Finite-differential models used as a basis for IT support in problem-solving

Abstract

The study confirms that the forecast quality of social-economic indicators estimated with multiple linear models does not often seem to be satisfactory. The coefficients of regression-differential models are poorly interpreted economically. This paper covers the issue of improving the forecast quality by modifying regression-differential models to the sum of linear and autoregressive models. For brevity, the resulting models are called finite-differential models of the first-and second-orders, respectively. The authors use the coefficients of multiple linear models for an initial estimate when they determine the unknown coefficients of finite-differential models. The performance of the suggested method is tested with 60 sets of statistical data. The study proves that the use of the first-order finite-differential model results in the significant increase in the quality of forecasting and reduction of the approximation error for 65 per cent of cases, and there are 80 per cent of cases when a second-order finite-differential model is used. This indicates that the authors suggest a valid modification of the algorithm for determining the coefficients of models so that it can be used in further research