In the Tom Cruise movie Minority Report, police were able to accurately predict a crime, its location, and the criminal in advance of the event in time to send police to prevent the crime from occurring. Science fiction at its best, huh? Actually, that’s somewhat of a reality now through predictive analytics. Predictive analytics uses a variety of decision tools and techniques—such as neural networks, data mining, decision trees, and Bayesian networks—to analyze current and historical data and make predictions about the likelihood of the occurrence of future events. Along the lines of Minority Report, police in Richmond, Virginia, are using predictive analytics to determine the likelihood (probability) that a particular type of crime will occur in a specifi c neighborhood at a specifi c time. Using the system, the mobile task force of 30 offi cers is deployed to the areas with the greatest likelihood of crimes occurring. According to Richmond Police Chief, Rodney Moore, “Based on the predictive models, we deploy them [the mobile task force] almost every three or four hours.” Sixteen fugitives have been arrested directly as a result of the system’s prediction of the next time and location of a crime. Moreover, in the fi rst week of May in 2006, no homicides occurred, compared to three in the same week of the previous year. The predictive analytics system uses large databases that contain information on past calls to police, arrests, crime logs, current weather data, and local festivals and sporting and other events. From an IT point of view, the system is a combination of software—SPSS’s Clementine predictive analysis software and reporting and visualization tools from Information Builder—and decision support and predictive models developed by RTI International. The Richmond police afford just one of many examples of the use predictive analytics. Some others include the following: Blue Cross Blue Shield of Tennessee—uses a neural network predictive model to predict which health care resources will be needed by which postoperative patients months and even years into the future. According to Soyal • Momin, manager of research and development at Blue Cross Blue Shield, “If we’re seeing a pattern that predicts heart failure, kidney failure, or diabetes, we want to know that as soon as possible.” FedEx—uses a predictive analytics system that is delivering real and true results 65 to 90 percent of the time. The system predicts how customers will respond to new services and price changes. It also predicts which customers will no longer use FedEx as a result of a price increase and how much additional revenue the company will generate from proposed drop-box locations. University of Utah—uses a predictive analytics system to generate alumni donations. The system determines which of its 300,000 alumni are most likely to respond to an annual donation appeal. This is particularly appealing to most higher-education institutions as they have limited resources to devote to the allimportant task of fund raising. Donations increased 73 percent in 2005 for the University of Utah’s David Eccles School of Business as a result of the system. The future of predictive analytics is very bright. Sales of predictive analytics software exceeded $3 billion in 2008. Moreover, businesses are beginning to build predictive analytics into mainstream, operational applications—such as CRM, SCM, and inventory management—which will further increase their use. According to Scott Burk, senior statistician and technical lead for marketing analytics at Overstock.com, “Predictive analytics is going to become more operational. We’re definitely doing things a lot smarter than we were six months ago.” Overstock.com uses its predictive analytics system to predict demand levels for products at various price points. 41
1. Many predictive analytic models are based on neural network technologies. What is the role of neural networks in predictive analytics? How can neural networks help predict the likelihood of future events? In answering these questions, specifically reference Blue Cross Blue Shield of Tennessee.
2. What if the Richmond police began to add demographic data to its predictive analytics system to further attempt to determine the type of person (by demographic) who would in all likelihood commit a crime? Is predicting the type of person who would commit a crime by demographic data (ethnicity, gender, income level, and so on) good or bad?
3. In the movie Gattaca, predictive analytics were used to determine the most successful career for a person. Based on DNA information, the system determined whether or not an individual was able to advance through an educational track to become something like an engineer or if the person should complete only a lower level of education and become a janitor. The government then acted on the system’s recommendations and placed people in various career tracks. Is this a good or bad 2. 3. use of technology? How is this different from the variety of personal tests you can take that inform you of your aptitude for different careers?
4. What role can geographic information systems (GISs) play in the use of predictive analytics? As you answer this question, specifically reference FedEx’s use of predictive analytics to (1) determine which customers will not respond positively to a price increase and (2) project additional revenues from proposed drop-box locations.
5. The Department of Defense (DoD) and the Pacific Northwest National Laboratory are combining predictive analytics with visualization technologies to predict the probability that a terrorist attack will occur. For example, suspected terrorists caught on security cameras who loiter too long in a given place might signal their intent to carry out a terrorist attack. How can this type of predictive analytics be used in an airport? At what other buildings and structures might this be used?