Created by Ansh Bhargava, Nikhil Golla, Prajay A Sachdev, and Armaan Kapoor for the 2023 Cubist Hackathon.
Classify 'Motor Vehicle Collisions' from the NYC Open Data Collective as 'fatal/nonfatal'. Select, preprocess, and encode features (spatial coordinates, vehicle type code, weather, time) to train gradient-boosted decision trees, LSTM networks, and deterministic statistical models and carry out classification/inference.
Utilized the Weather Query Builder from Visual Crossing to join weather data with collision data by the time of the collision. This enabled us to analyze how various weather conditions may have potentially influenced the severity of the collision.