Forecasting accuracy is one of the most effective and most important factors in order to select an appropriate forecasting model. Therefore, one of the most important issues in recent studies is improving accuracy of forecasting. Due to importance of providing accurate and reliable forecasting results in different kind of sciences and especially in decision making processes, various techniques have been developed in order to improve accuracy of forecasting in recent years. Combining different models or using hybrid models is one of the most well-known and widely used methodologies to lift limitations of individual models, using their unique advantages simultaneously and especially improving forecasting accuracy. Obtained results of various studies which have been done in modeling and forecasting areas indicate that considerable improvement can be achieved in producing more accurate forecasting and consequently improvement of decision quality, made by decision- makers and managers by using such that methods. Despite widely-used hybrid models, proposing an efficient and comprehensive combination method is yet one of the vague points in this area. Therefore, the main purpose of this thesis is to design and implement of an efficient and accurate hybrid structure in time series modeling and forecasting and finally disillusion of the introduction an efficient hybrid structure in the literature of hybrid models. Thus, in the first step, the possible hybrid structures of individual models are designed and by using the basic concepts of the statistical ltr"
