@Article{JICS-15-146, author = {Lei Zhao}, title = {Research on user behavior recognition based on 2D-CNN}, journal = {Journal of Information and Computing Science}, year = {2024}, volume = {15}, number = {2}, pages = {146--153}, abstract = {In this paper, we studied the application of two-dimensional convolutional neural networks to the classification of multivariate time series. Time series sample data is usually a set of measurement values of a single attribute or multiple attributes at continuous time points separated by uniform time intervals. It is a set of structured data, usually non-discrete, time-related between data Features such as sex, feature space, and large dimension. At present, most methods for time series classification problems need to go through an extremely complex data preprocessing process and related feature engineering and do not consider the long pattern information hidden in different time dimensions of time series data, and the different characteristics of multivariate time series data Relevant information in the space dimension between. By converting multivariate time series data into matrix form, this paper proposes an end-to-end deep learning model Pyramid-CNN based on two-dimensional convolutional neural networks, which uses two-dimensional convolution kernels to extract the spatial dimensions of multivariate time series data and the relevant information in the time dimension, and applied it to the user behavior recognition time series data set. The experimental results show that for this data set, compared with the existing methods, the model proposed in this paper has higher performance Accuracy and robustness, with a good classification effect. }, issn = {1746-7659}, doi = {https://doi.org/}, url = {http://global-sci.org/intro/article_detail/jics/22390.html} }