Development of a PSO–SA hybrid metaheuristic for a new comprehensive regression model to time-series forecasting

Forecasting has always been a crucial challenge for organizations as they play an important role in making many critical decisions. Much effort has been devoted over the past several decades to develop and improve the time-series forecasting models. In these models most researchers assumed linear relationship among the past values of the forecast variable. Although the linear assumption makes it easier to manipulate the models mathematically, it can lead to inappropriate representation of many real-world patterns in which non-linear relationship is prevalent. This paper introduces a new time-series forecasting model based on non linear regression which has high flexibility to fit any number of data without pre-assumptions about real patterns of data and its fitness function. To estimate the model parameters, we have used hybrid metaheuristic which has the ability of estimating the optimal value of model parameters. The proposed hybrid approach is simply structured, and comprises two components: a particle swarm optimization (PSO) and a simulated annealing (SA). The hybridization of a PSO with SA, combining the advantages of these two individual components, is the key innovative aspect of the approach. The performance of the proposed method is evaluated using standard test problems and compared with those of related methods in literature, ARIMA and SARIMA models. The results in solving on 11 problems with different structure reveal that the proposed model yields lower errors for these data sets.