An end-to-end data science and machine learning project focused on leveraging graph theory and network analysis to improve route optimization and accurately predict Estimated Times of Arrival (ETAs) for delivery networks.
Academic Context: This project was successfully completed as part of Summer Projects '26, organized by the Consulting & Analytics Club, IIT Guwahati. The official certificate of completion is included in this repository.
Accurate delivery ETAs are critical for logistics efficiency and customer satisfaction. This project analyzes complex delivery networks by modeling them as mathematical graphs. By evaluating node metrics (hubs) and edge weights (routes), we identify bottleneck regions, optimize Full Truckload (FTL) carting strategies, and build machine learning models to dynamically predict ETAs.
Key Highlights:
- Graph Network Modeling: Represents delivery routes and traffic nodes using advanced graph structures to calculate Betweenness Centrality and pinpoint SLA breach rates.
- Dynamic ETA Prediction: Calculates optimized travel times by analyzing temporal delays and network constraints.
- Interactive Dashboard: A comprehensive Streamlit web application that visualizes hub details, node metrics, and strategic memos.
The codebase is modularized into specific tasks and capstone scripts:
task1_2_graph.py: Initial data preprocessing and graph network construction.task3_eta.py: Implementation of baseline ETA calculations and delay metrics.task4_ftl_carting.py: Analysis and optimization logic for FTL (Full Truckload) carting routes.task5final.py: Final integration of task workflows and network intelligence logic.