
Table of Contents .......................................................................................................................................... ii
List Of Table ................................................................................................................................................ iii
List Of Figure............................................................................................................................................... iv
List of Abbreviation ....................................................................................................................................... v
Executive Summary...................................................................................................................................... vi
1. Introduction .......................................................................................................................................... 1
1.1. Background ................................................................................................................................... 1
1.2. Motivation Of Study ...................................................................................................................... 3
1.3. Statement of the Problem ............................................................................................................ 4
1.4. Research Question ........................................................................................................................ 5
1.5. Objective ............................................................................................................................................ 5
1.5.1 General Objective ........................................................................................................................ 5
1.5.2. Specific Objective ........................................................................................................................ 5
1.6. Significance of the Study ............................................................................................................... 5
1.7. Scope Of The Study: ...................................................................................................................... 7
2. Research Methodology ......................................................................................................................... 8
2.1 proposed architecture of machine learning model .......................................................................... 11
3 Related Works And Literature Reviews. .................................................................................................. 11
3.1. Cost Break Down And Time Table .................................................................................................... 14
3.1.1. Cost Break Down ....................................................................................................................... 14
3.1.2. Time Table ................................................................................................................................. 15
Reference .................................................................................................................................................... 16

ii
List Of Table
Table 1. Related work ................................................................................. Error! Bookmark not defined.
Table 2. Cost break down ........................................................................................................................... 14
Table 3. time table....................................................................................................................................... 15

iii
List Of Figure
Figure 2g. 1. System Architecture ............................................................................................................ 11

iv
List of Abbreviation
ML-Machine learning
SVM-Support Vector Mchine
ARIMA- AutoRegressive Integrated Moving Average
MAE-Mean Absolute Error
MSE-Mean Squared Error
RMSE-Root Mean Squared Error
R2-R-squared
MAPE-Mean Absolute Percentage Error
AIC-Akaike Information Criterion
BIC-Bayesian Information Criterion

v
Executive Summary
This is a research proposal for carrying out a time series crime prediction analysis using machine
learning techniques at Wolaita Sodo City Police Headquarters. The research work will thus be
directed at the analysis of historical criminal activities and their temporal patterns and trends, as
well as the development of a predictive model that will enable crime prediction.
The main aim of the research is to improve the efficiency of law enforcement agencies in
Wolaita Sodo City through accurate prediction of criminal activities and better resource
allocation. The current study will apply machine learning algorithms in an attempt to produce
insights that bear value for policy-making and proactive strategies in crime prevention.
The proposed methodology is for gathering the Wolaita Sodo City Police Department's past
records of criminal events, which will contain details about the type of crime, where it occurred,
date, and time. The preparation of the data will be done to address outliers and discrepancies. In
order to create the prediction model, multiple machine learning approaches, including ARIMA,
random forests, and neural SVM, will be trained and evaluated.
To assess the performance of the model, appropriate metrics will be used. The research will
follow the ethical guidelines for protecting the confidentiality and anonymity of participants' data
and obtaining all the required permissions and approvals from the police department.
The research will confine itself to an analysis of historical crime data and the development of a
predictive model through the use of machine learning techniques. The findings will provide
insights into crime patterns and aid in resource allocation and proactive crime prevention
measures for law enforcement agencies in Wolaita, Sodo City.
These findings have big implications in that accurate crime prediction will lead to better law
enforcement strategies, improved public safety, and better allocation of resources. This research
study will help in the domain of crime analysis and provide valuable inputs to policymakers and
law enforcement agencies, not only for Wolaita Sodo City but for similar contexts around the
globe.
In summary, this research proposal wants to respond to the necessity of effective crime
prediction in Wolaita Sodo City by applying a time-series analysis approach with machine
learning techniques. Analyses of historical crimes and developing a predictive model that can
help law enforcement agencies gain valuable insights and practical implications for improving
public safety and crime prevention strategies.

