Algorithm of Division into Training and Testing Samples in the Construction of Adaptive Neuro-fuzzy Network

The question of division of the generalized sample into training and testing is considered. A division algorithm based on the Chi-square criterion is described. The work of the algorithm in two experiments is demonstrated by the example of a hybrid adaptive neuro-fuzzy network. In the first experiment, the total sample is divided into testing and training with minimal difference. In the second experiment, the total sample is divided into testing and training with the maximum difference.