2025_f1_predictions

🏎️ F1 Predictions 2025 - Machine Learning Model

Welcome to the F1 Predictions 2025 repository! This project uses machine learning, FastF1 API data, and historical F1 race results to predict race outcomes for the 2025 Formula 1 season.

🚀 Project Overview

This repository contains a Gradient Boosting Machine Learning model that predicts race results based on past performance, qualifying times, and other structured F1 data. The model leverages:

• FastF1 API for historical race data
• 2024 race results
• 2025 qualifying session results
• Over the course of the season we will be adding additional data to improve our model as well
• Feature engineering techniques to improve predictions

📊 Data Sources

• FastF1 API: Fetches lap times, race results, and telemetry data
• 2025 Qualifying Data: Used for prediction
• Historical F1 Results: Processed from FastF1 for training the model

🏁 How It Works

1. Data Collection: The script pulls relevant F1 data using the FastF1 API.
2. Preprocessing & Feature Engineering: Converts lap times, normalizes driver names, and structures race data.
3. Model Training: A Gradient Boosting Regressor is trained using 2024 race results.
4. Prediction: The model predicts race times for 2025 and ranks drivers accordingly.
5. Evaluation: Model performance is measured using Mean Absolute Error (MAE).

Dependencies

• fastf1
• numpy
• pandas
• scikit-learn
• matplotlib

File Structure

• For every race the end of the file will be numbered in correlation to the race on the calendar, ex. prediction1 - Australia, prediction2 - China, etc.

🔧 Usage

Run the prediction script:

python3 prediction1.py

Expected output:

🏁 Predicted 2025 Australian GP Winner 🏁
Driver: Charles Leclerc, Predicted Race Time: 82.67s
...
🔍 Model Error (MAE): 3.22 seconds

📈 Model Performance

The Mean Absolute Error (MAE) is used to evaluate how well the model predicts race times. Lower MAE values indicate more accurate predictions.

📌 Future Improvements

• Incorporate weather conditions as a feature
• Add pit stop strategies into the model
• Explore deep learning models for improved accuracy
• @mar_antaya on Instagram and TikTok will update with the latest predictions before every race of the 2025 F1 season

📜 License

This project is licensed under the MIT License.

🏎️ Start predicting F1 races like a data scientist! 🚀