Assessment of Region Economic Development on the Basis of Neural Network Model

Abstract

Abstract The paper considers the possibility to use neural network modeling for assessing the economic development of regions exemplified by the Arctic region of the Russian Federation – Murmansk region. The paper presents assessing and reasoning of this usage, describes its opportunities and threats. The author analyzes four indicators as main economic characteristics: gross regional product, an amount of investments into the fixed capital, retail turnover, foreign trade turnover. The study shows which factors have the most significant effect on these characteristics and comments on the obtained results. The author describes methodology of building the model and checks it empirically. In order to assess the model more accurately, it includes autoregression elements, which allows estimating not only direct interaction, but monitoring possible temporary inertia. Assessment results based on neural network modeling are compared with the results obtained on the basis of ADL equations (autoregressive distributed lag model). Accuracy of the final calculations is analyzed with using the mean absolute percentage error (MAPE) in further preference to the model of neural networks.