Context

There's currently a lack of large and easily-accessible amounts of detailed data on gun violence.

Content

This project aims to change that; we make a record of more than 260k gun violence incidents, with detailed information about each incident, available in CSV format. We hope that this will make it easier for data scientists and statisticians to study gun violence and make informed predictions about future trends.

The CSV file contains data for all recorded gun violence incidents in the US between January 2013 and March 2018, inclusive.

Acknowledgements

Where did you get the data?

The data was downloaded from gunviolencearchive.org. From the organization's description:

> Gun Violence Archive (GVA) is a not for profit corporation formed in 2013 to provide free online public access to accurate information about gun-related violence in the United States. GVA will collect and check for accuracy, comprehensive information about gun-related violence in the U.S. and then post and disseminate it online.

How did you get the data?

Because GVA limits the number of incidents that are returned from a single query, and because the website's "Export to CSV" functionality was missing crucial fields, it was necessary to obtain this dataset using web scraping techniques.

Stage 1: For each date between 1/1/2013 and 3/31/2018, a Python script queried all incidents that happened at that particular date, then scraped the data and wrote it to a CSV file. Each month got its own CSV file, with the exception of 2013, since not many incidents were recorded from then.

Stage 2: Each entry was augmented with additional data not directly viewable from the query results page, such as participant information, geolocation data, etc.

Stage 3: The entries were sorted in order of increasing date, then merged into a single CSV file.

Inspiration

I believe there are plenty of ways this dataset can be put to good use. If you have an interesting idea or feel like messing around with the data, then go for it.

I was originally inspired to compile it in the wake of the Parkland shooting and the mass media coverage that followed. Reports like this and this showed that Nikolas Cruz had exhibited plenty of warning signs on social media before the shooting; what if we could build a machine learning system that preemptively detected such signs?