The Development of a CFM Hybrid Artificial Sale Forecasting Model

Recently, the number of Convenience Store (CVS) increases dramatically and CVS’s business operation is facing various competitive environments in Taiwan. Sales forecasting of daily fresh foods of CVS is highly complex due to the influence of internal and external environments. However, reliable sales forecasting can improve the quality of decision making and increase its competitiveness. The purpose of the study is to develop a hybrid artificial intelligent sales forecasting model of daily fresh foods for CVS. The Self Organization Map (SOM) neural network and Back Propagation Neural Network (BPNN) is combined to build a hybrid artificial model called Cluster and Forecast Model (CFM). The model is evaluated by a half-year sales data set of daily fresh foods of a chained CVS in Taiwan. The results of the proposed model are compared with the results of BPNN, Fuzzy Neural Network (FNN), Generic Algorithm (GA) and linear logic regression. The results reveal that the proposed model can not only solve the over fit problem and incremental data rescanning but also forecast brand new fresh foods efficiently.