🎮 Steam Game Recommendation System
🔍 An AI-powered recommendation system that suggests similar Steam games based on metadata, user playtime, and tags.

---

📖 About This Project

This project is an AI-driven game recommendation system that leverages machine learning & NLP to suggest games based on:
✔ User Playtime & Preferences
✔ Developers & Publishers
✔ Genres & Categories
✔ Short Descriptions & Tags
✔ Metacritic & User Review Scores
✔ Real Steam Library Integration

It processes Steam game data, extracts meaningful features, and calculates similarities using cosine similarity & Word2Vec.

---

⚙️ Features

✅ Suggests similar Steam games based on a given game’s Steam ID
✅ Fetches a user's Steam library and analyzes playtime for personalized recommendations
✅ Uses TF-IDF, DictVectorizer, Word2Vec & Cosine Similarity
✅ Playtime-weighted recommendations for accurate results
✅ Filters out low-playtime games dynamically
✅ Web-based interface using Streamlit
✅ Lightweight and fast recommendation engine

---

🚀 How to Use?

1️⃣ Install Dependencies

pip install -r requirements.txt

2️⃣ Run the Web App

streamlit run app/app.py

3️⃣ Enter a Steam ID or App IDs

• Enter your Steam ID to fetch your library and get personalized recommendations
• Enter a list of Steam Game IDs to get recommendations based on selected games

---

📂 Project Structure

📂 steam-game-recommender /
│── 📂 app/
│   ├── 📜 app.py              # Streamlit Web App        
│   ├── 📜 app_local.py        # Local Streamlit App    
│
│── 📂 data/
│   ├── 📜 games.csv           # Original dataset csv file
│   ├── 📜 games.json          # Processed dataset json file
│
│── 📂 models/
│   ├── 📜 game_recommendation.h5 # Precomputed similarity matrix
│   ├── 📜 game_recommendation_creative.h5 # Creative model precomputed similarity matrix
│
│── 📂 src/
│   ├── 📜 main.py              # Core ML pipeline
│   ├── 📜 creative_main.py     # Creative ML pipeline
│
│── 📜 requirements.txt                 
│── 📜 README.md              

📌 Dataset & Model

The complete dataset and model are hosted on Hugging Face: 👉 Hugging Face Dataset

📌 How to Use the Precomputed Model

1. Download the .h5 file from Hugging Face and load the model:

import pandas as pd

with pd.HDFStore("game_recommendation.h5", "r") as store:
    df = store["df"]
    similarity_matrix = store["similarity_matrix"].values

2. Run the recommendation function:

recommended_games = recommend_multi_games(app_id_list, playtime_weights=playtime_weights, top_n=10, min_similarity=0.2)

---

🛠 Technologies Used

• Python (Pandas, NumPy, Scikit-learn, Gensim, Streamlit, Requests)
• Machine Learning & NLP (TF-IDF, Word2Vec, Cosine Similarity)
• Steam API Integration (Fetching real user libraries)
• GitHub LFS (for large dataset storage)

---

🚀 Ready to find your next favorite game? Let’s go! 🎮

---

Note: This project uses data from the Steam Games Dataset by Fronkon Games, available under the MIT License.