In this study, a clustering-based sales forecasting scheme is proposed. The proposed model contains training and
testing stages. In the training phase, the training dataset is divided into several disjoint groups and the SVR is used in
each group to build forecasting model. The detailed procedure of the training phase is as following steps.