PRODUCT CLUSTERING ANALYSIS ON THE MARKETPLACE USING K-MEANS APPROACH (CASE STUDY: SHOPEE)

Abstract

The business world has experienced a paradigm shift towards a more modern concept. Many business processes are carried out through the internet or commonly known as e-commerce, by utilizing a platform known as Marketplace. One of the marketplaces that are quite well-known and in great demand in Indonesia is Shopee. The high online shopping activity in the current marketplace indirectly encourages business actors to understand the online market. However, one of the obstacles that are quite often faced by sellers, especially new sellers who are starting to enter the digital realm, is the emergence of confusion in the selection of products to be sold due to a lack of information regarding the demand for what products are in demand in the market.

The process of searching for information related to the demand for products of interest is carried out through clustering analysis to find out the groups of products that are of interest to those that are less attractive to the public. The data used is product data from 6 categories in the Shopee market which was taken using web scraping techniques. The clustering processes used the K-means approach by determining the number of K and the optimal center point through the calculation of Sum Square Error (SSE) by looking at the elbow graph. The final results show the optimal number of K clusters that are different in each category, namely in category women’s clothing, men’s clothing, and electronics are at K=4 then for products in the category of Muslim fashion, care & beauty and household appliances are at K=3. Based on the validation results using the Davies Bouldin Index, values were obtained in6 categories, namely 0.391, 0.438, 0.414, 0.357, 0.387, and 0.377, which means that the cluster structure and the level of information formed in each category using the K-Means method is quite good.