Development of an effective adaptive forecasting system based on the combination of neural network and genetic algorithm

Abstract

Abstract The paper is presented an effective adaptive forecasting system based on combining mathematical modeling tools, including neural networks and genetic algorithm. The construction of the neural network structure that is best relative to the selected criterion makes it possible to improve the procedure for finding a solution to the problem in terms of a number of parameters. Each individual in the genetic algorithm is encoded as a vector with data on the number of neurons on the intermediate layers of the neural network. The evolution of the population occurs in the genetic algorithm, information in the chromosomes changes as a result of the probabilistic application of genetic operators. As a result, such a structure of the neural network is formed, at which the convergence to a given level of error of 1.0% is the fastest. Applied calculations were carried out on the monthly statistical data of investments in human capital (education, healthcare and culture) of the Udmurt Republic. The proposed adaptive system, applied to the construction of forecasts of socio-economic indicators, can be used in the construction of development strategies both at the regional level and at the country level.