Multi-Objective Forecast Combination

• Activity Number: 341
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Gang Cheng*+ and Yuhong Yang
• Companies: University of Minnesota at Twin cities and University of Minnesota at Twin Cities
• Keywords: multi-objective ; forecast combination ; adaptive loss function ; forecast evaluation

Abstract

Forecast is commonly used in many fields such as economics and finance to serve a specific goal or multiple goals simultaneously. In the literature of forecast combinations, there have been only a few studies exploring how to serve multiple goals simultaneously in a single forecast combing process. In this work, we propose a multi-objective combing method, which build on the AFTER algorithm (Yang, 2004a) by using an adaptive integrated loss function to serve multiple goals. The theoretical results for many versions of this method are obtained, and they suggest that the combined forecasts automatically and consistently perform as well as the best individual among the pool of forecast candidates under each of the multiple evaluation criterions interested. Simulations and data examples show that the proposed method outperforms AFTER when multiple evaluation criterions are of similar importance and performs similarly if one of the criterions is dominant. Besides, the simulation shows the potential advantages of the multi-objective AFTER algorithm comparing with some commonly used combing methods.