Title: Application of ensemble methods for detection of changepoints in generalized linear models
Authors: Asanao Shimokawa - Tokyo University of Science (Japan) [presenting]
Takuma Kurosawa - Tokyo University of Science (Japan)
Etsuo Miyaoka - Tokyo University of Science (Japan)
Abstract: The focus is on the estimation of the number of changepoints and detection of their locations under a generalized linear model. If the number of changepoints is small and the size of data is not so large, there are several proposed methods for detecting them, like hierarchical binary splitting algorithm or dynamic programming algorithm. However, if the size of data is too large to search all possibility of change points and/or if it is necessary to estimate the number of change points, these algorithms are difficult to run. To address these problems, we consider to use ensemble methods for estimating the number of change points and detecting those locations at same time. The performance of these methods are examined by simulation studies. Additionally, we show the results of applying the methods to actual data.