🕵️♂️ LAPD Crime Data Analysis
This project presents a comprehensive analysis and predictive modeling of crime case solvability using the Los Angeles Police Department (LAPD) Crime Dataset. The study covers the full data science pipeline: data cleaning, feature engineering, EDA, advanced visualizations, and machine learning.
Tools & Techniques:
- Data cleaning & preprocessing (Pandas, handling missing/categorical data)
- Exploratory Data Analysis (EDA) with heatmaps, barplots, histograms
- Feature engineering: spatial, temporal, demographic variables
- Machine Learning: Random Forest, Logistic Regression, SHAP
- Model evaluation: accuracy, recall, AUC-ROC, class balance
Key Insights:
- Victim demographic patterns (ethnicity, age, gender) are strong predictors of case solvability.
- Random Forest achieved 70.4% accuracy and AUC-ROC 0.75, outperforming Logistic Regression.
- Case resolution rates are higher in certain districts, hours, and for specific crime types.
- Spatial and temporal patterns inform actionable recommendations for resource allocation.
Business Impact: Results inform resource allocation, targeted training, and public policy development for law enforcement, supporting decision-making based on predictive analytics.
📊 Notebooks: