An empirical survey of data augmentation for time series classification with neural networks.

Brian Kenji Iwana,Seiichi Uchida,Friedhelm Schwenker

doi:10.1371/journal.pone.0254841

Abstract

In recent times, deep artificial neural networks have achieved many successes in pattern recognition. Part of this success can be attributed to the reliance on big data to increase generalization. However, in the field of time series recognition, many datasets are often very small. One method of addressing this problem is through the use of data augmentation. In this paper, we survey data augmentation techniques for time series and their application to time series classification with neural networks. We propose a taxonomy and outline the four families in time series data augmentation, including transformation-based methods, pattern mixing, generative models, and decomposition methods. Furthermore, we empirically evaluate 12 time series data augmentation methods on 128 time series classification datasets with six different types of neural networks. Through the results, we are able to analyze the characteristics, advantages and disadvantages, and recommendations of each data augmentation method. This survey aims to help in the selection of time series data augmentation for neural network applications.

Highlights

Time series classification attempts to categorize time series into distinct categories, and it is used for a wide range of applications
Recent work has shown that feedforward networks such as Multi-Layer Perceptrons (MLP) and temporal Convolutional Neural Networks (CNN) [11] can achieve competitive and sometimes better results for time series recognition [3, 12, 13]
It is notable that the data augmentation with MLP and Long Short-Term Memory (LSTM)-Fully Convolutional Network (FCN) was mostly detrimental, sometimes significantly

Summary

Introduction

Time series classification attempts to categorize time series into distinct categories, and it is used for a wide range of applications. Artificial neural networks have had many successes in time series classification [2, 3]. Recurrent Neural Networks (RNN) [4] have had many recent successes on time series in gait recognition [5, 6], biosignals [7, 8], and online handwriting [9, 10]. Recent work has shown that feedforward networks such as Multi-Layer Perceptrons (MLP) and temporal Convolutional Neural Networks (CNN) [11] can achieve competitive and sometimes better results for time series recognition [3, 12, 13]. It has been shown that increasing the amount of data can help with improving the generalization ability as well as the overall performance of the model [14, 15]

Objectives

Results

Conclusion