Customer segmentation with RFM models and demographic variable using DBSCAN algorithm

Abstract

The aims of this research was to identify prospective customers by conducting customer segmentation based on recency, frequency, monetary (RFM) values and demographic variables. The step were selected the data and normalized. The normalized data were clustered using the density based spatial clustering of applications with noise (DBSCAN) algorithm. The k-dist graph was utilized with RStudio tools to identify the best values for epsilon and MinPts. The outcome of utilizing epsilon 0.06 and MinPts 3 was the identification of 5 clusters and 31 data points considered as noise, resulting in a silhouette index (SI) value of 0.4222. Based on the average RFM values, cluster 1 was categorized as prospective customers, while clusters 2, 3, 4, and 5 were designated as loyal customers. Furthermore, according to demographic analysis, the majority of customers are between the ages of 35 to 45, female, married, and housewives. Women, groceries, such as rice and cooking oil, were the most popular products. Besides, the customers were mostly lecturers and lived in Pekanbaru. This was compatible with the customer target of people from upper middle class, such as lecturers, and with the location of the mart as well, which was near a campus.