Feature-Based Dividing Symbolic Time Series Representation for Streaming Data Processing

Abstract

Along with the continuous development of infor- mation technology, massive numbers of detecting instruments or systems in various fields are continuously producing plenty number of streaming time series data. In recent years, many representation approaches for time series has been proposed with the main objective of dimensionality reduction to support various data mining algorithms in the domain of time series data processing. Symbolic Aggregate approXimation(SAX) is a major symbolic representation and dimensionality reduction algorithm which has been widely used in many application scenarios of time series data mining, such as motifs discov- ery,outlier detection, etc. In this paper, we propose a symbolic representation method of streaming time series based on VTP- diving with sliding window and a similarity measurement algorithm for the proposed representation method which lower bounding the Euclidean distance on the original data. Extensive experiments on different kinds of typical time series datasets have been conducted to demonstrate the advantages of our proposed method.