At present, in the task of processing multivariate time series classification, the RNN (Recurrent Neural Network) structure has been widely used. The TCN (Temporal Convolutional Network) performs better than the RNN structure in the processing of time series data. TCN uses a causal convolution neural network and has a strong memory capacity for past and future information. At the same time, because TCN can perform parallel computing, the processing speed is also very fast. Therefore, TCN has great advantages when dealing with time series with strong time correlation, which is the reason why TCN is introduced in this paper. Many models do not take into account the interrelationship between different variables when making classification predictions on multivariate data. In order to enhance the model’s feature extraction degree in variable dimensions, this paper proposes a variable attention mechanism to solve this problem. Therefore, this paper will use the VATCN (Variable Attention Mechanism TCN) network that combines TCN and variable attention mechanism to study the classification of multivariable time series from the time and variable dimensions.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Multivariate time series classification based on time and variable dimensions


    Contributors:
    Xu, Jinjing (author) / Zhang, Bin (author)


    Publication date :

    2021-10-20


    Size :

    798934 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English