vi
 1. Introduction
1.1. Background
Any violent or illegal conduct directed by a perpetrator against any other person with the
potential to cause harm or damage to property is illegal and subject to punishment under the laws
of the country in which the crime is committed.[1] There are several different categories of
crime, including crimes against property (theft, burglary, and robbery) and crimes of
aggression.[2] Traditionally, statistical methods comparable to those used for forecasting other
aspects of daily life have been used to anticipate crimes.[3]
Governments and law enforcement have access to a wealth of records that may be used to lower
the rate of crime. Crime pattern theory states that rather than venturing into new terrain,
criminals would prefer to commit violent, opportunistic crimes by taking advantage of well-
known sites they have already visited. Furthermore, there is no uniform or random distribution of
crimes within a town or its surrounding areas. A hotspot is defined as a location or area where
illegal activities may occur. In light of this data, it is imperative to emphasize that mapping crime
hotspots can aid researchers in understanding the causes of persistent crimes in a certain
location.[1]
The Wolaita Sodo City Police Department is an excellent law enforcement organization
dedicated to protecting and safeguarding its inhabitants. In the prosperous Ethiopian city of
Wolaita Sodo, it is situated. A growing number of people live in the community, which increases
the need for efficient crime prevention measures. To fulfill this requirement, I suggest reading
"Time Series Crime Prediction Analysis Using Machine Learning: A Case Study of the Wolaita
Sodo City Police Department."
Through the Wolaita Sodo City Police Department, a wealth of historical crime data has been
gathered, including information on various crime classes, dates, hours, and places. The
invaluable insights in this information, when used correctly, can help identify trends, styles, and
potential hotspots. Our goal is to use device studies to build a robust time-collection crime
prediction model that will allow the agency to take preemptive steps to prevent crimes from
happening.
Machine learning algorithms have been developed based on more than just statistics to uncover
hidden patterns and produce original forecasts. I need to improve the branch's ability to

1
anticipate and respond to potential fraud situations by applying those algorithms to Wolaita Sodo
City's specific situation.
By developing a deep understanding of the temporal and spatial dynamics of crime in Wolaita
Sodo, I can create a predictive framework that takes into account factors like seasonality, time of
day, day of the week, and geographic influences. With the aid of this framework, the Wolaita
Sodo City Police Department may be able to successfully hire personnel, strategically allocate
resources, and implement targeted crime prevention efforts.
The effective application of the periodic collection crime prediction version could severely
impair the city's security and safety. It will enable the agency to employ data-driven insights to
effectively combat crime, enhance public safety, and foster network consideration. Finally, the
goal of this study is to provide the Wolaita Sodo City Police Department with a useful tool to
enable it to carry out its mandate of protecting the well-being and tranquility of the city's
residents.
Crime is a prevalent social issue that has an impact on people's quality of life. As crime rates
rise, it becomes more important to develop a model that can predict how many crimes may be
committed in a given time frame, identify the traits of a potential criminal, and pinpoint potential
crime scenes.[4].
Living in a secure and crime-free environment is essential to human survival and cannot be
avoided. Particularly, living in an urban environment increases the possibility of safety concerns
because of things like street crimes. Studies show that locals are worried about street crimes like
snatch-and-run offenses, which tend to be concentrated[3].
For classification and prediction tasks in a variety of packages, support vector machines (SVMs)
and other device learning techniques are frequently used. Crime prediction is the primary
application of Support Vector Machines (SVMs), as they can recognize the high-quality
reparability within training inside datasets. SVMs can examine relevant spatial attributes and
historical crime data to find patterns and relationships that aid in predicting crime hotspots.
One component of the task may involve the analysis of spatial facts, which include
characteristics like socioeconomic prominence, historical crime data, and demography. The
algorithm might potentially be trained using a dataset that includes information on previous
criminal episodes and the locations associated with them. Second, the approach might enable
new places to predict crime hotspots.

2
A powerful system for predicting crimes is required in place of traditional crime analysis because
traditional methods cannot be applied when crime data is high dimensional and complex queries
are to be processed. Therefore a crime prediction and analysis tool were needed for identifying
crime patterns effectively.[5]
We expect that our research will add to the body of knowledge presently available on device
learning techniques for crime prediction. The results of the study will provide insight into how
well the Model of Machine Learning anticipates crime hotspots and could be used in the future to
improve public safety measures.

