Time Series Classification Using Compression Distance of Recurrence Plots

Abstract

Abstract-There is a huge increase of interest for time series methods and techniques. Virtually every piece of information collected from human, natural, and biological processes is susceptible to changes over time, and the study of how these changes occur is a central issue in fully understanding such processes. Among all time series mining tasks, classification is likely to be the most prominent one. In time series classification there is a significant body of empirical research that indicates that k-nearest neighbor rule in the time domain is very effective. However, certain time series features are not easily identified in this domain and a change in representation may reveal some significant and unknown features. In this work, we propose the use of recurrence plots as representation domain for time series classification. Our approach measures the similarity between recurrence plots using Campana-Keogh (CK-1) distance, a Kolmogorov complexitybased distance that uses video compression algorithms to estimate image similarity. We show that recurrence plots allied to CK-1 distance lead to significant improvements in accuracy rates compared to Euclidean distance and Dynamic Time Warping in several data sets. Although recurrence plots cannot provide the best accuracy rates for all data sets, we demonstrate that we can predict ahead of time that our method will outperform the time representation with Euclidean and Dynamic Time Warping distances.

Extracting Texture Features for Time Series Classification

“…Even without extracting texture features, the most similar work to ours is presented by Silva et al [2] with the Recurrence Patterns Compression Distance (RPCD) for time series classification. The RPCD applies a video compression based distance measure (CK-1) in an 1-NN algorithm to estimate the similarity between two time series represented by recurrence plot.…”

“…In summary, TFRP uses a Support Vector Machine (SVM) algorithm with four techniques for extracting texture features from recurrence plots. We show in our experimental evaluation with 38 time series data sets that TFRP is very competitive with state-of-the-art methods such as 1-NN with Euclidean distance, Dynamic Time Warping and Recurrence Patterns Compression Distance (RPCD) [2]. RPCD is our previous attempt to classify time series using recurrence plots and CK-1 [3], a distance measure between images that uses video compression algorithms.…”

“…These data sets have standard partitions of training and testing, further facilitating the execution of experiments and the comparison of results. Similarly to [2], we decided to exclude all synthetic data sets since they were generated for a specific purpose or algorithm. Thus, we use 38 data sets in our experiments.…”

Recurrence Based Similarity Identification of Climate Data

“…Furthermore, synchronicity and convergence are also examined among member nations of the Euro region for GDP using cross recurrence analysis [25]. Silva et al [26] gave an overview of recurrence plots as a representation domain for time series classification, in which CampanaKeogh (CK-1) and Kolmogorov complexity based distances are used to measure the closeness between recurrence plots and to estimate image similarity, respectively.…”

FESM: An Analytical Framework for Elastic Similarity Measures Based Time Series Pattern Recognition

“…There are structural patterns in recurrence plots which can be used to determine the similarity between two video sequences, which is necessary for classification, so measuring the similarity between two recurrence plots, needed for video classification in order to event detection in video sequences. In [5] the CK-1 distance used to measure the similarity between unthresholded recurrence plots that were generated from time series and results show the combination of the CK-1 distance measure together with unthresholded recurrence plots results in higher classification accuracy for time series which represent the shape. Related to energy and power [6], check the effect of similarity measures is a necessary step for the optimized design and development of efficient clustering based models, predictors and controllers of time dependent processes such as building energy consumption patterns.…”