Model Mining: Exploration of search and enumeration methods of local models from super-high dimensional data

2015 Fiscal Year Final Research Report

• Project/Area Number: 26540116
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Intelligent informatics
• Research Institution: Osaka University
• Principal Investigator: Washio Takashi 大阪大学, 産業科学研究所, 教授 (00192815)
• Co-Investigator(Renkei-kenkyūsha): SHIMIZU Shohei 大阪大学, 産業科学研究所, 准教授 (10509871) ; KAWAHARA Yoshinobu 大阪大学, 産業科学研究所, 准教授 (00514796)
• Project Period (FY): 2014-04-01 – 2016-03-31
• Keywords: データマイニング / 機械学習 / ビッグデータ / モデリング / 高次元データ / サンプリング / アンサンブル

Outline of Final Research Achievements

This study aimed at the exploration of model mining principles, which enable fast search of candidate models representing sub-processes embedded in super-high dimensional and large scale data, and their implementations into some algorithms for applying to experimental problems including medical fields. We established novel principles of random sub-sampling and ensemble modeling for fast and accurate model mining from the large scale data, and developed the methods of half-space mass and and mass based similarity measures by implementing the principles. Finally, by applying these methods to heart disease data in medicine, we succeeded to mine a model of a occurrence mechanism of the heart disease. These outcomes have been presented in Machine Learning:the world top journal of machine learning, ICDM: the world top international of data mining and a major medical journal.

Free Research Field

Data Mining