This study explores the mechanism and moderating effect of data-driven psychological service models on enhancing college students’ stress coping efficacy, providing methods for improving college students’ mental health. This article collects data through a questionnaire survey and uses the cutting-edge Variational Mode Decomposition Residual Long Short Term Memory Network (VMD-TCN-LSTM) evaluation model to analyze the data, studying the relationship between college students’ understanding of social support, stress coping efficacy, coping efficacy, and stress perception. At the same time, after modifying a certain feature value of the model, observe the degree of decrease in the evaluation effect, determine the key influencing factors, and compare the evaluation effect of the model with other algorithms. The experimental results show that the proposed VMD-TCN-LSTM model performs outstandingly in evaluating the stress coping efficacy of college students, with higher accuracy, G-mean, and F1 score than other benchmark algorithms. After shuffling each feature, the differences in consumption level and consumption behavior have a significant impact on the model’s performance. Research has shown that the data-driven psychological service model can effectively enhance college students’ sense of stress coping efficacy through mechanisms such as precise identification, personalized intervention, and dynamic monitoring, as well as effects such as social support, self-regulation of coping efficacy, and feedback regulation of psychological services.
