Time Series Symbolization Method for the Data Mining K-Means Algorithm

Abstract

Time series is a data type frequently encountered in data analysis. With the current depth and breadth of the data and the improvement in computer processing capabilities, the dimensionality and the complexity of time series are getting higher and higher. Time series symbolization is to cluster and assign complex and lengthy time series in the form of symbols to achieve the purpose of reducing the dimensionality of the sequence or making the sequence easier to process. Considering the excellent performance of the K-means algorithm in data mining and processing, as well as in the allocation algorithm for clustering, we plan to develop a simple method for the symbolization of time series for the K-means algorithm and hope that this method can realize the high-dimensional time series dimensionality reduction, processing of the special points in time series, and so on. Based on this, this article proposes an improved sans algorithm based on the K-means algorithm and discusses the representation method and the data processing of time series symbolization. Experimental results show that this method can effectively reduce the dimensionality of high-dimensional time series. After dimensionality reduction, the information retention rate contained in the elevation of the sequence can reach more than 90%, which is very effective for the detection of outliers in low-dimensional sequences.