МОДЕЛИРОВАНИЕ ТЕНДЕНЦИЙ РАЗВИТИЯ СОЦИАЛЬНО-ЭКОНОМИЧЕСКИХ СИСТЕМ НА ОСНОВЕ НЕЙРОННЫХ СЕТЕЙ

Abstract

When forecasting socio-economic processes the question of choosing a forecasting model that allows the most objectively reflect the development trends of socio-economic systems appears relevant. The most common forecasting models are: linear model, quadratic model, exponential model, autoregressive model, Holt-Winters model. They are based on extrapolation - extending into the future the trend observed in the past and present. The fundamental non-linearity of socio-economic systems that are managed under conditions of uncertainty and incomplete observability of processes of functioning, allowed to formulate the appropriateness of their prediction using neural networks. Comparative evaluation of the effectiveness of different prediction models was carried out by the example of forecasting of socio-economic development of the Astrakhan region in 2014-2015. As the initial data there were taken the following figures: GRP, industrial output, gross agricultural products, the amount of investments in fixed assets, volumes of construction works, the average monthly wages, the consumer price index, unemployment rate in 2001-2012. The obtained minimum values of the error in forecasting social-economic development of the region based on neural networks indicate a higher degree of objectivity of the results of neural network prediction, compared to other models.