Title: Decision forest: an algorithm for classifying multivariate time series

Authors: Ning He; Le-Yang Li; Osamu Yoshie

Addresses: Graduate School of Information, Production and Systems Fukuoka Waseda University, 2–7 Hibikino, Wakamatsu-ku, Kitakyushu, Japan ' Shanghai Mobile Technology Company, Shanghai, China ' Graduate School of Information, Production and Systems Fukuoka Waseda University, 2–7 Hibikino, Wakamatsu-ku, Kitakyushu, Japan

Abstract: Nowadays with time series accounting for an increasingly large fraction of world's supply of data, there has been an explosion of interest in mining time series data. This paper proposes an approach of creating a new data structure automatically, for multivariate time series classification. For more accurate and comprehensive classification, induction of valuable rules named soft discretisation decision forest is illustrated comparing with other machine learning methods such as traditional neural network, SVM and nearest neighbour algorithms. Moreover, some real time series instances from the training dataset will be selected as class dedicated patterns. And a splitting stage using fuzzy theory is prepared for comparing attributes of time series. The ideas of authors are confirmed by simulation results with a set of Japanese vowel time series capably.

Keywords: classification; decision forest; multivariate time series; supervised clustering; soft discretisation; fuzzy partitioning; data mining; machine learning.

DOI: 10.1504/IJBIDM.2012.049555

International Journal of Business Intelligence and Data Mining, 2012 Vol.7 No.3, pp.203 - 216

Published online: 12 Nov 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article