Applying General-Purpose Data Reduction Techniques for Fast Time Series Classification

Abstract

The one-nearest neighbour classifier is a widely-used time series classification method. However, its efficiency depends on the size of the training set as well as on data dimensionality. Although many speed-up methods for fast time series classification have been proposed, state-of-the-art, non-parametric data reduction techniques have not been exploited on time series data. This paper presents an experimental study where known prototype selection and abstraction data reduction techniques are evaluated both on original data and a dimensionally reduced representation form of the same data from seven time series datasets. The results show that data reduction, even when applied on dimensionally reduced data, can in some cases improve the accuracy and at the same time reduce the computational cost of classification.