Subscribe to DSC Newsletter

Crime Prediction - Predictive Modeling in Law Enforcement

No predictive model is going to be 100% accurate unless by chance.  The nature of predictive modeling is to learn from the past and see into the future.  Essentially, predictive modeling is just modeling.  Think about why we use statistical models - so we can fit the data into a pattern of behavior and anticipate future results.  It's all about how you use and interpret this model.

 

Crime analysts may use a tool similar to the following example on a robbery series:

 

Data

Date   DOW            DBO         Time    Location Money Stolen
3/1/2005   Tue -- 18:30    1502 E Broadway $1,120
3/8/2005   Tue 7 17:25    1454 S Main St $1,130
3/12/2005   Sat 4 19:00    1903 S First St $1,370
3/17/2005   Thu 5 20:00    1989 N Fourth St $1,420
3/24/2005   Thu 7 20:25    1048 N Third St $1,700
4/2/2005   Sat 9 21:45    1336 E Kimberly Ave $980
 

 

  Average 6 20:09    
  Minimum 2 17:18    
  Maximum 9 23:47    
  St. Dev 1.91 1:36    

 

Predicting Next Hit

  Min Earliest 7/14/2005 17:18    
  1 St. Dev Earliest 7/16/2005 18:33    
  1 St. Dev Latest 7/20/2005 21:45    
  Max Latest 7/21/2005 23:47    
  Correlation 0.79381602      
  Slope 0.00588819      
  Y-Inctercept -1.231583      
  # of Days til Next 3.18      
  Predicted Date 7/15/2005      

 

Based on the frequency of the robberies and the amount stolen, the next predicted date is in 3.18 days.  Now, would you tell your patrol officer that they can expect another hit in exactly that time?  No.  But you can use this information to provide a possible date range.  You can also provide the most likely time for a robbery to occur.  Past data helps us to identify patterns for the future but it is not always the rule to follow for future events.  This data is used in conjunction with intelligence reports from patrols.

 

See the Resources page for other templates like this one this template on Robbery Series.

 

Data is concrete.  It gives us the probability of future events occurring.  Add in variables that you know will affect the probability, and you have a predictive model that is even less concrete.  Explaining this to your "customer" should go something like this:

 

This is what happened [insert data here].  This is the probability of it happening again [insert probability here].  Based on the pattern, the next most likely event occurrence is [insert data range]. 

 

In conclusion, it's all about conclusions.  You will not be able to give a definite "next step", though you can come pretty close.  A marriage of data analysis and intelligence analysis will give you a higher probability of catching and preventing criminal actions.

Read this article on Data Driven 365.

Views: 1666

Tags: analysis, crime, prediction, robbery, series, time

Comment

You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

On Data Science Central

© 2020   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service