This study proposes a case-based reasoning model for corporate bankruptcy prediction and compares its performance with those of traditional statistical models and artificial intelligence models. While prior studies considered only the differentiation of input variable weights, this study additionally considers the covariance structure of input variables. Experiments were conducted using data from small and medium-sized manufacturing firms collected from 2001 to 2003. For the case-based reasoning model, a total of four models were applied by considering two experimental factors (considering/not considering the covariance structure among variables × differentiating/not differentiating variable weights). The number of neighbors included in the nearest neighbor was determined to be 15 through a search process. Using a simulation technique, the model yielding the best results on the training data was selected, and then the model performance was compared on the validation data. The experimental results showed that the case-based reasoning model incorporating both the covariance structure among variables and variable weights exhibited higher prediction accuracy than existing models.