Multivariate LSTM-FCNs for time series classification

• Published in: Neural networks Vol. 116; pp. 237 - 245
• Main Authors: Karim, Fazle, Majumdar, Somshubra, Darabi, Houshang, Harford, Samuel
• Format: Journal Article
• Language: English
• Published: United States Elsevier Ltd 01.08.2019
• Subjects: Convolutional neural network; Interrupted Time Series Analysis - classification; Long short term memory; Multivariate Analysis; Multivariate time series classification; Neural Networks, Computer; Recurrent neural network; Multivariate time series classification; Recurrent neural network; Convolutional neural network; Long short term memory
• ISSN: 0893-6080; 1879-2782; 1879-2782
• DOI: 10.1016/j.neunet.2019.04.014

Over the past decade, multivariate time series classification has received great attention. We propose transforming the existing univariate time series classification models, the Long Short Term Memory Fully Convolutional Network (LSTM-FCN) and Attention LSTM-FCN (ALSTM-FCN), into a multivariate time series classification model by augmenting the fully convolutional block with a squeeze-and-excitation block to further improve accuracy. Our proposed models outperform most state-of-the-art models while requiring minimum preprocessing. The proposed models work efficiently on various complex multivariate time series classification tasks such as activity recognition or action recognition. Furthermore, the proposed models are highly efficient at test time and small enough to deploy on memory constrained systems.