An integrated fuzzy mathematical programming-analysis of variance approach for forecasting gasoline consumption with ambiguous inputs: USA, Canada, Japan, Iran and Kuwait

Gasoline as the most important vehicle's fuel has a direct effect on economic development. In this study a fuzzy mathematical programming-analysis of variance approach is proposed to forecast gasoline consumption in the USA, Canada, Japan, Iran and Kuwait. The approach of this study utilises gross domestic production (GDP), annual population, number of vehicles and actual price of gasoline as the most standard independent variables. In this algorithm, gasoline consumption data from 1992 to 2005 for five mentioned countries are used to show its applicability. Proposed approach can select the best regression model between fuzzy and conventional methods for each country by means of analysis of variance (ANOVA), simultaneous Turkey test and mean absolute percentage error (MAPE). Results show that fuzzy regression provides better solution than conventional approaches. Moreover, it has more applicability toward gasoline consumption because it considers uncertainty and ambiguousness within the inputs and data sets. This is the first study that considers an integrated fuzzy mathematical programming-regression-ANOVA for gasoline consumption with uncertain inputs in both developed and developing countries.