IMPLEMENTASI K-MEDOIDS DAN MODEL WEIGHTED-LENGTH RECENCY FREQUENCY MONETARY (W-LRFM) UNTUK SEGMENTASI PELANGGAN DILENGKAPI GUI R

Abstract

The k-medoids algorithm is a partition-based clustering algorithm that groups n objects as much as k clusters. The algorithm uses medoids as the center point (partition) of the cluster. Medoids are actual objects that are randomly selected as the most centered object in a cluster so that the k-medoids algorithm is robust against outliers. Grouping objects in cluster analysis based on similarities between objects. Measurement of similarity between objects can use the euclidean and manhattan distances. The use of distance in cluster analysis can affect cluster results. Validation of cluster results using internal validation, namely the silhouette index. The Weighted-Length Recency Frequency Monetary (W-LRFM) model is a model that applies the relative importance (weight) of the LRFM model according to the importance of each variable in the LRFM model. LRFM model is a model used for customer segmentation based on customer behavior which consists of variables length, recency, frequency, and monetary. The relative importance (weight) of the W-LFRM model uses the Analytics Hierarchical Process (AHP) method. The W-LRFM model is used to calculate the Customer Lifetime Value (CLV) of each cluster. The implementation of k-medoids and the W-LFRM model in this study are used for customer segmentation based on the length, recency frequency, and monetary variable. The formation of these variables is the result of transformation of customer behavior data such as transaction id, date of purchase, and a total amount of 41,073 rows into variable length, recency, frequency, and monetary as much as 5,108 rows. The criteria of the best cluster formed are k = 2 using the manhattan distance with the average of coefficient values = 0.62. The weights on the W-LRFM model produced based on the AHP method are 0.16, 0.29, 0.47, and 0.08 for the variable length, recency, frequency, and monetary. CLV formed from two clusters, namely 0.158 and 0.499. CLV in the second cluster is bigger so that the second cluster becomes the main priority in the marketing strategy. The second cluster has the characteristics 0.29, 0.47, and 0.08 for the variable length, recency, frequency, and monetary. The second cluster has the characteristics  means a loyal customer group. The first cluster has characteristics  means a potential customer group. This research is assisted by using Graphical User Interface (GUI) R to facilitate analysis