Abstract:A long standing challenge for time series analysis is to develop representation techniques for dimension reduction while still preserving their fundamental features. As an effective representation technique, Symbolic Aggregate Approximation (SAX) has been widely used for dimension reduction in time series analysis. However, SAX always maps time series data into symbols by definite breakpoints. As a result, the similar points close to the breakpoints cannot be well represented, and thus lead to poor Tightness of Lower Bounds (TLB). To fill this crucial void, in this paper, we develop a time series representation method, named Random Shifting based SAX (rSAX), which has the ability in significantly improving the TLB of representations without increasing the corresponding granularity of representations. Specifically, the key idea of rSAX is to generate a group of breakpoints by random shifting rather than definite breakpoints. Therefore, the points close to each other will have higher probabilities to be mapped into the same symbols, while the points far away from each other will have higher probabilities to be mapped into different symbols. In addition, we also theoretically prove that rSAX can achieve better mapping performances and TLB than SAX. Finally, extensive experiments on several real-world data sets clearly validate the effectiveness and efficiency of the rSAX approach.

Symbolic Representation Algorithm for Time Series Based on Sliding Window and Local Features

A novel piecewise linear segmentation for time series

Entropy-based Symbolic Aggregate Approximation Representation Method for Time Series

A New Representation and Distance Measure for Financial Time Series

A Novel Method for Time Series Symbolization Based on Singular Event Features Clustering

Time Series Symbolization Based on Singular Event Feature Clustering

Dynamic Symbolization of Streaming Time Series

Integrating Data-Driven Segmentation, Local Feature Extraction and Fisher Kernel Encoding to Improve Time Series Classification

Towards Unsupervised Time Series Representation Learning: A Decomposition Perspective

Accurate Symbolization of Time Series

Power Quality Time Series Clustering Analysis Based on Symbolic Feature Representation Method

Entropy-Based Symbolic Representation for Time Series Classification

Representation and analysis of time-series data via deep embedding and visual exploration

Piecewise Statistic Approximation Based Similarity Measure for Time Series

Time-Series Classification Through Histograms of Symbolic Polynomials

Research on Vector Symbolization of Traffic Flow Time Series by Data Mining Oriented Method

An integrated approach implementing sliding window and DTW distance for time series forecasting tasks

A Novel Trend Symbolic Aggregate Approximation for Time Series

Satellite Images Analysis with Symbolic Time Series: A Case Study of the Algerian Zone

Imaging feature-based clustering of financial time series

Time series representation: a random shifting perspective