Abstract: The authors of the article describe the features of the use and advantages of the genetic algorithm for learning the neural-fuzzy network. The authors review the literature sources that consider modifications of the genetic algorithm adapted to solve various problems. The authors found that the existing approaches to the implementation of the genetic algorithm contain a number of drawbacks for learning the neural-fuzzy network, so when forming a chromosome, the interval containing the peak of the membership function is encoded with 1, otherwise it is 0. This affects the resolving power when searching for the optimal solutions. The authors of the article also discuss in detail such aspects as the operators of the basic algorithm and give the scheme of the combined method of learning the network. To teach a neural-fuzzy network based on a genetic algorithm, the authors propose an approach for coding membership functions based on the α-level. As a fitness function of the genetic algorithm, the root-mean-square error was used. The novelty of the research is caused by the fact that the authors offer their own approach of coding membership functions based on the α-level which allows to increase the resolution of the algorithm when searching for the optimal solution. The main conclusion of the research is the fact that the approach proposed in the article makes it possible to adjust the parameters of the membership functions of linguistic variables for the neural-fuzzy network and to obtain more adequate values for the parameters of the output layer of the network.