New York City Car Collision Prediction

Overview:

In this master's Project, I delved into predicting dangerous car accidents in New York City achieving over 60% accuracy using traditional machine learning techniques. The project aimed to provide data-driven insights into the core reasons and contributing factors behind these accidents.

Data Sources:

The project Utilized multiple data sources: the NYC Motor Vehicle Collisions - Crashes dataset, Weather data, and Speed Limit data.

Tools and Technologies Used:

The project employed a range of techniques and models, including:

• Data integration from multiple sources
• Comprehensive data processing and cleaning, addressing issues like missing data and feature preparation
• In-depth data analysis, visualization, and storytelling using Matplotlib and web-based presentations
• Exploration of temporal patterns, geographic hotspots via map plotting (Folium), and accident severity trend analysis
• Development of interactive dashboards (Bokeh)
• Implementation of effective data storytelling and visualization methods
• Implementation of class balance methods
• Development of a data preprocessor pipeline
• Leveraging traditional machine learning models like Logistic Regression, Random Forest Classifier, and Decision Tree Classifier
• Assessment of feature importance

Key Highlights:

• Developed a machine learning model to predict the severity of car accidents.
• Designed a website with interactive dashboards for data storytelling and visualization.
• Explored important features contributing to severe accidents, such as street names, vehicle types, and weather conditions and identified key insights for accident prevention.

Learn More

For an in-depth and interactive exploration of this project, including methodologies, code, and more, please visit the project website here..