AeroInsight

AI-Powered Airline Intelligence Platform

Data-driven decisions for the modern airline industry — built on BTS domestic airline data (2025 Q1-Q3).

---

Key Findings

Question	Answer
Why does Southwest dominate domestic?	Lowest CASM (8.7c, 22% below median), single fleet type (737), highest gauge (175 seats/dep)
Which aircraft has the best ROI?	B737 MAX 8 (7.3c CASM). Boeing 737 ≈ Airbus A320 families (9.70c vs 9.69c)
Where can airlines save money?	UA: replace A320/B737-800 with B737 MAX 9 (saves 5.9c/ASM). DL: target A321neo (8.5c)
What will next quarter's costs be?	AA Q4 fuel: $2.36/gal (+6%), total fuel cost ~$806M. JetBlue flagged highest risk

Modules

1. Competitive Intelligence

Market share analysis, hub dominance mapping, route competition (HHI), and underserved route identification across 44 carriers and 17,868 routes.

2. Aircraft ROI Optimizer

Aircraft cost rankings by route category, Boeing 737 vs Airbus A320 family comparison, transcontinental analysis, and XGBoost ML recommender (93.7% accuracy).

3. Cost Savings Calculator

Route-specific fleet replacement suggestions, hub optimization (UA ORD: 92 routes analyzed), break-even analysis, and what-if scenario modeling.

4. Predictive Forecasting

Q4 fuel cost forecasts with confidence intervals, budget planning with best/expected/worst scenarios, and carrier risk scoring using cross-carrier panel trend extrapolation.

Quick Start

# Install
pip install -r requirements.txt

# Run data pipeline (if starting from raw data)
python -m src.data.run_pipeline
python -m src.features.feature_engineering_pipeline

# Train ML model
python -m src.models.aircraft_recommender

# Launch dashboard
streamlit run dashboards/app.py

# Run tests
pytest tests/ -v

Data Sources

Dataset	File	Records	Grain
T-100 Domestic	operations2025.csv	376K	Carrier × Route × Aircraft × Quarter
P-5.2 Financial	T_F41SCHEDULE_P52.csv	1,769	Carrier × Aircraft × Region × Quarter
P-12A Fuel	T_F41SCHEDULE_P12A.csv	473	Carrier × Quarter

Source: Bureau of Transportation Statistics — 2025 Q1-Q3.

Project Structure

AeroInsight/
├── dashboards/
│   ├── app.py                    # Unified Streamlit app (6 pages)
│   ├── pages/                    # Page modules (home, competitive, roi, cost, forecast, explorer)
│   ├── components/               # Shared UI components (styles, header, footer, metrics)
│   └── module[1-4]_*.py          # Standalone module dashboards
├── src/
│   ├── data/                     # Pipeline: loader, cleaner, integrator
│   ├── features/                 # Feature engineering (cost, performance, competitive, temporal)
│   ├── modules/                  # Analytical modules (4 modules, 40+ functions)
│   ├── models/                   # ML models (XGBoost aircraft recommender)
│   └── visualization/            # 16 reusable plotting functions
├── notebooks/                    # 8 Jupyter notebooks (data setup through forecasting)
├── tests/                        # 62 tests (unit, integration, data quality)
├── docs/                         # 10 documentation files + 29 figures
├── data/
│   ├── raw/                      # 3 BTS CSV files
│   ├── processed/                # Cleaned parquets + enriched_dataset.parquet
│   └── merged/                   # master_dataset.parquet (351K rows × 76 cols)
└── requirements.txt

Tech Stack

Category	Tools
Language	Python 3.10+
Data	pandas, NumPy, PyArrow
ML	scikit-learn, XGBoost
Visualization	Plotly, Matplotlib, Seaborn
Dashboard	Streamlit
Testing	pytest

Documentation

Document	Description
Dashboard User Guide	How to use the unified dashboard
Technical Docs	Architecture, pipeline, API reference
Data Dictionary	Column descriptions and data sources
Insights Summary	Top 10 findings and 4 key answers
Module 1: Competitive Intel	Market analysis methodology
Module 2: Aircraft ROI	Fleet economics methodology
Module 3: Cost Savings	Replacement analysis methodology
Module 4: Forecasting	Forecasting methodology and limitations
Phase 1: Data Pipeline	Data cleaning and integration
Phase 2: Feature Engineering	Feature computation details
Project Report	Full project report
Demo Script	3-minute walkthrough script

Known Limitations

• Domestic only — T-100 dataset covers US domestic routes. No international data.
• Air ops only — P52 covers ~40-65% of total airline costs (excludes ground ops, admin, depreciation).
• 3 quarters — Q1-Q3 2025 financial data. Forecasts are trend extrapolations, not time-series models.
• Scope mismatch — P52 covers all operations (domestic + international) while T-100 is domestic-only. Widebody and B757 CASM unreliable.
• No revenue data — ROI uses estimated proxies based on industry load factors.

License

This project is for educational and portfolio purposes. BTS data is public domain.