Detection and Characterization of Triggers of Changes of Relationship or Banking Behavior
PNC Bank collects large amounts of high resolution data from many sources that reflect customer activity and interactions with the Bank. Analytical teams at PNC regularly process this data to generate reports, profile customers and build predictive models, in support of a variety of business ‐ critical objectives, but these analyses typically rely on coarsely aggregated data. We propose to build on our prior small project and, working closely with partners at PNC, conduct focused research and evaluation of relevant machine learning tools on specific case studies of the highest interest to PNC. We will focus on finding and characterizing models of triggers of changes in relationship or banking behavior, using multi ‐ source fine ‐ grain transactional data and focusing on its sequential character. We will represent triggers of change as multi ‐ component temporal association rules. This 2 ‐ year endeavor will enable transitioning of our most effective methods to daily practice of PNC analysts and managers.
