Recommender Systems using Hybrid Demographic and Content-Based Filtering methods for UMKM Products

Abstract

Marketing digitization such as e-commerce is needed by micro, small and medium enterprises (UMKM) in Bogor City and Regency so that the products are more easily accessible to consumers. One of the digital marketing that is commonly used by consumers is an e-commerce website. The Recommendation System is implemented into e-commerce websites to increase consumer convenience in online shopping. The recommendation systems method applied is Demographic Filtering and Content-based Filtering. Demographic Filtering uses IMDB Weighted Rating calculations which generate recommendations globally and give recommendations based on each product's IMDB Weighted score. Content-based Filtering uses Cosine Distance calculations which generate personal recommendations and give recommendations based on the score cosine distance of each product in the form of a presentation of the similarity of products that have been purchased with other products. This research uses 107 UMKM fashion and craft product data that was obtained from Bogor City Regional Craft Center which sells various kinds of UMKM products from Bogor City and Regency. Data preprocessing is then carried out on the raw data, with the Data Cleaning, Data Transforming and Data Splitting stages which divide the data in a ratio of 80:20. The accuracy of Demographic Filtering Recommendation System reaches 82.7% and Content-based Filtering Recommendation System reaches 100%.