Machine Learning Medal Prediction

Overview

This project aims to predict the number of medals a country will win at the Summer Olympics using historical data and machine learning techniques. The workflow includes data preparation, feature engineering, and building a regression model to forecast medal counts.

Datasets

• athlete_events.csv: Raw dataset containing Olympic athlete event results, including athlete demographics, event details, and medals won.
• teams.csv: Processed dataset aggregating country-level statistics per Olympic year, such as number of athletes, average age, previous medals, and more.

Data Preparation

The data preparation is performed in data_prep.ipynb:

• Filters for Summer Olympics only.
• Aggregates athlete-level data to country-year level.
• Computes features: number of events, athletes, average age, height, weight, and medals.
• Adds features for previous medals and rolling average of medals over the last 3 Olympics.
• Outputs the processed data to teams.csv.

Machine Learning Approach

Implemented in machine_learning.ipynb:

• Loads and cleans the processed data.
• Explores correlations and visualizes relationships between features and medal counts.
• Splits data into training (pre-2012) and test (2012 and later) sets.
• Trains a linear regression model using features such as number of athletes and previous medals.
• Evaluates model performance using mean absolute error.
• Analyzes prediction errors by country.

Project Structure

├── athlete_events.csv   # Raw Olympic athlete data
├── teams.csv            # Aggregated country-year features
├── data_prep.ipynb      # Data preparation and feature engineering
├── machine_learning.ipynb # Model training, evaluation, and analysis

Requirements

• Python 3.x
• pandas
• numpy
• scikit-learn
• seaborn

Install dependencies with:

pip install pandas numpy scikit-learn seaborn

Usage

1. Prepare the data:
   • Run data_prep.ipynb to generate teams.csv from athlete_events.csv.
2. Train and evaluate the model:
   • Run machine_learning.ipynb to train the regression model and view results/analysis.

Example

• Predicts medal counts for each country in recent Olympics (e.g., 2012, 2016).
• Example features: number of athletes, average age, previous medals.
• Example output: predicted vs. actual medals for countries like USA, IND, etc.

Contact

For questions or contributions, please open an issue or contact Mike Ayoti