Rivals-Data-Mining

Project for Data Mining course at CSUN, scraping and analyzing match records for Marvel Rivals.

Dataset

• 330,914 matches with 105 features (19 team stats + 86 hero composition)

Models

• Tree Methods: Random Forest, XGBoost, CatBoost
• SVM: LinearSVC with RBF kernel approximation
• Neural Network: Custom 2-layer feedforward network (NumPy)
• RAG Pipeline: Natural language queries via Ollama + ChromaDB

Usage

pip install -r requirements.txt
python benchmark.py      # Run tree model comparison
python rag.py            # Start RAG CLI (requires: ollama pull llama3.1 && ollama pull mxbai-embed-large)
python3 -c "import rag; rag.main()" <<< $'What is the average match duration?\nWhat is the average K/D ratio in matches?\nHow much damage is dealt in a typical match?\nquit'