This paper proposes a values sentiment analysis model based on RoBERTa-BiLSTM-Attention based on the mechanism and path of values process evaluation under sentiment analysis. After extracting and presenting the values and sentiment features data, we build the values dataset of college students, finish the training of the model, and obtain the classification results of values and sentiment. The college students’ value identity model is established to analyze the value shaping of education at the national, social, and individual levels. The correlation coefficients of the influencing factors of college students’ values were analyzed, and the correlation coefficients of the influencing factors were less than 0.75, and the p-value was less than 0.01. The hypotheses were put forward, and the standardized regression coefficient of the “social role model” was 0.2658, which was significantly higher than that of the other indicators, and the standardized regression coefficient of “emotional satisfaction” was 0.2658. The regression coefficient of “emotional satisfaction” is 0.1987, second only to cognitive identity. In view of the existence of the problem of values, the proposed political education intervention guidance strategy, after the political education intervention, 71.3625% of the students think that the values of guidance have a great help, but there are still some shortcomings, compared with this paper puts forward the optimization of the values of guidance recommendations.