DEVELOPMENT OF DATA PROCESSING METHOD FOR MODELING AND FORECASTING GAS CONSUMPTION IN THE REGION

Guriev Ruslan Makharbekovich, Dzgoev Alan Eduardovich, Karatsev Stanislav Taimurazovich

Development of methods for forecasting the aggregate monthly demand for consumer gas in the region. Modeling and forecasting of future gas consumption using the method of least squares and classical correlation and regression analysis is carried out. In the course of the study, factors affecting the monthly demand for natural gas were identified. Eight regression models have been developed: linear and nonlinear equations of the second degree. The analysis of independent variables for intercorrelation and the analysis of the dependent variable for autocorrelation is carried out. The criteria of adequacy and quality of models are calculated. An assessment of adequacy by Fischer's F-criterion was carried out. The significance estimates for the main regression coefficients are determined: Correlation coefficient corr(Y,YR); Coefficient of determination R2 and R2 adjusted; Standard error (MSE); Average modeling error (MAPE). New useful adequate regression models have been identified, on the basis of which forecast estimates of gas consumption are calculated 30% ahead of the initial data.

Keywords:small data sample, regression model, least squares method.