Time series classification using MACD-Histogram-based SAX and its performance evaluation

Time series classification is one of the most well-known grand challenges in many different application domains. Time series classification is the task of assigning a discrete class label to an unclassified time series. Three important key points should be considered in the design of time series classifiers: the feature expression for the time series, the definition of the distance function, and the classification strategy. Many researchers of time series have been focusing on Symbolic Aggregate approXimation (SAX), which is a state-of-the-art feature expression for time series. SAX is a high-level symbolic representation for time series that allows for dimensionality reduction. SAX allows symbol-based approaches, which have been studied in depth to be applied in time series classifiers. In this paper, we propose a novel method for time series classification using a SAX-based symbolic representation. The proposed method includes: Moving average convergence divergence (MACD)-Histogram-based SAX and Nearest Neighbor (1-NN) utilizing the extended Levenshtein distance. To evaluate the performance of the proposed method, we implemented it and conducted experiments using the UCR time series classification archive. The experimental results showed that the proposed method outperforms not only other distance-based 1-NNs, but also other state-of-the-art methods.