-
Notifications
You must be signed in to change notification settings - Fork 0
Home
In today's society there are mass shootings seemingly every day and since the Sandy Hook shooting in 2012 there have been 2,645 deaths because of shootings. This project was made to try and stop the shootings before they happen. It does this by using data from previous mass killers to determine if a user is at risk or not.
The original idea for this happened in 2018 when the Stoneman Douglas shooting gained popularity. I started the project in July of 2019 and started to get finalized in December of that year. The idea started with searching Twitter and morphed into using machine learning/neural networks.
The following lists what this repository can be used for.
- Finding potential mass shooters
- Finding popular mass murder's manifesto's
- Making Classification Report and Confusion Matrix *Making an ROC curve
I found manifestos created by mass shooters and put the segments containing hate speech or threatening language in the training set labeled Class_B. I found harmless tweets using the twitter API and labeled them Class_A. Then, I used the Wayback Machine to find 8chan posts and labeled them Class_B. After data collection was done I had 108 lines in the training set. Then, I used the twitter API again to find tweets to test on. For the Regional Fair I used the Naive Bayes classifier and in the State Fair I used Random Forest.