Skip to content

Repository files navigation

Project Uber: Machine Learning-Based Ride-Hailing Optimization (Delhi NCR)

Overview

Project Uber is a comprehensive research and analytics project focused on optimizing ride-hailing operations within the Delhi NCR region using machine learning. The study integrates forecasting, revenue modeling, and operational efficiency with fairness and sustainability evaluation. It leverages real-world ride data to identify demand trends, improve fleet allocation, and reduce emissions while maintaining service reliability.


Repository Structure

Project-Uber/
│
├── Final Submissions/             # Complete research package
│   ├── Thesis_Submissions/        # Core working directory with all essential research materials
│   │   ├── Code/                  # Python scripts for data analysis, modeling, and figure generation. Drawio files.
│   │   ├── Documentation/         # Weekly reports, analyses, and markdown summaries
│   │   ├── Graphs/                # Final figures used in the paper and presentation
│   │   ├── Graphics/              # Supporting visuals and conceptual diagrams
│   │   └── Illustrations/         # Other presentation visuals (AI Generated)
│   ├── Project_Uber_Thesis_Paper.docx
│   └── Machine_Learning_Based_Forecasting_RideHailing_DelhiNCR.pptx
│
├── data/                          # Processed and interim datasets used for training and validation
│   ├── interim/
│   └── processed/
│
├── models/                        # Trained models and parameters for forecasting and revenue prediction
│   ├── week5_best_rf.pkl
│   ├── week5_best_xgb.json
│   ├── week6_rf_revenue.pkl
│   └── week6_xgb_revenue.json
│
├── reports/                       # Week-by-week documentation of analytical stages and findings
│   ├── week1_data_audit.md
│   ├── week5_demand_summary.md
│   ├── week6_revenue_summary.md
│   ├── week7_ops_brief.md
│   ├── week8_results.md
│   └── week9_discussion.md
│
├── analysis_results/              # Final summarized outputs of model performance and analysis tables
│   ├── project_summary.txt
│   └── summary_table.txt
│
├── requirements.txt               # Python dependencies
└── README.md                      # Repository guide and navigation document

Environment Setup

To reproduce this project locally:

# Clone the repository
git clone https://github.com/kangss1/Project-Uber.git
cd Project-Uber

# Create a virtual environment (optional but recommended)
python -m venv venv
source venv/bin/activate   # for macOS/Linux
venv\Scripts\activate    # for Windows

# Install dependencies
pip install -r requirements.txt

Core Dependencies

  • pandas
  • numpy
  • scikit-learn
  • matplotlib
  • seaborn
  • xgboost
  • shap
  • statsmodels
  • pmdarima

Weekly Workflow and Reproduction Guide

Week 1 — Data Intake and Quality Audit

Script: week1_data_audit.py

  • Load raw dataset and validate structure.
  • Detect missing values, duplicates, and anomalies.
  • Generate a structured data audit report.
  • Deliverables: week1_data_audit.md, quality checklist.

Week 2 — Data Cleaning and EDA

Script: week2_cleaning_eda.py

  • Clean missing and invalid records.
  • Visualize trends, correlations, and distribution patterns.
  • Save cleaned dataset ncr_ride_bookings_cleaned.csv.
  • Deliverables: Cleaned dataset and EDA visualizations.

Week 3 — Feature Engineering

Script: week3_feature_engineering.py

  • Create time-based and rolling features.
  • Encode categorical and scale numeric features.
  • Generate enriched dataset for modeling.
  • Deliverables: Feature-engineered dataset.

Week 4 — Baseline Models

Script: week4_baselines.py

  • Build ARIMA and Linear Regression baselines.
  • Evaluate MAE/RMSE and residuals.
  • Establish reference models.
  • Deliverables: week4_baselines_summary.md and residual plots.

Week 5 — Advanced Demand Forecasting

Script: week5_advanced.py

  • Train Random Forest and XGBoost models for demand prediction.
  • Perform grid search tuning and model comparison.
  • Generate SHAP and feature importance plots.
  • Deliverables: Best model artifacts and demand forecasts.

Week 6 — Advanced Revenue Modeling

Script: week6_revenue_advanced.py

  • Build Random Forest and XGBoost revenue prediction models.
  • Analyze feature impacts using SHAP and PDP plots.
  • Segment results by vehicle type and time of day.
  • Deliverables: Revenue model summary and visualizations.

Week 7 — Operational Insights

Script: week7_operational_insights.py

  • Generate demand and cancellation heatmaps.
  • Examine peak hour performance and service reliability.
  • Relate findings to fleet allocation efficiency.
  • Deliverables: Operational performance heatmaps and brief.

Week 8 — Method Integration and Results

Script: week8_synthesis_methods_results.py

  • Integrate results from demand, revenue, and ops analysis.
  • Prepare thesis Methods and Results sections.
  • Deliverables: Consolidated results tables and draft figures.

Week 9 — Discussion and Reflection

Script: week9_interpretation_reflection.py

  • Interpret findings; identify key limitations.
  • Discuss fairness and sustainability implications.
  • Deliverables: week9_discussion.md and week9_limitations.md.

Week 10 — Finalization and Presentation

Files: Thesis Paper & Presentation

  • Compile final thesis and slides.
  • Review figures and confirm metrics.
  • Prepare oral defense materials.
  • Deliverables: Final thesis and presentation deck.

Project Scope

  • Demand Forecasting: Predict hourly and daily ride demand.
  • Revenue Modeling: Understand fare variability and contributing factors.
  • Operational Efficiency: Analyze fleet utilization and cancellations.
  • Sustainability Analysis: Evaluate emissions and fairness across regions.

Key Outcomes

  • Forecast Accuracy: Ensemble models achieved the lowest MAE and RMSE.
  • Operational Efficiency: Average service fulfillment ~93% with fewer peak-hour cancellations.
  • Fairness: 12% reduction in regional trip allocation disparities.
  • Sustainability: 9% reduction in total fleet emissions.

Citation

Kang, S. (2025). Machine Learning-Based Forecasting and Optimization of Ride-Hailing Systems in Delhi NCR. Project Uber Research, 2025.


Contact

Author: Sandeep Kang
LinkedIn: linkedin.com/in/sandeepkang
Year: 2025


Project Uber — A data-driven framework for forecasting, optimization, and sustainability in ride-hailing systems.

About

Uber Ride-Hailing: Forecasting Demand and Revenue for Efficiency in Delhi NC

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages