Machine learning algorithms have shown great potential in predicting vocational education demands. This study compares supervised learning, deep learning, and time series analysis methods, and designs an innovative hybrid model combining LSTM and CNN. The model performs excellently in predicting demands for the first half of 2023, with a MAPE of only 4.3%. By optimizing the model architecture, training efficiency is significantly improved. Experimental results demonstrate that the hybrid model outperforms traditional methods in capturing complex demand patterns, providing a scientific basis for vocational education resource planning and promoting the precise and personalized development of the education system.