Komparasi Algoritma Hierarchical, K-Means, dan DBSCAN pada Analisis Data Penjualan Melalui Facebook

- The use of the internet in Indonesia to access social media has increased from the previous four years, where 36.36% of users still use social media Facebook. The average social media users are teenagers with smartphones. Facebook has features that are favored by its users for buying and selling activities, so that it can increase user engagement and sales data. To analyze the increase in sales data, this study uses data mining with clustering methods. By using secondary data from the UCI Repository, a comparative analysis of three different algorithms was carried out to find out which is the best among the Hierarchical, K-Means, and DBSCAN algorithms. The results showed that the Hierarchical algorithm obtained the highest silhouette score, namely 0.884, a fairly thin difference with the silhouette score obtained by K-Means, which was 0.872. Furthermore, the results of comparisons made using performance indicators show that K-Means is the best algorithm with an average execution time of 0.402 seconds, a considerable difference from the other two algorithms. Based on the two indicators that have been used, it can be seen that the best algorithm for analyzing sales data via Facebook is the K-Means algorithm. Finally, the appearance of the number of clusters 2 from the K-Means algorithm can group sales data via Facebook into two categories, namely "Busy Posts" and "Lone Posts".