International Journal of Industrial Engineering & Production Management

  In today’s world, customer purchasing behavior forecasting is one of the most important aspects of customer attraction. Good forecasting can help to develop marketing strategies more accurately and to spend resources more effectively. The creation of a customer recognition system (CRS) model concerns a difficult task due to the large number of possible features. Furthermore, there is a high need to create a CRS that have both low complexity and good forecasting abilities at the same time. Thus, the purpose of this paper is to develop a hybrid CRS (HCRS) model that is computationally efficient and effective. The novelty of the HCRS lies in the design and implementation of the mentioned system by combining a pruned regression tree (PRT) that increases computational speed to select suitable subset of features and designing an improved Feedforward neural network (IFFNN) that practically provides better forecasting results. Since customer identification is one of the concerns in insurance industry, an insurance company dataset has been used. The obtained results show that the HCRS selects just 7% of the available features in this way considerably reducing computation costs. In addition, the results show that making the IFFNN led to more accurate forecasting than the methods compared .