Implementation of decision support systems based on statistical classification/regression models with sparse regularization

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330049
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Statistical science
• Research Institution: The Institute of Statistical Mathematics
• Principal Investigator: Kawasaki Yoshinori 統計数理研究所, モデリング研究系, 教授 (70249910)
• Co-Investigator(Kenkyū-buntansha): Ueki Masao 久留米大学, バイオ統計センター, 講師 (10515860) ; Akashi Kentaro 学習院大学, 経済学部, 教授 (50610747)
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: スパース正則化法 / 分類・パターン認識 / 変数選択 / 変数グルーピング / 高次元交互作用 / リスク解析 / 多重共線性 / 予測モデル

Outline of Final Research Achievements

We promoted the smooth-threshold estimation equations (STEE) to develop a prediction model with high accuracy even in high dimensional regression problems. In the analysis of bank telemarketing data, STEE as well as other methods like lasso reveals the features of the customers who are likely to subscribe bank deposits. We also conducted comparative study of prediction accuracy among competing methods changing the variable selection methods, with or without grouping. We also worked on Genome Wide Association Study where we developed a variant of STEE, smooth-threshold multivariate genetic prediction model to forecast the personal risk related to a certain disease. Finally we proposed a method to retain multiple rival models where some of the explanatory variables are highly correlated.

Free Research Field

統計科学