In this study, we applied a modified rule table method to a classification problem by comparing its performance and features with those of a feedforward neural network (FNN). In general, classification problems that predict the output against a new input are considered after the learning process using the input and output pairs of the learning dataset. We adapted an existing rule table method for the problems to correspond with the FNN, which uses a criterion to minimize the cross entropy. Our results confirmed the usefulness of the modified method in discussing the common and different features between them.