1.2. Motivation Of Study

The urgent need to address the crime-related issues in Wolaita Sodo city and assist the local
police department in their efforts to lower crime led to the drive for this investigation. Crime is a
multifaceted, intricate problem that affects people's safety and well-being on an individual and
community level. my objective is to apply machine learning and time series analysis to give the
Wolaita Sodo City Police Department a proactive tool for anticipating criminal situations.
Giving law enforcement agencies access to cutting-edge analytical tools that enable them to
consistently anticipate crime patterns is the main goal. By using system learning algorithms and
scrutinizing historical crime data, I developed the ability to identify latent patterns, associations,
and inclinations inside the data that might have remained obscured through the use of
conventional tactics. By utilizing this data, the police department will be able to more precisely
allocate resources, plan patrols, and implement targeted crime prevention strategies.
Enhancing the trendy safety and security of Wolaita Sodo city is another driving factor. Our goal
is to develop an accurate time series crime prediction version that will provide the police
department with useful data to reduce the likelihood of crimes and proactively target potential
crime hotspots. This preventive approach can also improve community confidence, make the
ecosystem safer, and provide both residents and visitors with a sense of security.
Furthermore, this study attempts to contribute to the growing field of system learning packages
in law enforcement and criminal forecasting. Utilizing a case study that is exclusively focused on
Wolaita Sodo city, I am prepared to create findings and suggestions that may be specific to the
specific functions and obstacles of the area as a whole. The findings from this study may serve as
the basis for future tests, rules, and the use of information-driven methods in crime prevention
operations.

3
 1.3. Statement of the Problem
Crime has an impact on human development, economic progress, and quality of life. The trend
indicates that there has been a periodic increase in the crime rate. With the resources that police
stations currently have at their disposal, it is challenging to conduct an effective analysis given
the volume of crime scenes, the coverage of crimes, and the complexity of crimes. All
governments should give serious consideration to combating crime Crime can be reduced by
prediction and taking preventive actions. Understanding the crime domain and having prior
knowledge for the crime data will have a big view for crime analysis.[6]
The Wolaita Sodo City Police Department found it extremely difficult to predict criminal activity
and put an end to it before it puts at risk the safety and well-being of the community. The present
methods for crime analysis cannot be able to precisely recognize patterns or forecast the
development of new crimes, even with hardworking applications. The department considers it
harder to decide on resources and begin implementing proactive measures to prevent crime as a
consequence of this limitation.
As the majority of studies on the issue have been carried out in developed nations, understanding
is absent in the field of time series crook prediction a.pplying technology studying, which is
particularly relevant to Ethiopia, including cities like Wolaita Sodo. The absence of predictive
patterns and specific investigations by the branch limits its capacity to fight crime effectively.
There is not much research on predicting crimes over time in Wolaita Sodo City, but it's a big
deal. this place is different from others, so we need to make our way of guessing what might
happen next, based on what's going on here and what the people here think. The model was then
developed and examined, and it was utilized to train the algorithm and sample the data set.[7]
Our objective is to enhance and boost the efficacy of Wolaita Sodo's crime prediction systems
via the implementation of advanced computer programs and past crime data. This study will
allow the department to use resources that will avoid crime before it occurs and will help us
discover more about crime trends. Guiding the people of Wolaita Sodo to reside in a safer and
more secure society is the main objective we have.

4
 1.4. Research Question
 How to gather and get ready previous criminal activity information for feature
engineering-based crime prediction in the future?
 Which machine learning algorithms are most successful at predicting criminal behavior
over time, and how may their accuracy be raised?
 How well do the top outperforming machine learning models perform in comparison to
each other when we assess their effectiveness?
 Which Metrics is suitable to evaluate the developed models?

1.5. Objective
1.5.1 General Objective
Developing a machine learning model that can anticipate criminal activity within the Wolaita
Sodo City Police Department is the Main objective of this research.

1.5.2. Specific Objective

The main aim of this study is to use machine learning to predict crimes in the Wolaita Sodo City
Police Department in Ethiopia. the aim is to reach the following specific targets:
• To gather and prepare historical crime data: By using feature engineering.
• To determine which machine learning models are most effective in forecasting crime in
time series.
• To develop a machine learning models to make accurate predictions about crime in the
future.
• To analyze the state of Arts of machine learning with time series criminal prediction
model.
• To evaluate the developed model accuracy of the system for time series criminal
prediction.

