A performance comparison between classification techniques with CRM application

Complaints Management (CM) is one of the important elements in Customer Relationship Management (CRM) system of any organization which helps in customer retention for the longest possible period of time. In this research, a system called Complaint Classification System (CCS) is implemented to discuss how Data Mining Techniques (DMT) can be used to classify and direct complaints to the departments responsible for them. This may help to renew the client confidence with the organization. To achieve this, many algorithms are used in classification and are compared to use the most efficient of them in the practical system. The used algorithms are the centroid-based classifier, the Voting k-Nearest Centroid Neighbor (VK-NCN) algorithm, the Weighted k-Nearest Centroid Neighbor (WK-NCN) and the Local Mean K-Nearest Centroid Neighbor (LM-KNCN) algorithm. Centroid-based classifier algorithm is proved to have the best performance during the usage of cosine measure while LM-KNCN algorithm is proved to have the best performance during the usage of Euclidean distance measure.