Identifying Factors for Pricing Models

The California State Automobile Association (CSAA) will demonstrate how Salford Systems' CART® and MARS® predictive modeling software addresses analytics challenges in the retail and insurance industries at Predictive Analytics World on March 14th.

Noe Tuason, CSAA customer research manager, will present how he used predictive modeling software to identify factors for pricing models with different distributional assumptions, and how to retain high-profit customers. He specializes in database marketing, data mining, advanced analytics, qualitative research and predictive modeling.

Case studies
Predictive Analytics World is the business-focused event for predictive analytics professionals, managers and commercial practitioners. The event will take place in San Francisco, March 14-15 2011. Over 30 sessions across 2 tracks including case studies from Bank of America, Bank of the West, Fidelity, Google,, Orbitz, PayPal/eBay, plus insights from Anheuser-Busch, the Social Security Administration and Netflix.
Huge databases
“PAW asked me to present how data mining software solves real-world problems,” Tuason explained. “Many companies have vital information hidden in their databases with no way to access it. As a real-world example, I’ll demonstrate how CSAA uses CART data mining software to sift through our huge database to classify and profile prime customers: those who combine good profit with low turnover.
New patterns
 “I’ve been working with CART and MARS software for years. I find CART data mining software useful in finding surrogate variables—mainly demographics—that were not in the original model. The program also helps in reducing variables to a few important ones to enter into linear models and find interactions. With MARS software I can find new patterns and automatically derive new features that are more useful in predicting my target.”

Read more…