🥇 Olympic Medal Predictor

A machine learning project that predicts the number of Olympic medals a country's team will win, using Linear Regression trained on historical Olympic data.

🌐 Live Demo → TiwariShreya05.github.io/olympic-medal-predictor

---

📌 Project Overview

This project analyzes Olympic team data and builds a predictive model using two key features:

• Number of athletes sent by a country to the Olympics
• Previous medals won by that country in prior Games

The model is trained on data before 2012 and evaluated on 2012 and later years.

---

🌐 Web App

The project is deployed as a fully interactive website on GitHub Pages — no installation required. Just open the link and start predicting!

Features of the web app:

• Sliders and number inputs for athletes and previous medals
• Instant medal prediction with a visual meter
• Estimated gold / silver / bronze breakdown
• Quick-select presets for countries like USA, China, India, France
• Works entirely in the browser — no backend needed

---

📁 Project Structure

olympic-medal-predictor/
│
├── index.html        # Deployed web app (GitHub Pages)
├── Medals.py         # Main script: data prep, model training, evaluation  
├── app.py            # Streamlit app (local)
├── teams.csv         # Dataset (Olympic team statistics)
├── model.pkl         # Saved trained model (generated after running Medals.py)
├── requirements.txt  # Python dependencies
└── README.md         # Project documentation

---

🧠 Model Details

Property	Value
Algorithm	Linear Regression
Features	athletes, prev_medals
Target	medals
Train set	Years < 2012
Test set	Years ≥ 2012
Evaluation	Mean Absolute Error (MAE)

How the prediction works

The model learns a linear equation from historical data:

medals = intercept + (coef_athletes × athletes) + (coef_prev_medals × prev_medals)

Approximate learned coefficients:

• intercept ≈ -0.9
• coef_athletes ≈ 0.02 → each additional athlete contributes ~0.02 medals
• coef_prev_medals ≈ 0.72 → past performance is the strongest predictor

Predictions are clipped at 0 (no negative medals) and rounded to whole numbers.

Key insight: Previous medals carry far more predictive weight (~0.72) than athlete count (~0.02), meaning a country's track record is the strongest signal of future performance.

---

🚀 Getting Started (Local)

1. Clone the repository

git clone https://github.com/TiwariShreya05/olympic-medal-predictor.git
cd olympic-medal-predictor

2. Install dependencies

pip install -r requirements.txt

3. Add the dataset

Place your teams.csv file in the root of the project directory.

4. Train the model

python Medals.py

This will:

• Train the model on pre-2012 data
• Evaluate on 2012+ data
• Print MAE and per-team error ratios
• Save the model as model.pkl

5. Run the Streamlit app (optional)

streamlit run app.py

---

📊 Sample Results

• Predictions are clipped at 0 and rounded to whole numbers
• Per-team error ratios identify which countries are hardest to predict
• Countries like USA and IND are spot-checked individually

---

🛠 Tech Stack

• Python — data processing and model training
• scikit-learn — Linear Regression
• pandas / numpy — data manipulation
• Streamlit — local web app
• HTML / CSS / JS — deployed GitHub Pages version (no dependencies)

---

📬 Contact

• 💼 LinkedIn: linkedin.com/in/shreya-tiwari-520b692b3
• 📧 Email: shreyatiwari0801@gmail.com
• 🐙 GitHub: github.com/TiwariShreya05