In this paper, K-means clustering technology is used to process the data of legal litigation cases, and the similarity of different types of learners is calculated and clustered. After that, a personalized recommendation algorithm is used to rate legal litigation cases, accurately recommend learning resources based on the evaluation results, and test the recommended results. The clustering results show that calm students are objective and fair in judging legal cases, and the results of the rating are real and reliable (35-80 points), so various types of legal cases can be recommended for them to evaluate and learn. Weak students are composed of lazy, young, and timid students. The results of their resource rating are generally biased. Challenging students are mostly active males who are more interested in learning about lawsuits, with scores ranging from 35-90, and for whom more complex cases with more points of view can be recommended. Lazy students, whose scores on cases span a wide range (0-95 points), should be recommended a small number of simple learning resources of different categories for their reference and learning. Students’ scores improved by 15-17 points after personalized recommendation learning, proving that the recommendation method of this paper improves students’ learning effect on legal proceedings, which is of great significance to students’ ideological and political education on the rule of law.