A0592
Title: Aggregation value regression and its clustering
Authors: Kei Hirose - Kyushu University (Japan) [presenting]
Hidetoshi Matsui - Shiga University (Japan)
Hiroki Masuda - University of Tokyo (Japan)
Abstract: In various practical situations, forecasting aggregate values rather than individual values is often the main focus. A novel forecasting method is introduced, specifically focused on aggregate values in the linear regression model, which is the aggregation value regression (AVR), and it is constructed by combining all regression models into a single model. With AVR, a large number of parameters are estimated when the number of regression models to be combined is large, resulting in overparameterization. To address this issue, a hierarchical clustering technique is introduced, referred to as AVR-C (C stands for clustering). In this approach, several clusters of regression models are constructed, and AVR is performed within each cluster. AVR-C introduces a novel bias-variance trade-off theory under the assumption of a misspecified model. In this framework, the number of clusters characterizes model complexity. Monte Carlo simulation is conducted to investigate the behavior of training and test errors of the proposed clustering technique. The bias-variance trade-off is also demonstrated through the analysis of real data.
