E-Commerce Customer Segmentation Using K-Means Algorithm and Length, Recency, Frequency, Monetary Model

Abstract

The growth of the e-commerce business sector presents competitors and creates company competition. Customers are the company's main asset that must be maintained. Understanding the different characteristics, behaviors and habits of each customer segment is important for companies to identify potential customers, establish important strategies, manage customer relationships and increase company profitability. The needs and desires of each customer are different, so that determining a strategy requires a method of segmenting customers according to their respective similarities. Using the clustering method with the K-Means algorithm helps determine customer segmentation based on transaction history data. Determination of the optimal k cluster randomized K-Means doesn't always give good result, so the Elbow, Silhouette and Davies-Bouldin Index methods are used. The determination of the test variables is based on the LRFM model (Length, Recency, Frequency, and Monetary), so that the customer segmentation obtained is more accurate in recognizing customer behavior and loyalty. The results of test 3606 dataset through the preprocessing stage using these methods results in three groups of customers that is New Customers, Lost Customers and Core Customers adjust the Customer Loyalty Matrix LRFM