📈 LAX Airport – Growth Prediction (2002–2017)

Time-Series Forecasting with Prophet

This project demonstrates end-to-end airport passenger forecasting using historic monthly traffic data for Los Angeles International Airport (LAX). The goal is to showcase data cleaning, metric engineering, time-series modelling, backtesting, visualization, and forecasting skills suitable for recruiters and portfolio reviewers.

---

🗂 Project Structure

Project_D_LAX_Airport_Growth_Prediction/
│
├── data/
│   └── raw/
│       └── all_airlines-LAX.csv
│
├── images/
│   ├── lax_prophet_backtest_24m.png
│   ├── lax_prophet_components.png
│   ├── lax_prophet_forecast_5y.png
│   └── lax_total_passengers.png
│
├── notebooks/
│   ├── 01_data_preparation.ipynb
│   └── 02_forecasting_lax.ipynb
│
├── output/
│   └── clean/
│       └── all_airlines_lax_clean.csv
│
├── README.md
└── requirements.txt

---

📊 Dataset Overview

Source: Monthly LAX passenger statistics (all airlines combined)
Period Covered: 2002-10 to 2017-03
Columns Included:

• Domestic / International Passengers
• Domestic / International Flights
• ASM / RPM metrics
• Total passengers (created during cleaning)

This dataset predates COVID-19; forecasts here reflect historical patterns only up to 2017.

---

🧹 1. Data Preparation

All work in 01_data_preparation.ipynb

Key Steps

• Load raw CSV (data/raw/all_airlines-LAX.csv)

• Parse dates into monthly periods (MS)

• Standardise column names to snake_case

• Create total metrics:

  • pax_total
  • flights_total
  • asm_total
  • rpm_total

• Ensure continuous month-to-month index

• Remove missing rows when needed

• Save cleaned dataset to:

output/clean/all_airlines_lax_clean.csv

Sanity Check Plot

A quick visualization was used to confirm dataset continuity and expected seasonality.

---

🤖 2. Forecasting Pipeline

All work in 02_forecasting_lax.ipynb

Model Used

Prophet (Meta)

• Yearly seasonality enabled
• Daily/weekly seasonality disabled (not relevant for monthly data)

Train/Test Split

A 24-month holdout period was used for backtesting:

• Train: 2002-10 → 2015-03
• Test: 2015-04 → 2017-03

---

📐 3. Backtest Performance (24-month holdout)

Metric	Value
MAE	175,934 passengers
RMSE	209,874 passengers
MAPE	5.31%

Performance is strong given airport-level macro data and no exogenous variables.

Backtest Plot

Shows actual vs forecast during the 24-month unseen window:

---

🔮 4. 5-Year Growth Forecast

The model projects a steady upward trend based entirely on 2002–2017 historical patterns.

---

🌀 5. Seasonality & Trend Components

Prophet’s decomposition reveals:

• Strong annual seasonality (peak: Jul–Aug)
• Clear long-term upward trend
• No weekly/daily components (disabled)

---

📝 6. Key Insights

1. Clear Growth Trend (2002–2017)

LAX experienced steady growth driven by both domestic and international demand.

2. Well-defined Seasonality

Summer months show strong peaks; February is consistently the lowest traffic month.

3. Model Fit is Strong

MAPE of 5.31% indicates robust predictive ability for airport-level forecasting.

4. Forecast Limitations

Since the dataset ends in early 2017:

• The model does not capture 2017–2019 rapid growth
• Nor the 2020 COVID collapse and 2021–2024 rebound
• Real LAX traffic reached 76.6M passengers by 2024, far above the ~40M predicted by historic-only data

This is intentional — the goal is demonstrating modelling, not replicating real-world 2024 numbers.

---

📦 7. Tech Stack

• Python 3.x
• Prophet (Meta)
• pandas, numpy
• matplotlib
• scikit-learn
• Jupyter Notebooks

---

📄 8. Installation & Requirements

Install dependencies:

pip install -r requirements.txt

Minimal requirements.txt:

pandas
numpy
matplotlib
prophet
scikit-learn

---

🧭 9. How to Run

1. Open 01_data_preparation.ipynb
2. Run all cells → generates cleaned dataset
3. Open 02_forecasting_lax.ipynb
4. Fit Prophet model, evaluate, generate visuals
5. View plots in /images folder
6. Forecast & insights ready for your portfolio

---

✔ Summary

This project showcases an end-to-end workflow for airport passenger forecasting using a controlled dataset. It demonstrates:

• Data wrangling
• Time-series modelling
• Backtesting
• Seasonality analysis
• Forecast visualization
• GitHub-ready project structure