Feature extraction by grammatical evolution for one-class time series classification

Stefano Mauceri,James Mcdermott,James Sweeney,Miguel Nicolau

doi:10.1007/s10710-021-09403-x

Abstract

When dealing with a new time series classification problem, modellers do not know in advance which features could enable the best classification performance. We propose an evolutionary algorithm based on grammatical evolution to attain a data-driven feature-based representation of time series with minimal human intervention. The proposed algorithm can select both the features to extract and the sub-sequences from which to extract them. These choices not only impact classification performance but also allow understanding of the problem at hand. The algorithm is tested on 30 problems outperforming several benchmarks. Finally, in a case study related to subject authentication, we show how features learned for a given subject are able to generalise to subjects unseen during the extraction phase.

Highlights

Feature extraction should normally be customised to the problem at hand
We investigate grammatical evolution (GE) [43], an evolutionary algorithm related to genetic programming (GP) [2, 27], as a means to achieve data-driven feature extraction from time series in the context of classification
We propose a data-driven evolutionary algorithm for feature extraction from time series

Summary

Introduction

Often, when modellers deal with a new problem they do not know which features could enable the best classification performance. A common approach is to construct an initial set of features, and select the subset yielding best performance [34]. In contrast to the manual approach, there is growing interest in algorithms that enable the data-driven discovery of features, as made possible by deep learning methods [28]. The advantage is that modellers can redirect their efforts from the construction of the solution to the construction of the learning framework. While the former may be useful solely on a particular problem the latter may be effective on many

Objectives

Methods

Results

Conclusion