1.6. Significance of the Study

This clarifies The definition of a study problem and the justification of the value of a solution are
done in this first stage of the design science process. It is employed in the creation of a solution
artifact. Having in-depth knowledge of the field is beneficial for identifying the issue. We have
studied several academic works and examined our daily routines to gain insight into the current
situation and the significance of the fix. Studies that have been conducted to forecast crimes in
order to stop them before they happen.[6]

5
The Wolaita Sodo Police Department's use of machine learning to anticipate crime has several
positive consequences for our community. Let's investigate them in a way that highlights the
similarities between us.
The most important aspect is how good we are at accurately predicting crimes, which keeps us
safe and discourages crime. By forecasting when and where crimes could happen, our police
officers may develop planning and take necessary steps to keep the community secure. By
determining the times and places when a crime is most likely to occur, our law enforcement
personnel will be prepared to prevent crime from occurring and keep our streets safer. I can
establish an environment of safety where everyone is comfortable if we collaborate.
And reality, there is yet another key advantage: planning for resources management. Since we
don't have many resources, we need to make sure all of our police officers can do their duties
professionally. I may place our police on the most likely occasions and places by utilizing
machine learning to identify where crimes are more likely to take place. In simple terms, it will
additionally boost our productivity but also make sure that we're making the most effective use
of our resources, helping us to do better for the community.
People of Wolaita Sodo: through careful study, I have a chance to understand trouble in new
ways. By tracking what happens over the years, this research will help us see repeating problems.
When we know the roots that keep causing fear and hurt, real change can grow.
This work lets our leaders make wiser rules and plans. If forecasts clearly show us when and
where issues tend to begin, answers may come. The studies help uncover what truly drives
distress in our streets, stores, and homes.
With open hands and hearts, a peaceful tomorrow starts. If we make space to understand each
experience, wisdom emerges about how to lift each other. As one person stands strong beside
one another, our town becomes what we all wish to see. This research plants those first seeds.
Now, what bright blooms may we nurture from the soil of good will and care for our neighbors.
There is light ahead, my friends if we walk as one.
Because of the ability to forecast crimes over time, better inquiries and reactions toward
incidences are inevitable. I can distribute investigative resources in a better way and respond
quickly as soon as I have an insight into criminal incidents. By doing so, I can increase our
chances of catching offenders and preventing further crimes which will then give victims and

6
their families some sense of justice and peace. I create an atmosphere where the neighbors feel
secure and supported when I work with them.
Trust and involvement of the community are essential to effective policing. This is why I employ
sophisticated crime forecasting devices that help me demonstrate my commitment to proactive
crime prevention, and community safety among other things. Consequently, trust has been
created amongst citizens who are now empowered to act to prevent crimes through interaction
with law enforcement agencies within given societies. In line with this concept, I ensure that
working together leads to strong alliances for the safety of all individuals within our local area
Effective governance is about saying goodbye to guesswork and embracing the use of data in
making decisions. I could adopt a data-driven, machine-learning-based crime-predicting model
that would help us arrive at informed choices. These are useful tools for both strategic and
operational decision-making processes, which facilitate evidence-based decisions regarding
resource allocation, policy formulation, or even implementation of preventive measures within
our institution.
This research contributes to crime analysis through theory and technical development. By
exploring the appropriateness and effectiveness of machine learning techniques in a specific
place such as Wolaita Sodo, I deepen the knowledge base on these models while giving some
insight into challenges and opportunities for their real-world application by law enforcement
agencies. The findings can prompt further investigation that will benefit police organizations
around the world and encourage international collaboration.
In conclusion, the scalability and applicability of these findings and methodologies are
fundamental. The acquired knowledge in this study could serve as a stepping stone for further
inquiries and developments related to crime prevention and prediction. My work may be useful
to law enforcement agencies of Wolaita Sodo. Thus, there is a lot that I can change if we team
up together across boundaries.

1.7. Scope Of The Study:

The scope of study for this research proposal basically involves analyzing historical data about
criminal activities in Wolaita Sodo City and developing predictive modeling using machine
learning techniques. The study will look into the temporal patterns and trends in criminal
incidents and assess the effectiveness of machine learning algorithms in predicting future
criminal activities.

