This paper first outlines the concept of teaching Japanese grammar according to the Japanese language talent cultivation standard and the current dilemma in teaching Japanese language in colleges and universities. Secondly, on the basis of big data technology, considering the effective availability of the initial data of Japanese grammar, it is necessary to preprocess the language grammar learning data and, at the same time, according to the fuzzy comprehensive evaluation, determine the Japanese grammar evaluation affiliation function. Then, for the problem of an incomplete and inaccurate traditional fuzzy comprehensive evaluation, they proposed to construct a comprehensive evaluation model of the Japanese linguistics grammar system based on the bat algorithm and carried out research and analysis on the Japanese linguistics grammar system. The results show that on the evaluation model, the comprehensive evaluation score of the Japanese language grammar system is C=2.65, which belongs to the E2 level. On the statistical analysis, the mean value of all the scores of the concise class was higher than that of the standard Japanese class, and all of them showed significant differences (p=0.002, 0.000, and 0.047, all less than 0.05).