Dynamic Algorithm Design for Data Mining and Accurate Prediction of Customer Response

The problem of identifying and anticipating potential customers to be addressed at direct marketing programs has been considered as one of the popular and most important marketers' issue. Marketers who use these approaches are threatened by the severe reaction of those consumers who consider the direct marketing as an attack to their private lives, so it may even be possible to boycott companies that use these methods. Neural networks are known as a powerful tool for prediction, but as previously mentioned, as with other prediction algorithms, they tend to deviate toward imbalanced data. In this research, in order to enhance the ability to identify and predict potential customers by multilayer perceptron networks, using Random under-over sampling methods, which has been used frequently in other articles, we attempted cluster customers and create different combinations of them, and then from the observed results, we finally introduced an innovative and highly efficient method for identifying and rating potential customers. The results indicate that, in addition to the undeniable power of multilayer neural networks in the field of identification and prediction, imbalanced data has greatly damaged the results. In this regard, creating an optimal combination of customer data and implementing the innovative algorithm of the present study significantly improves the results and the performance of artificial neural networks has yielded a reliable consequences.