An emotional learning-neuro-fuzzy inference approach for optimum training and forecasting of gas consumption estimation models with cognitive data

This study introduces an optimum training and forecasting approach for natural gas consumption forecasting and estimation in cognitive and noisy environments by an integrated approach. The approach is based on emotional learning based fuzzy inference system (ELFIS), artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and conventional regression. Results are compared to show the suitability of the optimum training model in noisy and uncertain environment. The designated forecasting models use standard inputs and gas demand as their output. The training approach utilizes intelligent and emotional learning mechanism. Furthermore, analysis of variance (ANOVA), mean absolute percentage error (MAPE), normalized mean square error (NMSE) and Duncan's multiple range test (DMRT) are used to test a set of hypothesis and to select the optimum training model. Applicability and superiority of the approach is shown through applying the above models on actual gas consumption data in Iran from 1973 to 2006. The approach is capable of modeling sharp drops or jumps in consumption with appropriate cognitive and emotional signals. This is the first study that uses an integrated approach for optimum training of gas consumption estimation with noisy and cognitive data.