SPSS is a comprehensive mathematical and statistical package that was originally developed at University of Chicago in 1968. Acquired by IBM in 2009, the package comprises several individual products and solutions, including IBM SPSS Data Collection, IBM SPSS Statistics, IBM SPSS Modeler, and IBM SPSS Deployment. Taken as a whole, the package gives organizations the ability to collect, analyze, and react to almost any type of data. The package has seen widespread use in many industries and applications, including fraud investigation, healthcare, government planning, and education.
The statistical package also has a history of involvement with police work. According to Bill Haffey, technical director for the public sector at IBM, the company has worked with police departments of various sizes to establish crime analysis programs similar to what the MPD was looking for. These programs can draw data from a variety of sources, including raw crime data, city event schedules, and holiday schedules, as well as more unusual sources of data such as weather patterns.
The key to taking advantage of such a wide array of data sources is finding the connections between them. According to Haffey, IBM SPSS excels at this task by being data-driven rather than user-guided."In the discovery algorithms, the data-driven algorithms, you simply introduce a number of various factors that you suspect might have some effect on the outcomes," says Haffey. "The algorithms themselves will sort out the various relationships, the various causes and effects that might be important."
This emphasis on data-driven analysis can have some surprising results. In Richmond, Va., one of IBM's previous forays into crime analysis, data about the phases of the moon, was added into the predictive model. "It was almost a joke, initially," says Haffey. "It turned out that the phases of the moon were a critical portion of the model."
Another key feature of the package is the ability to create "business rules"-sets of criteria that trigger specific behaviors in the package, such as indicating to a user the need for further analysis. He uses the example of insurance claim investigation to illustrate their use.
"In the case of looking at insurance claims, obviously there are some very specific rules that a claims investigator would want to follow," says Haffey. "One example of a very simple business rule that you'd want to incorporate is if a car was totaled and yet there were no passenger injuries reported. ... [That] would be one hard and fast rule that this particular claim should be investigated."
Lastly, and most importantly, IBM SPSS can identify ongoing trends and use historical data and other information to build statistical models that can predict future activity. That sort of capability would be invaluable to the system the MPD was looking to create.
At the end of 2005, the decision was made to go ahead and provide the University of Memphis with crime data for the city, which it could then analyze with IBM SPSS Statistics. Then in 2006, an active pilot program called Blue CRUSH (Crime Reduction Utilizing Statistical History) was launched in one of the department's precincts, based on the university's work.
According to Williams, Blue CRUSH quickly demonstrated its efficiency. "We saw two things occur," he says. "We saw the increase in good quality arrests for those offenders who were using firearms to commit violent felonies. And also we saw a decrease in the overall crime in that area."
Based on the success of the pilot program, several additional programs were launched, and in 2008 the MPD finished training its staff of 12 analysts in the use of IBM SPSS Statistics and took over the operation of the program from the university. Blue CRUSH and IBM SPSS Statistics are now vital parts of the MPD's Real Time Crime Center, which generates both statistical and geographic data that helps the department track and combat crime more effectively.
Williams provides a hypothetical example of the sort of trends Blue CRUSH can identify. "We've got three robberies of individuals that are occurring in a particular area between 12 a.m. and 2 a.m. That's pretty good stuff for a police officer. So we're able, through that, to place those resources and have the officers in the areas where they're occurring," he says.
IBM SPSS Statistics also helps Blue CRUSH generate information packages that can be distributed throughout the department, simplifying communication demands and the need to distribute critical data. But most importantly, the MPD credits the initiative with a concrete impact on the amount of crime, which has dropped nearly 30% since 2006.
"We found that the use of SPSS in our Blue CRUSH initiative has been more successful," says Williams, "than any other saturation or zero tolerance policy in the history of the Memphis Police Department."