OPTIMIZATION OF MARKET BASKET ANALYSIS USING CENTROID-BASED CLUSTERING ALGORITHM AND FP-GROWTH ALGORITHM

Abstract

The proliferation of the food and beverage sales business requires the creativity of business owners to offer their flagship products to every consumer, both new and subscribed consumers. A large number of menu choices makes the ordering process long because consumers are confused about which menu will be the best choice. the seller to be able to provide the right recommendations so that orders can take place faster. Shopping cart analysis is an activity that has often been done to find out the items found that are sold simultaneously. The FP-Growth association method is a faster algorithm for generating association rules, but the association process in large dataset sizes tends to add large items so that the accuracy value of association rules decreases. So that in this study, the grouping of datasets was carried out using a clustering model with a centroid-based algorithm, namely k-means, k-medoids, and fuzzy c-means. This research was conducted through dataset collection, dataset preparation, clustering modeling, evaluation of clustering models using DBI and silhouette index, association modeling, and evaluation of association models using lift ratio. The results of this study showed that the clustering model with the best DBI and silhouette index values ​​was at k=3 for k-means, k=2 for k-medoids, and k=7 for fuzzy c-means. The number of association rules is generated from the grouped data set using fuzzy c-means, but the highest average lift ratio is in the association rules generated from the grouping data set using k-means. From the association model using k-means and FP-Growth, 32 unique association rules were found with the 4 most frequently found items, namely cireng chili oil, regal milk coffee, banana cheese, and vietnam drip.