7
The research will be based on historical data on criminal incidents from the Wolaita Sodo city
police department. The data will contain information such as the type of crime, location, date,
and time of occurrence. The paper will focus on a certain period, say five years ago, to ensure
representation of crime patterns.
The data gathered will be preprocessed to handle missing values, outliers, and inconsistencies.
Several machine learning algorithms will be trained and tested on the data, such as ARIMA,
random forests, and SVM, to develop a predictive model. The performance evaluation of models
will be performed using proper evaluation metrics.
The research will be carried out within Wolaita Sodo City, Ethiopia, using data from the police
department. The results from the study will be useful to the law enforcement agencies of Wolaita
Sodo City in understanding patterns and trends in crime, proactive resource allocation,
and proactive measure.
It is important to note that the scope of the study is centered on the analysis of historical crime
data and the development of a predictive model. The study has specifically not included policy
interventions or changed law enforcement strategies. The ethical considerations within this study
will be to ensure the anonymity and confidentiality of the participants' data and appropriate
permissions and approvals from the police department before accessing the data.\
In general, the scope of the study will entail the analysis of time series crime data, building a
predictive model through the application of machine learning techniques, and providing insight
to law enforcement agencies in Wolaita Sodo City to improve their crime prevention strategies.

2. Research Methodology
One kind of information technology research methodology that focuses on assessing the
performance of the result is called design science. It's a model for research in which producing
new artifacts and assessing existing ones is crucial.[6]
The following explains the approach that will be used to the Wolaita Sodo Police Department to
conduct time series crime prediction analysis, using machine learning techniques with relatable
human traits and experiences.
 Data Collection: First, let's get the police agency to provide us with past crime reports
that span at least five years and above 5000 past crime reports. These records, which
offer insightful information on previous criminal episodes, will form the base of my

8
 analysis. I will be collecting very vital details, such as what type of crime, location, date
and time, and any other important information. Compiling this data will enable me to
learn about the dynamics and patterns of crime within our neighborhood.
 Data Preprocessing: Now, let's deal with the outliers, inconsistent data, and missing
information for the dataset I have gathered. I make sure that the information is reliable
and accurate, so I can base my decisions on it. We will aggregate/separate features as
necessary as we clean the data so that we can make analyzing and extracting great insight
easier. I will normalize the date and time features to make our machine-learning
algorithms use them coherently.
 Feature Engineering: Now to engineering features, I will create new ones to be able to
include dependencies and patterns over time. For example, I will develop lag features for
episodes that would record trends over time to detect patterns. Further, I will be
extracting seasonality patterns, such as differences in crime rates by day of the week.
Besides, I will design domain-specific attributes that could impact the crime rate, such as
being close to bars. This will greatly enhance our models' predictive capacity by
thoughtfully engineering these characteristics.
 Selecting Algorithms: Now, let's visualize the machine learning algorithms that will
work well for our time series crime prediction analysis. I would analyze Random Forest
Regressor, support vector regression, and autoregressive integrated moving averages
algorithms. In the process, I am going to assess the viability and efficiency of the
algorithms on tiny data sets. Moreover, I will give priority to the algorithms that not only
work well but are also interpretable—that is, I can understand and explain the variables
that affect their predictions.
Support vector machine (SVM) is a nonlinear model and it has shown good performance
in regression, classification tasks, and time series prediction. In time series prediction, the
effectiveness of the SVM is contrasted with that of the autoregressive moving average
(ARMA) and the recurrent eural network model (RNN).[2]
Autoregressive integrated moving average (ARIMA) forecasts apply advanced
econometric modeling techniques to forecast time-series data by first back fitting to
historical data and then forecasting the future.

