Making Recommender Module in OpenCart Using Item-Based Collaborative Filtering Method

Abstract

This research aims to add a recommender module in OpenCart CMS. One of the method is item-based collaborative filtering that can reduce the execution time of calculation. This study uses adjusted cosine similarity to calculate the similarity between books, weighted sum method to calculate the books rate prediction, and mean absolute error to calculate recommendacy accuracy. In order to get the recommendation, user have to login and give ratings to the books. Then, adjusted cosine similarity calculates the similarity between books based on user’s rate. Based on the similarities between books, weighted sum method calculates the books rate prediction. Before a book is recommended to the user, type of book from the prediction are first matched with the type of book rated by the user. This research uses 300 books and 30 users. The result shows that only 17 users can get recommendations. Evaluation is conducted by analyzing the execution time and recommendacy accuracy. It is found that the execution time is 1.60 seconds and the mean absolute error is 0.15.