Customer Clustering by Combining the Particle Swarm And K-Means Algorithms and Analyzing Their Behavior on Commercial Websites
الموضوعات :
MohammadReza Mehrazma
1
,
Behrad Mahboobi
2
1 - Mechanical, Electrical and Computer faculty, Department of Electronic and Communication, Science and Research Branch Islamic Azad University, Tehran, Iran
2 - Communication, Computer and Industrial Network Research Center, Science and Research
Branch Islamic Azad University, Tehran, Iran
الکلمات المفتاحية: Clustering, k-means clustering, Particle Swarm Optimization (PSO), Commercial Websites, Customer Behavior Analysis,
ملخص المقالة :
One of the trends in modern management is the consideration of the principle of customer orientation and customer satisfaction. Following this approach, online shopping, as one of the goods and services distribution channels, is also inclined to maintain and expand relations with customers. Nonetheless, what has become even more highlighted in the competition arena is going beyond customer satisfaction by predicting the customers’ behavior in order to properly respond to their needs and ultimately, establish loyalty. One of the methods used to know the customers is the clustering approach. Clustering is a data mining technique that takes a number of items and places them in clusters based on their attributes. One of the problems of the k-means clustering is that it has no specific method for primary determination or calculation of the cluster centers. Therefore, in order to optimize the clusters, we use the particle swarm optimization (PSO) algorithm. In the end, we analyze these clusters using the RFM model to analyze the customers’ behavior. What is achieved by analysis of each cluster is finding the cluster of the most loyal customers. In this study, we specify the number of optimized clusters using the k-means clustering algorithm. Then, we use the obtained number of clusters for the primary adjustment in the particle swarm optimization algorithm. Finally, we score the achieved clusters by the RFM method so that we can identify the most loyal customers that are placed in the cluster with the highest score.
