320_Group_Project

NYC Motor Vehicle Collisions — Data Science Tutorial

CMSC 320: Introduction to Data Science | Spring 2026 | University of Maryland

Authors: Miskay Zelalem, Dinna Yeshitlla, Emmanuel Michael, Emmanuel Adedeji, Daniel Odetoye

Live tutorial: https://eadedeji8.github.io/320_Group_Project/

---

Overview

This repository hosts our final tutorial for the NYC Motor Vehicle Collisions (Crashes) dataset published by the NYPD on NYC Open Data. The tutorial walks through the full data science pipeline — acquisition and cleaning of ~2.25M crash records, exploratory data analysis across boroughs and time of day, a Welch’s t-test on night-vs-day crash severity, and a binary classifier (Logistic Regression baseline + Random Forest) that predicts whether a crash will result in at least one injury given borough, hour, contributing factor, and vehicle type. The end-to-end story connects spatial / temporal patterns in NYC traffic to concrete policy implications for the city’s Vision Zero initiative.

---

Dataset

• Source: NYC Open Data — Motor Vehicle Collisions (Crashes)
• Reported by: New York Police Department (NYPD), via form MV-104AN
• Size: 2,252,143 rows × 29 columns; cleaned working set is 1,564,912 × 12
• Features used: crash date/time, borough, latitude/longitude, contributing factor (vehicle 1), vehicle type (vehicle 1), persons injured/killed counts

The CSV is gitignored because of size — see the notebook for the loader call.

---

Files in this repo

File	Purpose
crashes.ipynb	The single self-contained tutorial notebook (Sections I–VI + Sources).
crashes.html	Static export of crashes.ipynb — what GitHub Pages serves. Regenerate with jupyter nbconvert --to html crashes.ipynb after any edit.
Final Project Assignment-2-2.pdf	The assignment rubric.
confusion_matrices.png, roc_curves.png, feature_importance.png, classification_report.png	Saved by plt.savefig() calls in Section V; embedded by the HTML export.
README.md	This file.

---