At present, the existing dynamic early-warning model of college students' psychological crisis cannot obtain psychological characteristics, and the error of psychological crisis prediction is large. The early-warning result of psychological crisis appears big deviation. A new dynamic early-warning model of college students' psychological crisis is designed based on characteristic attributes. In the process of modeling negative psychological crisis of college students, the fusion constraint fuzzy theory obtains the essence of college students' psychology from subjective and objective reasons, gives the different stages of college students' psychology, obtains the corresponding information entropy of the behavior hindrance in each stage of college students' psychology, and calculates the threshold value of the behavior tendency of college students' psychology. Based on the psychological state information of college students obtained by perception, the fuzzy comprehensive early-warning data model of feature attributes is constructed. According to the fuzzy comprehensive early-warning method and the early-warning index system, the early-warning factor set of college students' psychological crisis is constructed, and the dynamic early-warning model of college students' psychological crisis is constructed. Experimental results show that the mining results are consistent with the actual situation, and the grey relational degree, small error probability, and mean square ratio index are better than the present model.