new opportunities, Milana says, for advanced analytics with banks, in basic retail, and within healthcare—wherever companies have a lot of information about their customers. The most advanced technologies tend to be used in banks, where “they’re all about balancing risks,” Milana says.
For example, banks use Opera Solutions’ predictive analytics technology to help them decide whether to issue loans to customers with no credit history, by using “unstructured” Web data, i.e. data that has not been formatted in a database. Milana says the company usually customizes its technology to meet customers’ needs. For example, a nationwide auto reseller uses Opera Solutions’ analytics to determine the price that a used car of a particular make and model is likely to fetch in different cities throughout the United States. “We also have our fingers in fraud detection, in which accounts get compromised,” as well as in cyber security, “where again it’s searching for outliers and anomalous activity,” Milana says.
“A lot of time, the need drives the application,” Milana says. “Looking at healthcare, for example, there is a lot of data that’s not fully labeled (in contrast to “structured” data in which each field has a specific meaning and value). So you need to make a system that’s sort of focused on anomaly detection—that compares and contrasts. In the case of banks, you’re trying to make a prediction on whether someone will or won’t pay back their loan.”
A substantial part of the KDD 2011 conference is focusing on expanding the use of unstructured data, Milana says, including natural language processing, which analyzes data available in text messages, Twitter feeds, and blogs. The conference, which ends Wednesday, also has sessions focused on social media analytics, Internet ad systems, and bioinformatics. Local speakers include Paul Rejto, Pfizer’s San Diego-based director of computational biology in oncology research, and Dan Steinberg of San Diego-based Salford Systems, who will discuss how predictive modeling and marketing optimization can be used to improve retail sales.
The conference, as Milana puts it, “sits between academia and industry,” and this year includes an industrial practice expo that is intended to highlight the latest issues and implications of data mining and predictive modeling. He adds, “What people are talking about the most is one of the reasons people go to the conference.”
