Time Series Forecasting With a Learning Algorithm:
An Approximate Dynamic Programming Approach
A Talk by
Tuesday, September 10th
5:45 PM Registration
6:00 PM Seminar Begins
7:30 PM Reception
AbstractWe pose the problem of fitting a time series over a finite period of time as a dynamic stochastic optimization problem, in which the underlying cost functions depend on a measure of model approximation and variation in the selected parameters. We take advantage of the underlying Markov decision process to obtain a model that at optimality considers historical data as well as forecasts of future outcomes. By leveraging the theory of approximate dynamic programming we are able to obtain efficient methods that effectively react to changes in the data and consider the stream of future outcomes obtained from our past model decisions. This give rise to models calibrated to historical data which at any point in time would be optimally positioned to react to possible future data stream. We conduct a broad set of numerical experiments to test our methods on energy-related time series data. Our numerical results show our methods performing strongly against traditional time series forecast methods.
Dr. Ricardo A. Collado currently is an Assistant Professor at the School of Business from Stevens Institute of Technology, NJ. Dr. Collado graduated from the Rutgers Center of Operations Research (RUTCOR) at Rutgers University, NJ. He has previously served as a Professional Specialist at Princeton Laboratory for Energy Systems Analysis (PENSA) from the Department of Operations Research & Financial Engineering (ORFE) at Princeton University, NJ. Dr. Collado also held a position as Assistant Professor/Faculty Fellow at Stern School of Business, Department of Information, Operations & Management Sciences at New York University, NY. His research focus on the science of decision-making in the presence of risk and utilizes dynamic stochastic optimization as its main tool. This line of research impacts areas such as finance, management science, competitive energy markets, auction theory, and homeland security. Dr. Collado's applied research program focuses in the field of energy markets problems in dynamic pricing & demand response and optimizing the design and control of energy portfolios.
Special thanks to the Fordham University Gabelli School of Business for hosting and sponsoring the seminar.
About the Series
The IAQF's Thalesians Seminar Series is a joint effort on the part of the IAQF (www.iaqf.org) and the Thalesians (www.thalesians.com). The goal of the series is to provide a forum for the exchange of new ideas and results related to the field of quantitative finance. This goal is accomplished by hosting seminars where leading practitioners and academics present new work, and following the seminars with a reception to facilitate further interaction and discussion.