Journal of Operational Research and Its Applications ( Applied Mathematics ) - Lahijan Azad University

In recent decades, due to the continuous increase in the cost of attracting new customers, it is very important and sensitive for the profitability of organizations to pay attention to maintaining customers and increasing their loyalty. Therefore, organizations implement various programs to increase the durability of their valuable customers (customers with less resource loss and high profitability). The current research, considering the capabilities of data mining in management and design, implements a model for predicting the behavior of customers in the field of industry, using the CRISP-DM standard methodology based on the RFM model and Random Forest and Growing Trees techniques. Increasingly, it has searched the database of customers of an automobile company, who have had more than one product purchase contract with that automobile company. By applying a model based on Random Forest, Growing Trees and a hybrid prediction model technique, customers who tend to turn away are identified and effective marketing strategies are planned for this group. The analysis of customer behavior shows that the length of active customer relationship, the frequency of relative purchases and the average time interval between purchases are among the best predictors. Also, the hybrid prediction technique has shown a better response than random forest and growing trees techniques.