📍 Baltimore Crime Prediction

Tools used: Python, R, ArcGIS, Tableau

Analyzed 14 years of crime data (750,000+ reports) across 55 neighborhoods in Baltimore City. Combined police, census, and NOAA weather datasets, integrating spatial and temporal features using ArcGIS and Python.

Applied spline-transformed time features and engineered weather and demographic variables to model monthly crime counts. Evaluated XGBoost, Random Forest, KNN, and SVM models, achieving strong performance (R² = 0.85, MAPE ≈ 0.15–0.22) while avoiding overfitting through validation and tuning.

Presented findings through an interactive Tableau dashboard, supporting targeted public safety planning with insights into crime types, risk areas, and community-level trends.

📎 Project Files

📊 Tableau Dashboard

← Back to Portfolio