Development of Knowledge Discovery Basis for Massive Behavioral Data

2014 Fiscal Year Final Research Report

• Project/Area Number: 25730127
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Intelligent informatics
• Research Institution: Tokyo University of Science (2014); Gunma University (2013)
• Principal Investigator: ANDO SHIN 東京理科大学, 経営学部, 講師 (70401685)
• Project Period (FY): 2013-04-01 – 2015-03-31
• Keywords: 巨大情報資源マイニング / 系列パターンインデクシング / 時間非均質性

Outline of Final Research Achievements

This project devoted to building the foundation of the exploratory analysis for very large data sets. It achieved concrete results on behavior sensing data addressing the problems originating from velocity and variety common over very large data sets.
For learning discriminative models under sequence structured data with strong correlations between adjacent observations, we developed indexing based on primitive patterns which improved model interpretability and precision simultaneously in real-world data experiments. Furthermore, we developed cutting-plane method optimization in a multi-scale feature space for temporal data with heterogeneous time-scale and meta-feature generation method for anomaly detection in a multi-scale feature space. These developments made possible the reduction of prediction time and detection of multi-scale anomalies.

Free Research Field

知識発見・データマイニング