911 Calls Data Analysis & Predictor

I built this project to analyze 911 emergency call data from Montgomery County, PA. I performed end-to-end data analysis starting from raw data exploration all the way to deploying a live machine learning web app on Streamlit Cloud.

Dataset I worked with a dataset of 663,522 emergency calls recorded between 2015–2020. The data contains the following fields:-

lat : String variable, Latitude lng: String variable, Longitude desc: String variable, Description of the Emergency Call zip: String variable, Zipcode title: String variable, Title timeStamp: String variable, YYYY-MM-DD HH:MM:SS twp: String variable, Township addr: String variable, Address e: String variable, Dummy variable (always 1)

I extracted the call reason (EMS, Fire, Traffic) from the title column.

EDA I analyzed 663K+ 911 calls exploring distributions by reason, day, month and hour. I built heatmaps, time series plots per reason (EMS/Fire/Traffic), top township rankings, and boxplot to uncover call patterns across time and location.

Machine Learning I trained Logistic Regression, Decision Tree and Random Forest to predict call reason from time/location features. Random Forest and Decision Tree achieved ~56% accuracy.

Live App 911 Calls Predictor — Built and deployed on Streamlit Cloud. Enter location, time and day to get a live prediction with confidence scores.

TechStack Python • Pandas • Seaborn • Plotly • Scikit-learn • Streamlit