B1612
Title: Clusterwise joint lasso with penalty term for discriminating each cluster
Authors: Shinta Urakami - Doshisha University (Japan) [presenting]
Hiroshi Yadohisa - Doshisha University (Japan)
Abstract: In analysing high-dimensional data, it is difficult to interpret the results using ordinary multivariate analysis owing to a large number of dimensions. In this case, sparse estimation of the values to be estimated may facilitate the interpretation of the results. In addition, when the data potentially contain a cluster structure, clustering and stratified regression can be used to estimate the regression coefficients for each cluster, after which their coefficients can be interpreted. However, if we apply regression after clustering, we may end up estimating regression coefficients that do not correctly capture the cluster structure of the data. Clusterwise regression analysis, which performs clustering and estimation of regression coefficients simultaneously, may solve this problem. We propose a clusterwise joint lasso with a penalty term for discriminating each cluster. This lasso method simultaneously performs clustering and estimation of regression coefficients and regularizes the objective function to increase the difference in the values of the regression coefficients for each cluster. By applying this method, it is possible to clearly distinguish the values of the regression coefficients among clusters, which may make it easier to identify the characteristics of each cluster.
