PRODUCT RECOMMENDATION SYSTEM IN MARKETPLACE USING MULTI-FILTERING MODEL

• Main Author: Firdaus Arifi, Yusuf
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/70886
• Institution: Institut Teknologi Bandung

Providing appropriate product recommendations to increase user consumption is the goal of many companies today. People usually buy new products based on most purchased product, similar products, or feedback from other users. To do all this automatically, a recommendation system must be implemented. This final project aims to build a recommendation system specifically for both new users and old users right when after made purchase. The data is taken from a set of transaction data from US Superstore Marketplace in the fourth quarter of 2018 with selected column, which consists of 1033 rows of purchase transaction data by 418 users with 729 products. This recommendation system is built using multi-filtering model that combines several filtering techniques, called as global filtering, product-based filtering, and association-rule filtering. Global filtering uses sales information to provide recommendations for best-selling products. Product-based filtering techniques aim to provide recommendations for products that are most similar to products that have been purchased using category, subcategory, and product name information. Association-rule filtering recommends products that are frequently purchased together with other products using transaction history data information. Multi-filtering model makes it easier for users to select more product options at the same time. In its development, recommendation systems play a role in improving the process and quality of decision making. Thus, it has an impact on increasing the sales and revenue of the company.