The key of case-based reasoning process (CBR) is the case retrieval. With the increase of the number of cases, the efficiency of the case retrieval decreases. In order to ensure efficiency and stability of the CBR system, related clustering algorithms are introduced to make effective classification and improve the efficiency of the case retrieval. Results of many clustering algorithms are affected by selection of initial values. For example, different results of classification may be produced with the same case library. Therefore, it is probable that the most similar cases cannot be retrieved and the optimal number of case categories cannot be determined in common clustering algorithms. A case retrieval algorithm based on correlation analysis is proposed according to the process of case-based reasoning. In the algorithm, the gray correlation analysis is adopted to classify cases stored in the case library of the CBR system, and the method oft-distribution in mathematical statistics is adopted for determining the optimal number of case categories. Then, corresponding algorithms for case classification and retrievals are designed. Finally, comparison experiments are made to verify the stability and effectiveness of the algorithm. Theoretical analysis and experimental results show that with the number of cases in the case library, the algorithm has better stability and efficiency compared with classic algorithms. The model of the algorithm has some practical value.