Feature Extraction and Selection Techniques for Time Series Data Classification: A Comparative Analysis

Abstract

Listen

Translate article

Recently, time-series data mining has attracted tremendous interest and initiated various researches in real-time high dimensional data like, Stock market, Electrocardiogram, Electroencephalogram signal, noise detection, cryptocurrency, weather, and etc. Extraction of features in time series classification is mainly used to overcome the computational problem, storage, and easier to visualize the data. Time series data leads to better decision making in these emerging fields and to build a strong knowledge about the prediction. In this paper, we focused on exploring how time series data classification tests influence various techniques of feature extraction of time-series datasets. Different extraction techniques are studied with different approaches, and then different methods are also investigated. A comparison is made for effectiveness and improvements of classifiers in time series data. The findings suggested that the extraction and selection methods play a vital role in the resulting classification with the individual classification, SVM is the mostly used classifier and no approach outperforms each other consistently. While the main emphasis is on time series classifiers, various feature extraction methods have been discussed.