On New Year's Eve, shots ring out on the streets of Richmond, Va., as people turn their firearms skyward in celebration.
And, in the words of one law enforcement official, "what goes up must come down."
The spike in reported gunshots and the subsequent drain on resources had the Richmond Police Department seeking new ways to stem the flow of calls that require dispatches. As have many other police departments and government agencies, Richmond turned to predictive analytics.
Knowing which areas of the city are more prone to gunfire allowed the force to be proactive. It was able to station officers in those spots to prevent incidents.
This year, Richmond police were able to reduce the number of "shots-fired" complaints by 26% on New Year's Day and by 45% on New Year's Eve. Additionally, police confiscated 45 weapons, including six assault rifles -- up from 13 the previous year.
By identifying potential "hot spots" with SPSS Inc.'s Clementine application, the Richmond PD not only deterred crime and improved its reaction to incidents; it also was able to give some officers the night off.
"[With] this holiday, traditionally we said employees can't take the night off," said Colleen McCue, program manager with the department's crime analysis unit. "This year, we reallocated resources. Some people got the night off, which was a great morale booster."
Analytics also allow the Richmond PD to better allocate personnel resources throughout the year, a vital factor when the department is losing officers to federal agencies and military call-ups, and is facing pressure from terror alerts, McCue said.
Though the tools are most often thought of in terms of leveraging the spending habits of customers, data mining and predictive analytics software offer other methods of maximizing efficiency and cutting costs.
Adoption of business intelligence software is increasing. Initiatives to combat terrorism and fortify national security are fostering its use by government agencies, said Dan Vesset, an analyst with Framingham, Mass.-based International Data Corp. Vendors are beginning to realize this and are trying to offer tools tailored for government.
"There's certainly a big push," Vesset said. "[Vendors] realize that money is tight in the government, but there are certain mandates that are funded, especially around law enforcement."
More than ever, information is flowing into police departments, with homeland security alerts and improved data collection. For about 10 years, the Richmond department had been data mining through "brute force" and standard statistical packages, McCue said. Clementine, purchased two years ago, provides an intuitive and visually friendly alternative with algorithms not typically available, she said.
Software vendors usually direct their products toward businesses, yet the tool has required very little customization, McCue said.
"The biggest challenge has been that most of [the] literature has been about using it on business applications," she said. "We've pretty much been blazing the trail. We know our questions and where we need to get. It's usually just a matter of reformulating."
Additionally, Richmond is dealing with data from a wide variety of sources, including legacy reports written in a narrative style that were never intended for data mining. It's a cache of data as jumbled as that of many businesses. Clementine allows the department to do much of its querying on an ad hoc basis, McCue said.
Richmond has also used analytics for sharing knowledge from investigative rule sets. This practice helps identify motives and flag incidents where crimes are likely to escalate. For example, data analysis helped the department determine that people who burglarize the homes of single women, stealing small or insignificant items, may be likely to commit non-acquaintance rape -- the rape of women they do not know well. Such burglaries in fact constitute a stronger indicator of non-acquaintance rape than do a suspect's prior sex offenses.
While some detectives are still determined to conduct their investigations using note cards, the Richmond PD has been able to foster adoption by disseminating much of the Clementine-generated information with maps of the city, McCue said. For example, operational and analytical teams within the department teamed up to identify the drug markets and determine when activity was set to spike. They then shared that information with officers.
"It's such a paradigm shift for us," McCue said. "In so many places, crime analysis is counting crime and figuring out what happened. This is one of the first times we don't need to connect the dots. We identify the whole picture and determine what happens next."