San Diego Traffic Accident Analysis and Prediction (Classification)

This project focused on analyzing traffic accident patterns and developing a predictive model to identify high-risk locations in San Diego. By leveraging machine learning, geospatial analysis, and traffic data from San Diego’s Open Data Portal, I provided data-driven recommendations for road safety improvements.

​

Project Highlights:

• Analyzed 12+ Years of Traffic Data – Processed traffic volume and collision records, identifying University Ave and El Cajon Blvd as the most accident-prone areas, with 1,600+ recorded incidents.

• Developed a High-Accuracy Predictive Model (94%) – Implemented a Random Forest classifier, which outperformed Logistic Regression and SVM in accurately detecting high-risk accident zones.

• Built Geospatial Visualizations – Used GeoPandas and Folium to create interactive maps highlighting accident clusters, helping city planners make data-driven infrastructure decisions.

• Optimized Road Safety Strategies – Provided actionable insights for traffic control measures, pedestrian safety improvements, and speed regulation policies.​

​

Project Impact:

• 94% Accuracy in Predicting High-Risk Locations – The Random Forest model successfully identified accident-prone areas, allowing proactive intervention strategies.

• Enhanced Urban Planning & Traffic Safety – Geospatial analysis helped city planners prioritize high-traffic accident zones for road safety improvements.

• Actionable Insights for Policy Makers – The findings support data-driven decisions on traffic regulation, speed limits, and pedestrian safety enhancements.

• Scalable Model for Future Safety Initiatives – The framework can be expanded to other cities for continuous monitoring and predictive road safety planning.

​

This project demonstrates how machine learning, geospatial analysis, and predictive modeling can be leveraged for urban planning and public safety improvements.

Link for Code

Link for Report

- Programming & Machine Learning: Python (Pandas, NumPy, Scikit-Learn)

- Predictive Modeling: Random Forest, Logistic Regression, SVM

- Geospatial & Data Visualization: GeoPandas, Folium, Tableau, Matplotlib, Seaborn

- Data Sources: San Diego Open Data Portal (Traffic & Collision Records)