A method for generating switching operation sequences based on generative adversarial networks

To address the issues of low efficiency and accuracy in the generation of switching operation tickets, a method for generating switching operation sequences based on Generative Adversarial Networks (GAN) is proposed. Firstly, the SDAR (Sequence Decomposition And Reconstruction) module is used to perform multi-scale decomposition and reconstruction of the original operation task sequences, capturing both the overall and local features of the operation task sequences. Secondly, an A-LSTM (Attention-Long Short Term Memory) network is introduced to enhance the accuracy of the generated switching operation sequences. Finally, the generator of the proposed model is constructed based on the SDAR module and the A-LSTM network. Experimental results show that the proposed model achieves a 47.10% improvement in RMSE and a 4.26% improvement in R² compared to LSTM, and a 27.33% improvement in RMSE and a 1.73% improvement in R² compared to WGAN.