MoraaOntita/Logistic-optimization-Delivery-drivers-location-optimisation-with-Causal-Inference-

This project aims to optimize the placement of delivery drivers (referred to as pilots) for Gokada, the largest last-mile delivery service in Nigeria. By understanding the primary causes of unfulfilled delivery requests and applying causal inference methods, we aim to recommend driver locations that increase the fraction of completed orders. # Gokada Last Mile Delivery Optimization Project ## Overview The Gokada Last Mile Delivery Optimization Project aims to enhance delivery fulfillment rates by analyzing historical data and leveraging advanced analytics techniques. ### Business Need Gokada is the largest last mile delivery service in Nigeria, facing challenges with sub-optimal pilot (driver) placement leading to high rates of unfulfilled delivery requests. This project seeks to identify root causes of unfulfilled orders and propose data-driven solutions to improve operational efficiency. ### Key Objectives - Analyze historical data to understand factors influencing delivery success rates. - Implement machine learning models and causal inference techniques to predict and optimize pilot placements. - Create visualizations to communicate insights and recommendations effectively. ## Project Structure ### Tasks 1. **Data Exploration and Feature Engineering**: - Conducted exploratory data analysis (EDA) to identify patterns and trends in delivery requests. - Engineered features including time-based variables (e.g., hour of the day), weather conditions, and traffic data to enrich the dataset. 2. **Creative Visualization**: - Developed interactive visualizations to depict spatial and temporal aspects of delivery operations. - Inspired by projects like "A Day in the Life" to showcase pilot movements and delivery interactions. 3. **Causal Learning and Predictive Modeling**: - Built causal graphs to understand relationships between variables influencing delivery fulfillment. - Utilized machine learning models (e.g., XGBoost, Random Forest) to predict and optimize pilot placements. ### Tools and Technologies Used - Python (NumPy, Pandas, Scikit-learn) - Causal Inference (Judea Pearl’s Framework) - Data Visualization (Matplotlib, Seaborn, Datashader) ## Repository Structure - `data/`: Contains raw and processed datasets. - `notebooks/`: Jupyter notebooks for data exploration, modeling, and visualization. …