Associations

In data mining, association rules are useful for analyzing and predicting customer behavior. They play an important part in shopping basket data analysis, product clustering, catalog design, and store layout. For example, if a grocery store wished to sell more 1-liter bottles of Coke, they could examine transactions by customers who bought 1-liter bottles of Coke. By data mining associations, they might discover that customers who bought 1-liter bottles of Coke often bought 15-ounce bags of Lay’s Classic Potato Chips. Once this association is understood, the store could send out coupons for Lay’s Classic Potato Chips and offer a sale on potato chips. They could, then, be relatively certain that the sales of Coke would increase.

While the use of data mining and association rules has nothing to do with selling either Coke or potato chips in criminal justice, the goal of crime data analysis in law enforcement is to identify and visualize associations among criminal networks. For example, findings that 80% of individuals released from a particular prison were subsequently involved in automobile thefts within six months of leaving that prison would be valuable information for the police to have. Similarly, if a crime analyst determined that there was a strong association between young white men in their twenties who applied to purchase a handgun from a national chain store and rejection of that transaction with a subsequent murder, this information could lead to preventative efforts that might save lives.

The Associate Wizard (developed by Microsoft) is one example of software that helps a crime analyst create a data mining model using the Microsoft Association Rules algorithm. Such mining models are particularly useful for creating recommendation systems. How this works is that the Microsoft Association Rules algorithm scans a data set comprised of transactions or events, and finds the combinations that frequently appear together. There can be many thousands of combinations, but the algorithm can be customized to find more or fewer, and to retain only the most probable combinations.

This kind of association analysis can potentially be used to address several problems, including predicting who is likely to commit certain crimes and when these crimes might take place. Association rule mining can be used to generate rules from a crime data set based on the frequent occurrence of patterns to help lead to recommendations for preventive action. But, in general, discovering association rules helps investigators to recognize mutual implications among criminal occurrences.