In complex multi-machine,multi-task parallel production environments,capacity prediction faces significant challenges due to large-scale data and nonlinear spatiotemporal feature coupling.Existing methods,such as convolutional neural networks (CNN),gated recurrent units (GRU),long short-term memory networks (LSTM),and CNN-LSTM hybrid models,often focus on extracting isolated features,failing to fully exploit spatiotemporal correlations.This limitation leads to weak generalization and insufficient prediction accuracy.To address these issues,this study proposes an attention-based CNN-LSTM hybrid model (CNN-LSTM-Attention) and validates its performance using real-world production data from a factory workshop spanning 2018 to 2022.Experimental results evaluated with metrics including mean absolute error (MAE),root mean square error (RMSE),symmetric mean absolute percentage error (sMAPE),and accuracy demonstrate that the proposed model achieves over 0.9 accuracy in both short-and long-term predictions,significantly outperforming conventional methods.The findings verify the model′s feasibility and effectiveness in practical applications.