SPSS Enterprise Data Mining
Workbench, Clementine"> Unfortunately, the lack of sophisticated analytical tools was a stumbling block to doing more advanced crime prevention work, according to McCue, who described the technology in use at the time she arrived at the RPD as low-tech, descriptive statistics. Not for long. The spring of 2002 became a major turning point for the RPD. McCue obtained an evaluation copy of Clementine, an enterprise data mining workbench from SPSS, which enabled her to make rapid progress in her work.The software, according to the company, lets users develop predictive models using business expertise and deploy them into business operations to improve decision making.McCue played around with Clementine, tweaking it to better accommodate the law-enforcement-type results she was seeking. In some cases, she worked independently; at other times, she consulted with SPSS experts. At this point, a more robust integrator-enabled solution was yet to come. The Clementine workbench gave McCue the advanced IT tools she said she needed to explore her ideas. First, she looked at threat assessment in investigative support work, linking crime based on motive, and then she looked at firearms-related aggravated assaults. In both cases, she used Clementine for data mining predictable patterns of behavior to gain insight into where it made the most sense to deploy police resources. Digging for data The data mining tools allowed the crime analyst to look at the interaction among data, both present and past, such as arrest records, motive and crime at a particular location based on calendar day, time, and weather. So, for example, if the police were better able to understand under what conditions a person was most likely to be victimized, the police could deploy resources to deter the crime from happening. "Using Clementine was like using a power tool after using a hammer and chisel," McCue said. The RPD uses the Clementine Server and Solution Publisher components of Clementine 10. Click here to read about how face-recognition software is helping the LAPD fight gangs. It wasnt until after the RPD tested the value of the law enforcement work that McCue was doing that SPSS and Information Builders joined forces to develop a robust law enforcement application. Using Clementine for the preliminary law enforcement application required that new rules be developed: that is, what would be acceptable error, and so on, which ultimately became the final analytical framework in the application that the RPD uses today. McCue spent about a year using Clementine to look at aggravated assaults. On New Years Eve, 2004, and the Fourth of July that year, the RPD put the analytic application to the test by deploying police resources aimed at more aggressive enforcement of firearm safetybased on the applications statistics on assaults and random gun violence. Looking at historical data for the specific events on particular days, the application predicted with a high or low degree of probability where criminal activity was likely to occur, in the present, under similar circumstances. Based on the analytical information, police resources were strategically deployed. More specifically, the application looked at where it made sense to deploy police based on knowing the likelihood of a crime being committed at a certain time, place and so on. For example, if there is a high probability that an assault will occur outside a particular stadium after a football game, after 9 oclock on a moonless night, the RPD can put police in the area earlier in the evening to deter crime. Built upon IBMs DB2 technology, CrimeMaps tracks crime patterns in an effort to help San Francisco police recognize potential threats. Click here to read more. According to Monroe, there was a substantial reduction in the number of calls made to the police reporting sounds of gunshots, for example. "Calls were reduced by 70 percent," he said. On average, the RPD gets 500,000 calls for service annually, 30,000 of which actually involve criminal behavior. "The other 470,000 calls are [from] citizens saying they want police in the area because they perceive a problem," said Monroe. Without doing the math, Monroe said he knows that if the law enforcement application can target areas where crime might occur, the police department can deploy officers proactively, raise the level of safety in the defined areas and reduce the number of calls for assistance. At the same time, this would allow the RPD to use police officers time more effectively. Calls for service result in the departments sending two officers out in a patrol car for 20 to 30 minutes per call. Officers are the departments most expensive resource. The successful New Years Eve test of the law enforcement application and a second test on the Fourth of July prompted the vendor to take the application to the next level. "We always thought [McCues] work would be more useful with a map interface," said Bill Haffey, technical director for the public sector at SPSS. So did McCue, who said that, until Information Builders was brought into the picture, "We had to use brute-force technology to create visual representations of the SPSS analytical output." SPSS picked up the ball to make the application more visual. It took no time for Haffey to select a technology partner to add the visual capability that the law enforcement application lacked at that time. "We had worked in the past with Information Builders and recognized that they had the technology that was needed," he said. Next Page: Information Builders maps the data.