University Graduate Employment Quality Evaluation Based on Attention Mechanism Fusing Temporal Convolution and Long Short-Term Memory

In addressing the issues of inaccurate evaluation results and long evaluation times in college graduate employment quality assessments, this paper proposes an innovative data augmentation network model, ATCLSTM-TimeGAN. The model integrates attention mechanisms, temporal convolutional networks (TCN), and long short-term memory (LSTM) networks, and utilizes an adaptive mechanism to resolve the information loss problem present in traditional TimeGAN models. The model is tested using the Appliances dataset, and experimental results confirm that the proposed data augmentation method is effective in time series prediction tasks, particularly under small sample conditions. Furthermore, the paper demonstrates the superiority of the ATCLSTM-TimeGAN algorithm in the context of college graduate employment quality evaluation models. The experimental results show that the data generated by ATCLSTM-TimeGAN better matches the distribution of the original data and significantly improves both evaluation accuracy and efficiency. The evaluation accuracy exceeds 90%, with an average evaluation time of only 4.40 seconds, outperforming traditional LSTM-TimeGAN methods.