Location: Old Main Academic Building, Room 1030
Time: Saturday, May 7th, 8:40am - 9:40am
Title: Foresight: A Game-Theory Based System with Reinforcement-Learning for Predictive Analytics
Abstract:
In this talk, Foresight, a forward-looking hybrid predictive analytics algorithm is presented. Unlike machine learning based technologies, Foresight predicts an outcome based on the current dynamics and not based on analysis of the historic data. We have utilized game theory, expected utility and median vector theories, and reinforcement learning to develop the core algorithm. This system benefits from a reinforcement learning mechanism to model players' reasoning ability when it comes to taking risks. Foresight is designed to predict outcomes of complex problems with multiple stakeholders with conflicting interests in economic, business, or political issues. Several case studies and the system’s accuracy will be presented to illustrate its performance.