Peningkatan Visibilitas Produk pada Rekomendasi Long-Tail dengan Pendekatan Frequent Maximal Itemset

Abstract

Long-tail products are often overlooked in Collaborative Filtering recommendation systems due to their low purchase frequency and reliance on user interaction history. This study proposes the use of a Frequent Maximal Itemset (FMI) to improve the visibility of long-tail products in an online electronic cigarette (vape) store. Unlike Collaborative Filtering, FMI does not require user data and identifies historical transaction patterns to recommend relevant long-tail products alongside popular ones. Experimental results show that FMI is effective in identifying maximal itemsets that combine popular and long-tail products. Validation with 10 users revealed that 90% found the recommendations relevant to the main products they were searching for, and 90% indicated that they were likely to try the recommended long-tail products. The long-tail products included in the recommendations had logical associations with popular products, such as nicotine liquids with vaping devices. Thus, the FMI approach proves to be more flexible and effective in addressing popularity bias, while also providing long-tail products with greater visibility and increasing their potential for sales.