Paper Abstract and Keywords |
Presentation |
2017-03-07 10:00
Safe Pruning Rule for Predictive Sequential Pattern Mining and Its Application to Bio-logging Data Analysis Kaoru Kishimoto (NITech), Masayuki Karasuyama (NIT), Kazuya Nakagawa (NITech), Kotaro Kimura (Osaka Univ.), Ken Yoda (Nagoya Univ.), Yuta Umezu, Shinsuke Kajioka (NITech), Koji Tsuda (UTokyo), Ichiro Takeuchi (NITech) IBISML2016-105 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Recently, the analysis for time-series logging data of animal behaviors, called bio-logging data, has attracted a wide attention in ethology because of development of a variety of logging devices including GPS. Most of bio-logging data are records of time-series movements of animals which can be represented as a sequence of patterns. In this paper, we consider mining predictive patterns from the sequence data by building a sparse prediction model. However, a huge number of possible patterns exist by which a naive application of machine learning algorithms is prohibitive. We propose an efficient approach to the predictive pattern mining by combining sequential pattern mining and the safe screening technique, which enable us to prune a bunch of unnecessary sub-sequences without losing the model optimality. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Sequential pattern mining / Safe screening / Bio-logging data / / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 116, no. 500, IBISML2016-105, pp. 41-48, March 2017. |
Paper # |
IBISML2016-105 |
Date of Issue |
2017-02-27 (IBISML) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
IBISML2016-105 |
|