Parallel Distributed Trajectory Pattern Mining Using Randomized Algorithm

2014 Fiscal Year Final Research Report

• Project/Area Number: 25560147
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Geography
• Research Institution: Kobe University
• Principal Investigator: UEHARA Kuniaki 神戸大学, システム情報学研究科, 教授 (60160206)
• Co-Investigator(Kenkyū-buntansha): SEKI Kazuhiro 甲南大学, 知能情報学部, 准教授 (30444566)
• Project Period (FY): 2013-04-01 – 2015-03-31
• Keywords: 地理情報システム / データマイニング / 近似的アルゴリズム / GPS / クラスターアンサンブル / 行動分析 / 移動軌跡データ

Outline of Final Research Achievements

With the rapid increase of the number of mobile GPS devices, it is important to develop efficient and effective algorithms to analyze massive trajectory data streams. Although there are many algorithms that can find patterns by batch processes, what we need is a new algorithm with limited resources by online processes. This research aims at developing such an algorithm and attempts to discover stay points, or the places which are becoming crowded.

Currently, behavioral analysis using trajectory data are widely studied. However, raw GPS data consists of time series data of the coordinates, and does not have any semantic information. Furthermore, because of the problem of private protection, the personal attributes are covered by the data. This research also estimates semantic information of trajectory data using multiple unsupervised learning methods. It is useful as a technique of the privacy-protection data mining by using data without the meaning information.

Free Research Field

知能情報学