Analysing Clustering Algorithms Performance in CRM Systems

Customer Relationship Management technology plays an important role in business performance. The main problem is the extraction of valuable and accurate information from large customers' transactional data sets. In data mining, clustering techniques group customers based on their transaction's details. Grouping is a quantifiable way to analyse the customers' data and distinguish customers based on their purchases. Number of clusters plays an important role in business intelligence. It is an important parameter for business analysts. In this paper the performance of K-means and K-medoids algorithm will be analysed based on the impact of the number of clusters, number of dimensions and distance function. The Elbow method combined with K-means algorithm will be implemented to find the optimal number of clusters for a real data set from retail stores. Results show that the proposed algorithm is very effective when customers need to be grouped based on numerical and nominal attributes.