LA Crime Intelligence & Prediction Engine 🚔

A machine learning project designed to analyze and predict crime types across Los Angeles using over 1 million rows of historical data (2020-Present).

🚀 Overview

This repository contains a full data science pipeline, from exploratory data analysis to a tuned predictive model. The goal is to identify patterns in criminal activity and predict the most likely crime type based on location, time, and demographics.

📊 Key Results

• Data Scale: Processed and cleaned a dataset of 1,000,000+ records.
• Model Accuracy: Achieved a validated accuracy of 33.78% using an optimized Random Forest Classifier.
• Optimization: Utilized GridSearchCV to find the best hyperparameters (max_depth: 15, n_estimators: 100) while managing significant memory constraints.
• Feature Importance: Identified that Location (Premise) and Time of Day (Hour) are the strongest predictors of crime categories.

🛠️ Technical Stack

• Language: Python
• Environment: Jupyter Notebook / Anaconda
• Libraries: Pandas, Scikit-Learn, Seaborn, Matplotlib, Joblib
• Techniques: Hyperparameter Tuning, Label Encoding, Geospatial Visualization

📂 Project Structure

• dataanalytics.ipynb: Full analysis, data cleaning, and model training steps.
• crime_label_encoder.pkl: The saved encoder used to translate categorical data for the model.
• README.md: Project documentation.

⚠️ Important Note on the Model

The final trained Random Forest model (la_crime_predictor_best.pkl) is approximately 1.45 GB in size, which exceeds GitHub's file storage limits.

• The model is excluded from this repository to ensure stability.
• You can recreate the model by running the dataanalytics.ipynb notebook locally.
• A cloud download link for the pre-trained model is available upon request.

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Developed by Arinda Deogracious Data Science & AI Student | Founder-Builder