Unsupervised ML: Athlete Recommender & Analytics for Paris 2024

📌 Project Overview

This project applies unsupervised machine learning to analyze and recommend athletes for the Paris 2024 Olympic & Paralympic Games. Using clustering and similarity-based methods, we built an athlete recommender system that identifies similar athletes based on key performance, demographic, and social influence attributes.

🚀 Features

• Athlete Similarity Recommender: Finds and suggests athletes with similar profiles.
• Data Preprocessing & Normalization: Encoding, scaling, and handling categorical attributes.
• Chatbot Integration: Users can query the system using text inputs (e.g., "Football male athletes for JO2024 with Instagram influence >10K").
• Exploratory Data Analysis (EDA): Understanding athlete distribution based on multiple attributes.
• Visualization: Graphical representations of athlete clusters and similarities.

📊 Dataset

The dataset contains 600 athletes with 82 features, including:

• Demographics: Age, Gender
• Performance Metrics: Medals (Gold, Silver, Bronze)
• Competition Status: Participation in Paris 2024 (Qualified, Non-Eligible, etc.)
• Sport & Para-sport Category: Encoded categorical variables
• Influence Metrics: Social media following (normalized values)

🛠 Methodology

1️⃣ Data Preprocessing & Encoding

• Encoded categorical variables using the one-hot encoding (e.g., sports, gender, status).
• Normalized numerical features using MinMaxScaler (Age, Followers, etc.).
• Kept only relevant and non-missing features for scoring and similarity analysis.

2️⃣ Feature Engineering & Similarity Computation

• Built a feature matrix (X) combining sports, performance, and influence data.
• Used cosine similarity to measure athlete likeness.
• Identified the top 5 most similar athletes for any given athlete.

3️⃣ Chatbot for Athlete Search & Recommendations

• Developed a query-based chatbot that filters athletes based on textual queries.
• Users can search for athletes using conditions like sport type, gender, status, and social influence.
• Returns a ranked list of matching athletes with their details.

📈 Visualization & Analysis

• Heatmaps to illustrate athlete similarity scores.
• Clustering of athletes based on performance and social influence.
• Graphical representation of top athlete recommendations.

🏆 Example: Finding Similar Athletes

For example, given Hélène Noesmoen, the system recommends the following top 5 similar athletes:

1. Axel Mazella
2. Louise Cervera
3. Lou Berthomieu
4. Jean-Baptiste Bernaz
5. Charline Picon

These recommendations are based on shared attributes like sports category, performance, and social influence.

📌 Future Enhancements

• Integrate deep learning for enhanced similarity detection.
• Expand the chatbot with natural language processing (NLP) for better query understanding.
• Add real-time athlete data updates.

---

Made with ❤️ for Paris 2024 Data Analysis & Recommender Systems 🏅