9
 A flexible machine-learning method for predicting numerical values is called random
forest regression. In order to decrease overfitting and increase accuracy, it integrates the
predictions of several decision trees.
 Model Training and Validation: I will use 70% of the data to train the models I have
chosen, and the remaining 30% to validate the models. In this manner, I will evaluate our
models' performance on hypothetical data, guaranteeing their generalizability. I will use
grid search to optimize the models by adjusting their hyperparameters until we identify
the optimal setup that maximizes prediction accuracy. To provide a quantitative
assessment of My models' performance, I will assess them using measures such as Mean
Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE),
R-squared (R²) and Mean Absolute Percentage Error (MAPE) For Support vector
Machine, Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared
Error (RMSE), Mean Absolute Percentage Error (MAPE), Akaike Information Criterion
(AIC) and Bayesian Information Criterion (BIC) for ARIMA (AutoRegressive Integrated
Moving Average), Mean Absolute Error (MAE), Mean Squared Error (MSE), Root
Mean Squared Error (RMSE), R-squared (R²) and Mean Absolute Percentage Error
(MAPE) For Random Forest Regressor.
 Forecasting & Implementation: I will use our models to estimate future crime patterns
throughout one to three months once they have been trained and validated. The police
agency will receive forecasts and visualizations from us, giving them practical
knowledge. Through active communication and feedback-gathering with the police
officers, I can improve my models and increase their utility and impact in real-world
situations.
 Impact Evaluation: It is critical to evaluate how our forecasts will affect the distribution
of resources and the general nature of police operations. I will assess the extent to which
our forecasts have contributed to the efficient distribution of resources to regions with a
greater likelihood of crime. I Will also speak with police officers in person to learn more
about their experiences with the models, pinpoint any shortcomings, and get ideas for
enhancements. I can guarantee the applicability and efficacy of models by iteratively
refining them in response to feedback from real-world assessments.

10
  Ethics Review: Data privacy, confidentiality, and equitable results will be given top
priority during the research process. I will discuss the analysis's limits, societal effects,
and possible biases. I must share our findings with the community and the police
department and ask for their comments. I can make sure that My study is carried out
ethically and in line with the needs and beliefs of our community by having an honest and
open discussion.

2.1 proposed architecture of machine learning model

The suggested architecture offers a methodical framework for creating and implementing
machine learning models, guaranteeing efficient data processing, the extraction of pertinent featu
res and the training and validation of the model prior to its deployment for realworld applications
.

Figure 1g. 1. Architecture

3 Related Works And Literature Reviews.

Based on the application of machine learning techniques, the literature review will give a
comprehension in-depth on the current state of knowledge in time series crime prediction
analysis. This will lay the basis on which the proposed research will be built to guide the
selection of appropriate approaches and algorithms for time series crime data analysis and
identification of possible research themes.

11
Paper Author Title Finding Research Gap

[8] Betelhem Crime analysis The decision tree algorithm scored 84% on  adjust and change in step with the ever-
Zewdu and prediction the accuracy, with the random forest changing landscape of illegal activity.
Wubineh using the algorithm and K-nearest neighbor algorithm  Investigating how environmental elements
machine- coming a close second and third, respectively, and socioeconomic factors affect crime
learning with scores of 86.07% and 81%. prediction is necessary.
approach in the
case of the
Hossana Police
Commission
[1] Karabo Machine The study emphasizes the value of machine There aren't many thorough and organized
Jenga, Learning in learning in crime prediction while evaluating literature evaluations on crime prediction,
Cagatay Crime cutting-edge crime prediction methods. It despite the field's growing interest. The study
Catal, Prediction highlights how crucial precise crime emphasizes the necessity of a well-organized
Gorkem forecasting is to averting fatalities and and summarized body of prior research,
property damage. Assuming the availability supporting data, and obstacles that scientists
of labeled data, most of the studies examined must overcome. The authors hope to close this
in the study employed supervised machine- gap by undertaking a Systematic Literature
learning techniques. It does, however, also Review (SLR) and offering insightful
recognize the difficulties posed by sparse information for the next studies on machine
labeled data in practical situations. learning-based crime prediction.

12
[8] Meron Crime The goal of the study was to use a hybrid of deep The use of hybrid deep learning algorithms for crime
Tamiru Analysis and learning algorithms to predict the types of crimes. analysis and prediction was the main emphasis of the

