Baylor Model Helps Predict Crime in the City of Waco

Article excerpt

In 1991, the Waco Police Department and other city officials were looking at ways to reduce crime in Waco. "All of us wanted to make better use of the information contained in crime report records," says Robert Cervenka, the city's urban planner. "We approached Baylor University in the hope that they could provide the technical expertise to design and implement a system for analyzing and predicting criminal activity."

The project found its way to professors Kris Moore and Samuel Seaman in the Hankamer School of Business. "We welcomed the project," says Seaman. "We felt it was a unique opportunity to work on a significant real world problem using statistics and information technology."

The City of Waco wanted the university to develop a statistical-based model using crime reports, weather data, census data, unemployment data, school attendance records, and the lunar calendar. The software created would have both the statistical ability to predict crime and the database management capabilities required to better analyze crime records.

The Cooper Foundation of Waco took an interest in the project and agreed to provide the funding. The Waco City Council approved city participation in the project in the summer of 1992. Once these hurdles were cleared, work on the project was set to begin.

Building the Database

Police files providing over 100,000 individual crime reports were electronically transferred to Baylor during the fall of 1992. The files contained pertinent information about all major crimes reported in Waco over a six-year period. Names and addresses in the records were encrypted to insure the privacy of all individuals identified in the reports.

"The transfer and organization of files required a collaborative effort on the part of many individuals," says Seaman, who was in charge of the projects computing activities. These activities involved staff of the business school's Casey Computer Center, programmers from the university's computer science department, computer staff of the Waco Police Department, and faculty and graduate students of the Hankamer School of Business.

The database grew to include additional information about the weather, parolees, alcohol sales, and athletic and special event schedules. "We really wanted to find out if environmental or demographic factors could help us improve our ability to predict crime," says Moore. "We wanted to put aside any preconceived notions and to look at all viable factors for which information could be obtained. We even examined possible influences of the lunar phase and were somewhat surprised when a relationship was actually discovered."

Building the model

Most of 1993 was spent developing a variety of statistical models and testing their accuracies. This involved checking for relationships between different combinations of the variables, developing forecasts, and then checking these forecasts against the historical records. Models were developed for each of the 150 police reporting districts of Waco.

"Overall, these models provide better than 80% accuracy in their predictions of historical crime patterns," says Moore. "While some decrease in accuracy will likely occur when the model is implemented, we are happy with our results and feel confident that the police force can reliably use these predictions."

Managing the model

The data management system and all forecasting models are scheduled to be transferred to the police department this fall.

This author is working closely with the crime analysis unit to help smooth the transition. …