In the environment of education big data, how colleges and universities make good use of these data not only affects the orderly operation of the whole education and teaching system of colleges and universities. It will also become an inexhaustible driving force to help colleges and universities promote the reform and innovation of the education and teaching system. This paper takes student evaluation data and student online learning data as research objects. Focusing on the teaching operation and students’ autonomous learning, this paper uses the improved k-mode algorithm to cluster analyze the classroom teaching operation. This paper uses neural network algorithm based on machine learning to predict and compare students’ online course learning. It is hoped that it can provide meaningful reference for the construction of teaching management system and the reform and innovation of teaching management system in colleges and universities. Two research works are mainly carried out through the preliminary analysis and transformation of the data of student evaluation of teaching in a certain university. The improved cosine dissimilarity algorithm is used to eliminate the abnormal teaching evaluation data. The normalization method was used to standardize the teaching evaluation data. The traditional k-mode algorithm is used to cluster the teaching evaluation data. Some problems of k-mode algorithm are pointed out, and the traditional k-mode algorithm is improved. Experimental results show that the improved algorithm is more reasonable and effective.