Author(s): Olga V. Kitova,Ludmila P. Dyakonova and Victoria M. SAvinova
Aim of the study: Currently, decision support systems in economic management that have the ability to predict socio-economic indicators are relevant. One of the key areas is the development of time series scenario forecasting models. Various classical econometric and intellectual methods are used to solve this problem. In earlier studies, the authors cited the construction of hybrid models based on multiple linear regression, artificial neural networks, the support vector machine, and regression decision trees. The purpose of this work is to study the possibility of using neuro-fuzzy models for predicting indicators of the social sphere of Russia. Methodology: The main tasks are to build an ensemble of models based on the methods of multiple linear regression, artificial neural networks of direct propagation (perceptrons) and the ANFIS module, to verify the obtained models, and to compare the results of forecasting. Conclusion: The use of an artificial neural network has a significant drawback, which is the complexity of the interpretation of the computation results. Revealing the knowledge of the model is a complex and time-consuming task. The ANFIS model incorporates the advantages of neural networks, and also eliminates the problem of interpreting the results through the use of a fuzzy logic apparatus. In the works of many leading scientists, the results of the application of a neuro-fuzzy model for forecasting time series in various fields are presented. However, the works do not investigate the possibility of predicting the socio-economic indicators of the state in order to improve the quality of strategic decisions.