Classification and Prediction of Customer Categories Using Combination of LRFM Method, Quartiles and Multi-Class Data Mining Methods

Today, Customer Relationship Management (CRM( has become an inevitable necessity in organizations. However, one of the primary problems in this area is the lack of a clear criterion for classifying customers. Creating predictive models from customer categories is also one of the challenges of customer relationship management. For this purpose, in this paper, the combination of LRFM method with quartile concepts and multi-class data mining methods was used to categorize customers. In this solution, customer information records were first reviewed and refined to remove invalid data. Customers were then categorized by combining the concepts of the LRFM method with quartiles. In order to identify and classify new customers, the obtained outputs are subjected to feature selection and their additional features removed. Then, the remaining features were transferred to different classifiers to create predictive customer category models. Contrary to previous research achievements, in this paper, multi-class microscopic and macroscopic evaluation criteria were used to evaluate the performance of predictors. In order to evaluate the results, in a case study, part of the customer information of Sarchin Khorasan Company was used. Evaluations showed that the accuracy of the predicted models of the proposed method in customer classification was higher than the conventional method of customer classification using K-Means. In addition, the proposed method performed better in group detection of records.