Over the last 24h, we built a tool for cops to identify the most prominent offenders based on the history of their interactions with the police station.
Domestic Violence (DV) is a challenging issue. Every year,
- 1 in 3 women who is a victim of homicide is murdered by her current or former partner.
- More than 3 million children witness domestic violence in their homes.
Data around these crimes are under-exploited. This tool's impact could be HUGE.
A Tool for Cops & Community Workers
Once a week, an officer of Highpoint, NC might want to know what is the best way to patrol his city and whether or not to pay a visit to a certain location. He visits the website and quickly identifies the adresses likely to experience a serious DV incident in the very near future. His targeted preventive action could have a huge impact on the household we identified as likely to snowball into DV. #preventivepolice #violencesnowball
Once a month, a local community worker might want to target his preventive action to households most likely to experience a crisis in the near future. We provide the tool to do just that.
New Approach : Minority Report 2.0
Targeting of DV is traditionally based on neighborhoods income, education or racial composition but it misses the triggers and signs of violence at the household level.
Our approach is centered around the household's history. Key features of our prediction model include very specific individual knowledge such as the number of times a DV call to the PC originated from their house, when was the last one, whether this person has multiple police reports for drugs or alcohol, whether they committed violent offenses in the past and how violent these were, etc. The relevance of all these features is let to our Elastic Net Logistic Regression model. The model gets smarter every week of new data. The model already provides insightful predictions. For instance, individuals avoiding justice are quite likely to get arrested for domestic violence. #ML
- Interaction with DV specialists: cops, psychologists, sociologists to gather qualitative insights on profiling offenders. #QualitativeData
- Scaling to Other Cities and take into account their specificities #BigData
- Time Series modeling of the individual story