Python · Machine Learning

Los Angeles Crime Case Analysis

End-to-end analytical pipeline built to identify what factors influence crime case resolution in Los Angeles. The project combines exploratory analysis, feature engineering, and supervised classification to support better resource allocation decisions.

Dataset

LAPD historical crime records with victim, district, and event attributes.

Core Stack

Python, Pandas, Scikit-learn, Matplotlib, Seaborn.

Modeling Goal

Predict if a reported case is likely to be solved.

Delivery

Clean notebook workflow + interpretable visual storytelling.

Workflow

Prepared data by handling missing values, duplicates, and categorical normalization.
Built exploratory views to understand crime trends by location, time, and victim profile.
Created engineered features to improve signal quality for classification models.
Trained and compared Logistic Regression and Random Forest baselines.
Evaluated performance using accuracy, recall, F1-score, and AUC-ROC.

Key Findings

Demographic and contextual variables showed strong explanatory power for case outcomes. District and crime-type patterns revealed significant variation in resolution rates.

77% Best accuracy (Random Forest)

0.82 AUC-ROC

High impact Victim and district features

Visual Output

Main dashboard view for LAPD crime case analysis — Main analytic dashboard with case-solvability context.

Crime distribution by district — Distribution of crime events across city districts.

Feature importance chart from predictive model — Feature importance used to explain prediction drivers.

Download Python Files Back to Portfolio