F1 Dutch Grand Prix Prediction ML Project

Predicting Formula 1 race results from historical data, sessions, and simulation.

---

🔧 Tools Used

           

_{Core stack: Python for data wrangling & modeling, scikit-learn for ML, Pandas for feature engineering, and FastF1 for F1 timing/telemetry data.}

---

A machine learning project that predicts Formula 1 race results using historical data and free practice session information. This project uses Gradient Boosting (R-squared: 0.494) and Random Forest (R-squared: 0.553) models to forecast driver finishing positions for upcoming Grand Prix events.

🏎️ Project Overview

This project leverages the FastF1 library to access Formula 1 telemetry and timing data, building predictive models that can forecast race outcomes based on:

• Historical race performance data
• Free practice session lap times
• Qualifying results (when available)
• Driver and team form throughout the season
• Grid positions and starting positions

📁 Project Structure

The project is organized into a dedicated Dutch GP folder containing all race-specific analysis and ML models:

f1-simulation-ml-project/
├── Dutch GP/                           # All Dutch GP related files
│   ├── index.html                      # Main results comparison interface
│   ├── dutch_gp_results.js             # JavaScript logic for results
│   ├── load_dutch_gp_results.py        # Python data loading scripts
│   ├── dutch_gp_results.json           # Race results data
│   ├── F1_Dutch_GP_prediction_ML_GBR_Random_Forest.ipynb  # ML models
│   ├── DUTCH_GP_RESULTS_README.md      # Results system guide
│   ├── DUTCH_GP_ORGANIZATION.md        # Complete organization guide
│   └── NEXT_RACE_PREPARATION_CHECKLIST.md  # Next race checklist
├── index.html                          # Main project navigation hub
├── README.md                           # This project documentation
└── F1 Car Image Aug 27 2025 (1).png   # Project logo

🚀 Features

Data Processing

• Historical Race Analysis: Processes past race results to build driver performance profiles
• Free Practice Integration: Incorporates FP1, FP2, and FP3 session data for enhanced predictions
• Qualifying Data: Uses grid positions when available for more accurate predictions
• Driver Name Cleaning: Handles variations in driver names across different data sources

Machine Learning Models

• Gradient Boosting Regressor: Primary model with optimized hyperparameters

• Random Forest Regressor: Alternative model for comparison (Gave 0.55

• Cross-Validation: Uses GroupKFold to prevent data leakage between races

• Feature Engineering: Creates comprehensive feature sets including:

  • Season-to-date performance metrics
  • Form scores and qualifying performance
  • Practice session statistics
  • Historical finish positions and reliability

Prediction Capabilities

• Point Predictions: Direct finish position predictions
• Monte Carlo Simulations: 3000+ simulations for probability analysis
• Win/Podium Probabilities: Calculates chances for different finishing positions
• Expected Points: Projects championship points based on predicted finishes

📋 Prerequisites

Before running this project, ensure you have the following installed:

# Core dependencies
pip install fastf1
pip install scikit-learn
pip install pandas
pip install joblib
pip install tqdm
pip install matplotlib
pip install numpy

🚀 Quick Start

1. Open the main project hub: index.html - This provides navigation to all project components
2. Access Dutch GP Analysis: Navigate to the Dutch GP/ folder for complete race analysis
3. Run ML Models: Use the Jupyter notebook in the Dutch GP folder for predictions
4. View Results: Check the results comparison interface for Dutch GP analysis

🛠️ Installation & Setup

1. Clone or download the project files

2. Install dependencies:

   pip install -r requirements.txt

   Or install manually:

   pip install fastf1 scikit-learn pandas joblib tqdm matplotlib numpy

3. Open the Jupyter notebook:

   jupyter notebook F1_Dutch_GP_prediction_ML_GBR_Random_Forest.ipynb

📊 Usage

Basic Prediction

# Run prediction for a specific Grand Prix
race_order, summary = run_prediction_with_fallback(
    year=2025,
    grand_prix="Hungarian Grand Prix",
    mode="race-only",
    n_sims=5000,
    race_only=False,
    model_type='GradientBoosting'
)

Model Comparison

# Compare Gradient Boosting vs Random Forest
race_order_gb, summary_gb = run_prediction_with_fallback(
    2025, "Hungarian Grand Prix", mode="race-only",
    n_sims=5000, race_only=False, model_type='GradientBoosting'
)

race_order_rf, summary_rf = run_prediction_with_fallback(
    2025, "Hungarian Grand Prix", mode="race-only",
    n_sims=5000, race_only=False, model_type='RandomForest'
)

Prediction Modes

• race-only: Uses only historical race data (no practice sessions)
• with-grid: Includes qualifying results and grid positions
• no-grid: Excludes grid position information

🔧 Key Functions

Data Loading Functions

• load_past_race_results(): Retrieves historical race data
• load_qualifying_results(): Gets qualifying session results
• load_practice_features(): Extracts free practice session data
• build_season_table(): Creates season-wide performance metrics

Model Functions

• build_training_data(): Prepares labeled training dataset
• assemble_prediction_frame(): Creates feature matrix for predictions
• fit_model(): Trains and evaluates ML models
• convert_sim_preds_to_standings(): Converts predictions to race standings

Utility Functions

• clean_driver_name(): Standardizes driver names
• safe_div(): Handles division by zero safely
• session_key(): Creates unique session identifiers

📈 Model Performance

The models are evaluated using:

• Mean Absolute Error (MAE): Average prediction error in positions
• Root Mean Square Error (RMSE): Penalizes larger prediction errors
• R-squared Score: Measures model fit quality
• Cross-Validation: Prevents overfitting using GroupKFold

🏁 Supported Grand Prix

The project includes data for the following 2025 season races:

• Australian Grand Prix
• Chinese Grand Prix
• Japanese Grand Prix
• Bahrain Grand Prix
• Saudi Arabian Grand Prix
• Miami Grand Prix
• Emilia Romagna Grand Prix
• Monaco Grand Prix
• Spanish Grand Prix
• Canadian Grand Prix
• Austrian Grand Prix
• British Grand Prix
• Belgian Grand Prix
• Hungarian Grand Prix

🔄 Data Sources

• FastF1 Library: Primary data source for F1 telemetry and timing
• Official F1 Data: Race results, qualifying times, and session data
• Practice Sessions: FP1, FP2, FP3 lap times and performance metrics

⚠️ Important Notes

1. Data Availability: The project includes fallback logic to use previous year's data if current year data is unavailable
2. Caching: FastF1 caches data locally for improved performance
3. Driver Changes: The model handles driver name variations and team changes
4. Missing Data: Robust handling of missing practice sessions or qualifying data

🤝 Contributing

Feel free to contribute to this project by:

• Adding new features or models
• Improving data processing functions
• Enhancing prediction accuracy
• Adding support for additional Grand Prix

📄 License

This project is for educational and research purposes. Please respect F1 data usage terms and conditions.

🙏 Acknowledgments

• FastF1 library developers for providing access to F1 data
• Formula 1 for the official timing and telemetry data
• The open-source community for the machine learning libraries used

---

_{Note: This project is designed for educational purposes and race prediction analysis. Actual race results may vary due to numerous factors not captured in the model.}