How tharuntulasi06/ola-bike-ride-request-forecasting works
This is an Ola ride-demand forecasting and fleet-operations system. Its primary workflow ingests ride, GPS, and weather data; builds spatial clusters and temporal features; trains forecasting models; serves predictions and operational analytics through FastAPI; and presents them in a fleet dashboard. Evaluation, explainability, DuckDB analytics, and the ST-GNN are additional product capabilities. Several model and analytics modules were not excerpted, so those branches reflect README documentation; unsupported internal calls are omitted.
Main entry points: train.py Training entry point · main.py FastAPI service · page.tsx Dashboard page · data_loader.py Dataset loader
Components 14
Data ingestion
- Dataset loader
src/data_loader.py
Forecasting pipeline
- Spatial features
src/feature_builder.py - Training entry point
src/train.py - GBDT ensemble
src/trainer.py - ST-GNN model
src/st_gnn_model.py - Model evaluation
src/evaluate.py - Model explainability
src/explainability.py
API and analytics
- FastAPI service
api/main.py - DuckDB analytics
src/db.py
Fleet dashboard
- Dashboard page
dashboard/app/page.tsx - API client
dashboard/lib/api.ts - Fleet dashboard panels
dashboard/components
Other components
- Fleet operator
- Ride and weather data
Connections 11
- Ride and weather data → Dataset loader: provides datasets
README.md - Dataset loader → Spatial features: supplies ride data
README.md - Spatial features → Training entry point: builds features
README.md - Training entry point → GBDT ensemble: trains ensemble
src/train.py - Training entry point → ST-GNN model: uses model
README.md - GBDT ensemble → FastAPI service: serves predictions
api/main.py - FastAPI service → DuckDB analytics: queries analytics
api/main.py - FastAPI service → API client: provides REST data
dashboard/lib/api.ts - Fleet operator → Dashboard page: uses dashboard
README.md - Dashboard page → API client: requests data
dashboard/app/page.tsx - Dashboard page → Fleet dashboard panels: renders panels
dashboard/app/page.tsx