Recently, most researches about case-based reasoning are focused on case retrieval, while little attention is paid to the method of construting a proper case base. Better case base can improve both the efficiency and accuracy of case retrieval. Therefore, the construting of the case base is addressed in this paper. By removing redundant cases via FCM, neural network-CBR (NN-CBR) is improved. The simulation experiments conducted with UCI data verify the improvement of the proposed method in both accuracy and efficiency.