Financial Statistics/Marketing Seminar

ABOUT THE SPEAKER

Grigoris Karakoulas is founder and president of infoAgora Inc. that provides consulting services on customer behaviour analytics to financial services organizations. He is also adjunct professor in the Department of Computer Science at the University of Toronto. With more than twelve years of R&D experience in predictive modeling, Grigoris has a proven track record of creating and implementing customer decisioning solutions that positively impact financial performance and shareholder value. Prior to founding infoAgora Inc., he worked for a major Canadian financial institution as VP Customer Behaviour Analytics. Grigoris holds a PhD in Artificial Intelligence. He has published more than 40 papers in journals and conference proceedings in the areas of predictive modeling in banking, machine learning, data mining and E-commerce.

ABSTRACT

Financial institutions are becoming increasingly aware of the importance of customer behaviour modelling for risk management. In this talk we argue that advanced modeling techniques from machine learning (e.g. neural networks, kernel estimators and bayesian networks) can be more appropriate and tend to outperform traditional techniques (e.g. linear and logistic regression) for customer behaviour modelling. We provide support to this argument through two case-studies, a regression problem in mortgage prepayment behaviour and a classification problem in redemption of term deposits.