SEGMENTASI PELANGGAN MENGGUNAKAN METODE DBSCAN UNTUK MENDETEKSI POLA BELANJA

Abstract

Customer segmentation is one approach used to identify customer characteristics. Accurate segmentation allows companies to personalize offers, increase customer retention and optimize marketing costs. The purpose of this study is to group customer characteristics of a retail company using the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method. The DBSCAN method does not require initial determination of the number of clusters, is able to recognize clusters with irregular shapes and can identify outliers or customers with extreme patterns. The dataset used is an external dataset obtained from Kaggle. The dataset contains customer personalization analysis with a total of 2,240 rows and 29 columns. The results of the study show that the DBSCAN method can produce an eps value of 1.2 and produce the highest Silhouette Score of 0.080 with 4 clusters formed. Visualization of segmentation results with PCA dimension reduction techniques into two dimensions to facilitate interpretation. The PCA visualization produces 5 clusters, each of which represents its respective customer group. Thus, this approach offers an adaptive segmentation alternative that is more sensitive to complex behavioral patterns.

Keywords : Customer segmentation, DBSCAN, shopping patterns, silhouette scorem, PCA