Shibiru Prediction The author employed feedforward artificial neural thesis. The study advances the field by illustrating
network (FFANN) and long short-term memory how the hybrid model can be used to increase
Using Hybrid
recurrent neural network (LSTM-RNN) models to forecast accuracy. To improve the model's prediction
Deep Learning
analyze crime data that was gathered from Bahir power, other variables such socioeconomic
Algorithms
Dar city. When compared to the separate models, characteristics or temporal patterns may need to be
the hybrid model performed better in terms of incorporated, as well as attribute selection strategies
accuracy, mean squared error (MSE), and mean may require more investigation.
absolute error (MAE). The accuracy of the
suggested model was 95.07%.
[9] Romika Crime The use of hybrid deep learning algorithms for The volume, pace, and accuracy of crime data
Yadav, Prediction Using crime analysis and prediction was the main present management and analysis issues that are
Savita Auto Regression emphasis of the thesis. The study advances the highlighted in this study. It highlights how
Kumari Techniques for field by illustrating how the hybrid model can be sophisticated methods are required to manage and
Sheoran Time Series used to increase forecast accuracy. To improve examine the spatiotemporal nature of crime data. To
Data the model's prediction power, other variables such overcome these obstacles, the authors suggest
socioeconomic characteristics or temporal combining big data technologies with auto regression
patterns may need to be incorporated, as well as techniques. By offering a paradigm for more accurate
attribute selection strategies may require more crime prediction, the research advances the field.
investigation. Nevertheless, the literature that is currently available
makes no mention of the particular research deficit.

13
3.1. Cost Break Down And Time Table
3.1.1. Cost Break Down
No Budget item No of Cost per No of Total Description
item item day Day cost(Birr)
1 Gathering of Data 400 30 12000 gathering of data,
Data: collection professionals, and advisors
2 Pay for data 4 person 300 10 4500 To interact with other relevant
annotators and parties who are
(expert) geographically distant and
challenging to reach through
physical means, annotate
unstructured data.
3 additional 2500 2500 Paper, transportation,
equipment, lamination, photocopying, and
stationary, and other uses
logistics
4 Internet services 1000 6 months 6000 To get information from
several sources.
Total 25,000
Table 1. Cost break down

14
 3.1.2. Time Table
No. Task

AUGUST
MARCH

APRIL

JULY
JUNE
MAY
1. Review of the literature and creation of a
proposal

2. Analysis of related work

3. Gathering and analyzing data
4. Building a Model
5. Model testing and assessment
6. Final thoughts and suggestions
7. Documentation
Table 2. time table

15
Reference
[1] K. Jenga, C. Catal, and G. Kar, “Machine learning in crime prediction,” J. Ambient Intell. Humaniz.
Comput., vol. 14, no. 3, pp. 2887–2913, Mar. 2023, doi: 10.1007/s12652-023-04530-y.
[2] N. H. M. Shamsuddin, N. A. Ali, and R. Alwee, “An overview on crime prediction methods,” in 2017
6th ICT International Student Project Conference (ICT-ISPC), Johor, Malaysia: IEEE, May 2017, pp. 1–
5. doi: 10.1109/ICT-ISPC.2017.8075335.
[3] J. Klimaszewski and M. E. Aydin, “Long Short Term Memory for predicting regional crime rates”.
[4] Department of Computer Science and Engineering, Adama Science and Technology University,
Adama, Ethiopia. et al., “Designing Time Series Crime Prediction Model using Long Short-Term
Memory Recurrent Neural Network,” Int. J. Recent Technol. Eng. IJRTE, vol. 9, no. 4, pp. 402–405,
Nov. 2020, doi: 10.35940/ijrte.D5025.119420.
[5] “Crime Prediction and Analysis”.
[6] E. W. Muchie, “MSc. Thesis By Meron Tamiru Shibiru”.
[7] S. S, S. N, V. S, and S. P, “Crime Prediction using Machine Learning,” EAI Endorsed Trans. Internet
Things, vol. 10, Feb. 2024, doi: 10.4108/eetiot.5123.
[8] B. Z. Wubineh, “Crime analysis and prediction using machine-learning approach in the case of
Hossana Police Commission,” Secur. J., Mar. 2024, doi: 10.1057/s41284-024-00416-6.
[9] R. Yadav and S. Kumari Sheoran, “Crime Prediction Using Auto Regression Techniques for Time
Series Data,” in 2018 3rd International Conference and Workshops on Recent Advances and
Innovations in Engineering (ICRAIE), Jaipur, India: IEEE, Nov. 2018, pp. 1–5. doi:
10.1109/ICRAIE.2018.8710407.

16