Scribd
Upload
46 views10 pages

Machine Learning Algorithm for Delivery Estimation at Swiggy

The document discusses Swiggy's implementation of a machine learning algorithm to improve delivery time estimation and enhance customer satisfaction. It outlines the challenges faced, the data analysis and feature engineering processes, the selection of Gradient Boosting Machines for the model, and the deployment and monitoring of the model in production. Key lessons learned emphasize the importance of data quality, feature engineering, model explainability, and the need for continuous improvement.

Uploaded by

kritika singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
46 views10 pages

Machine Learning Algorithm for Delivery Estimation at Swiggy

The document discusses Swiggy's implementation of a machine learning algorithm to improve delivery time estimation and enhance customer satisfaction. It outlines the challenges faced, the data analysis and feature engineering processes, the selection of Gradient Boosting Machines for the model, and the deployment and monitoring of the model in production. Key lessons learned emphasize the importance of data quality, feature engineering, model explainability, and the need for continuous improvement.

Uploaded by

kritika singh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 10

Machine Learning

Algorithm for Delivery


Estimation at Swiggy
This presentation delves into the machine learning algorithm
employed by Swiggy to enhance delivery estimation accuracy and
optimize customer experience. We'll explore the challenges,
solutions, and key learnings from this project.

by Kritika Singh
Introduction to Swiggy's Delivery Estimation
Challenge
The Challenge Factors Influencing Delivery Times

Swiggy, India's leading online food delivery platform, Delivery times are influenced by numerous factors,
faces the challenge of providing accurate and reliable including distance, traffic conditions, restaurant order
delivery time estimates to its customers. This is crucial preparation time, and delivery driver availability. These
for customer satisfaction and trust. factors are dynamic and unpredictable, making it
difficult to estimate delivery times with precision.
Defining the Problem
Statement and Business
Objectives
1 Problem Statement 2 Business Objectives
To develop a machine Improve customer
learning model that satisfaction by providing
predicts accurate delivery accurate delivery time
time estimates for Swiggy estimates.
orders.

3 Business Objectives 4 Business Objectives


Increase efficiency by Reduce delivery
optimizing delivery routes cancellations and
and driver allocation. customer complaints.
Exploratory Data Analysis and Feature Engineerin
Data Exploration Feature Engineering

A comprehensive analysis of Swiggy's historical order New features were created by combining and
data was conducted to understand the key factors transforming existing data, including distance, time of
influencing delivery times. day, restaurant popularity, and traffic conditions.
Selecting the Appropriate ML Algorithm

Regression Algorithms
Linear Regression, Random Forest Regression, Gradient Boosting Machines
1 (GBM)

Neural Networks
2
Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM)

Final Choice
Gradient Boosting Machines (GBM) were chosen due to
3
their ability to handle complex relationships and high
accuracy.
Model Training and
Hyperparameter Tuning
Training Data
The model was trained on a massive dataset of historical
order data, labeled with actual delivery times.

Hyperparameter Optimization
Hyperparameters, such as the learning rate and tree
depth, were tuned to optimize the model's performance.

Cross-Validation
Cross-validation techniques were employed to ensure the
model's generalizability and prevent overfitting.
Evaluating Model
Performance and Iterating
Metrics Iteration
The model's performance Based on the evaluation
was evaluated using metrics results, the model was
such as mean absolute error iteratively improved by
(MAE), root mean squared refining features, adjusting
error (RMSE), and R- hyperparameters, and
squared. exploring alternative
algorithms.
Deploying the Model into
Production

Cloud Infrastructure API Integration


The trained model was deployed An API was created to allow the
on a cloud platform, ensuring model to be accessed by Swiggy's
scalability and reliability. mobile and web applications.

Real-Time Predictions
The model now provides real-time
delivery time estimates to
customers during the ordering
process.
Monitoring and Maintaining
the Model
1 Performance Monitoring
The model's performance is continuously monitored to
identify any drifts or changes in accuracy.

2 Data Updates
The model is retrained periodically with new data to
ensure its accuracy remains current and relevant.

3 Model Maintenance
Regular maintenance and updates are performed to
address any performance issues or changes in the
environment.
Lessons Learned and Future
Improvements

1 2
Data Quality Feature Engineering
The quality of data is critical for model Creative feature engineering can
accuracy. Ensuring clean and significantly improve model
comprehensive data is essential for performance. Exploring new features
effective model training. and combinations is key.

3 4
Model Explainability Continuous Improvement
Understanding why the model makes Machine learning models are not static.
certain predictions is important for trust Continuous monitoring, retraining, and
and debugging. Techniques like SHAP iteration are essential for optimal
values can help. performance.

You might also like