Mass Shooters - README

IMPORTANT:

• This repository uses data from the Non-Profit organization: 'The Violence Project'
• Because of the 'Terms of Use' outlined by the Non-Profit organization: 'The Violence Project', I will not be giving out the raw dataset nor the prepared version of their dataset
• If you wish to recreate anything from this repository, you must ask for the version 6.1 dataset from The Violence Project
  
  
  

Table of Contents:

Thanks/Citations

Executive Summary

Project Description

Project Goals

Hypothesis/Questions

Data Dictionary

Planning

Instruction To Replicate

Takeaways

Thanks/Citations

Back to 'Table of Contents'

• THANKS TO Jillian Peterson and James Densley from the Non-Profit organization: 'The Violence Project' for allowing me to use their dataset!
• Please visit their amazing website, The Violence Project, if anything within this repository peaks your interest!

Executive Summary:

Back to 'Table of Contents'

• Project Goals
  • Attempt to predict whether a mass shooter is of high volatility (>10 casualties) or is of low volatility (<= 10 casualties).
• Key Findings
  • Orientation
    • From 1966 - 2022, mass shooting events have increased as well as the average casualties per shooting
  • Key Visuals
    • Leads to increase in casualties:
      • More unique felon crimes committed
      • More unique traumatic events experienced
      • More abnormalities the shooter exhibits
• Summary

  • Since 1966, mass shooting events as well as the average casualties have been increasing.

  • It does appear that as an individual deviates further from normal life experiences, psyche, and has less inhibition to harm another, then the individual is more likely to become a highly volatile mass shooter.

• Recommendations

  • If you want the most accurate predictions overall, there's a 10.5% increase from the baseline at 73.7%...

  • If you want the most accurate predictions for highly volatile shooters, there's a 71.4% increase from the baseline at 71.4%...

  • Implementing this model can give key decision makers in a mass shooting event a stronger drive to allocate more resources and/or quicker actions to a mass shooting event should the shooter be identified as highly volatile in order to subdue the threat. However, due to the expansive information this model requires, I'd imagine this would need to be authorized through multiple legal filters such as nationally, locally, by OSHA, by HIPAA, etc. prior to implementation.

Project Description:

Back to 'Table of Contents'

Using a dataset of U.S. mass shooters from 1966 - JAN2023 given by the Non-Profit organization: 'The Violence Project', attempt to predict whether or not a mass shooter will be of high volatility (> 10 casualties) or of low volatility (<= 10 casualties).

Project Goals:

Back to 'Table of Contents'

Obtain and read the mass shooters dataset

Ensure the data is formatted and prepared properly

Identify patterns of mass shooters that lead to increase in volatility

Use patterns for modeling

Identify best model and compare to baseline

Repeat process as necessary before moving to next step

Hypothesis/Questions:

Back to 'Table of Contents'

Hypothesis:

I think that as a person is exposed to and/or participates in more violence, hatred, and essentially anything that is outside of the normal scope of a person's life, then that person will become more accostomed to as well as more willing to harm another person.

Questions:

Does a person's criminal history show an increase in volatility? (Yes)

Does a person's cumulation of traumatic events show an increase in volatility? (Yes)

Does a person's exposure and/or participation in violence show an increase in volatility? (Yes)

Can I discern the shooter's motivation from the data given and see if a particular motivation shows an increase in volatility? (To a particular extent)

Does a person's overall accumulation of events that are significant and abnormal to a normal person's life show an increase in volatility? (Yes)

Data Dictionary:

Back to 'Table of Contents'

IMPORTANT:

• Because of the 'Terms of Use' outlined by the Non-Profit organization: 'The Violence Project', I will not be going in to extrenuous detail of the dataframe especially as to avoid revealing the raw contents of the dataset...

Feature Name	Data Type	Description	Example
197 Binary Columns(Prepared)	int	If something is true or not for a shooter	1, 0
13 Aggregate Column(Prepared)	float	Average score of 0 - 1 from the sum of select columns for each shooter	0.45
3 Datetime Columns(Prepared)	datetime	Datetimes of various specific columns	2017-01-01 08:35:00
41 Object Columns(Prepared)	Object	Columns that contain descriptions, locations, elaboration of specific circumstances, etc.	AZ

Planning:

Back to 'Table of Contents'

Objective

Create a predictive model of a mass shooter's volatility

Methodology

Data science pipeline (Wrangle, Explore, Model)

Wrangle the data by properly acquiring the data then ensuring the data can be interpreted by both human and machine via preparation

Explore for key features, relationships, and patterns

Create clusters if and when necessary

Create models to best predict shooter's volatility

Deliver takeaways

Deliverables

final_report.ipynb

This github repository

Instructions To Replicate:

Back to 'Table of Contents'

IMPORTANT:

• Because of the 'Terms of Use' outlined by the Non-Profit organization: 'The Violence Project', I will not be giving out either the raw or prepared version of this data...
• HOWEVER, if you request the dataset from the Non-Profit organization: 'The Violence Project', and ensure you have version 6.1 of the mass shooter data, then you will be able to run anything and everything in this repository...
  
  
  

Takeaways:

Back to 'Table of Contents'

Summary:

• Since 1966, mass shooting events as well as the average casualties have been increasing.

• It does appear that as an individual deviates further from normal life experiences, psyche, and has less inhibition to harm another, then the individual is more likely to become a highly volatile mass shooter.
  
  

Recommendations:

• If you want the most accurate predictions overall, there's a 10.5% increase from the baseline at 73.7%...

• If you want the most accurate predictions for highly volatile shooters, there's a 71.4% increase from the baseline at 71.4%...

• Implementing this model can give key decision makers in a mass shooting event a stronger drive to allocate more resources and/or quicker actions to a mass shooting event should the shooter be identified as highly volatile in order to subdue the threat. However, due to the expansive information this model requires, I'd imagine this would need to be authorized through multiple legal filters such as nationally, locally, by OSHA, by HIPAA, etc. prior to implementation.
  
  

Next Steps:

1. (CURRENT) Predict volatility of mass shooters
   • Fully exhaust all exploration routes from this dataset (Only 1 excel sheet out of 8 sheets)
     • Attempt to identify stronger features
     • Attempt to improve model accuracy/recall
     • Create regression models to better predict casualties rather than binning them
   • Repeat this process for the 'true' full-dataset (All 8 excel sheets)
     • Attempt to improve findings from #1
   • Ensure the best possible model is created from this 'true' full-dataset
     • All possible exploration/modeling exhausted
2. (FUTURE) Predict shooter to mass shooter
   • Identify patterns of shooters
   • Create model for shooters
   • Attempt to use both shooter and mass shooter model to predict if someone will be a mass shooter as well as their volatility from a population of shooters
3. (FUTURE) Predict criminal to shooter
   • Identify patterns of criminals
   • Create model for criminals
   • Attempt to use criminal, shooter, and mass shooter models to predict if someone will be a mass shooter as well as their volatility from a population of criminals
4. (FUTURE) Predict civilian to criminal
   • Identify patterns of civilians
   • Create model for civilians
   • Attempt to use civilian, criminal, shooter, and mass shooter models to predict if someone will be a mmass shooter as well as their volatility from a population of civilians
5. (FUTURE) See similarities between countries???