Recurrent Auto-Encoder Model for Multidimensional Time Series Representation

Timothy Wong, Zhiyuan Luo

Abstract:

Recurrent auto-encoder model can summarise sequential data through an encoder structure into a fixed-length vector and then reconstruct into its original sequential form through the decoder structure. The summarised information can be used to represent time series features. In this paper, we propose relaxing the dimensionality of the decoder output so that it performs partial reconstruction. The fixed-length vector can therefore represent features only in the selected dimensions. In addition, we propose using rolling fixed window approach to generate samples. The change of time series features over time can be summarised as a smooth trajectory path. The fixed-length vectors are further analysed through additional visualisation and unsupervised clustering techniques.

This proposed method can be applied in large-scale industrial processes for sensors signal analysis purpose where clusters of the vector representations can be used to reflect the operating states of selected aspects of the industrial system.

TL;DR: Using recurrent auto-encoder model to extract multidimensional time series features

Keywords: recurrent autoencoder, seq2seq, rnn, multidimensional time series, clustering, sensor, signal analysis, industrial application