The development of Japanese language education provides a solid foundation for the friendly relations between China and Japan. In this paper, we first constructed a framework for optimizing teaching in Japanese language education based on big data and intensive learning intervention decision-making. We designed an achievement prediction model based on deep learning for Japanese language teaching effectiveness. The achievement prediction model is applied to the teaching achievement prediction and compared with the actual achievement, and its relative error range is between 0~0.2, compared with the traditional prediction model error reduction of about 0.2 or so in the Japanese language teaching achievement prediction analysis, the confidence and validity of the learning law and class performance are more significant than 0.8, the association rule is superior, and the optimized student prediction of the Japanese language teaching course learning achievement is basically distributed between 80 and 100 points, and the achievement level was significantly improved. This study provides an effective method for the precise reform of Japanese language teaching and realizes the prediction of the effectiveness of educational reform.