Uber Ride Cancellation Analysis

Automated data analysis pipeline that investigates ride cancellation patterns and presents the results in a Grafana dashboard.

Project Walkthrough — full reasoning behind every decision: problem framing, business objectives, EDA methodology, and key findings.

Usage

# 1. Run the analysis pipeline
python run_analysis.py

# 2. Start the dashboard
cd uber-analysis/grafana && docker compose up

Open http://localhost:3000 (admin / admin).

To stop Grafana: docker compose down from the same directory.

Key Findings

• 32% of bookings end in cancellation (~$960K estimated annual loss)
• VTAT >= 15 min → 100% cancellation (system auto-cancel)
• VTAT missing → 100% cancellation (early cancellations before vehicle assignment)
• Temporal features carry no signal — rate is flat across hours, days, months
• Vehicle type is not discriminative — all 7 types show ~32%
• Route feature fails cross-validation — high Cramer's V was a cardinality artifact

How It Works

Raw CSV → Clean → Enrich → Univariate → Bivariate → Multivariate
                                                          ↓
                                                     SQLite DB → Grafana

Every value in the dashboard is computed from data. Nothing is hardcoded.

Project Structure

├── run_analysis.py              # Run this
├── requirements.txt
└── uber-analysis/
    ├── PROJECT_WALKTHROUGH.md
    ├── DATASET_INFO.md
    ├── data/
    │   ├── raw/                 # Kaggle CSV
    │   ├── bronze/              # Cleaned parquet
    │   └── silver/              # Enriched parquet
    ├── grafana/
    │   ├── docker-compose.yml   # Grafana container
    │   ├── data/                # SQLite DB (generated)
    │   └── provisioning/        # Auto-provisioned datasource + dashboard
    ├── notebooks/               # Exploratory (historical reference)
    └── src/
        ├── analysis/            # Pipeline: cleaning, univar, bivar, multivar
        ├── grafana/             # SQLite export + dashboard JSON generator
        └── eda_utils/           # Reusable stats/plotting functions

Prerequisites

• Python 3.10+
• Docker

Setup

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

Dataset

• Source: Kaggle — Uber Ride Analytics Dashboard
• Size: 150,000 bookings from 2024
• License: CC BY-SA 4.0

License

MIT — see LICENSE.