User Collaborative Filtering Implementation for Purchase Recommendations using the Cosine Similarity Algorithm

Dewi Khairani,Luh Kesuma Wadhani,Yusuf Durachman,Zainal Muttaqin

The trend of internet usage has increased every year. Data noted that internet users in Indonesia increased by more than 50% in 2018 compared to the previous 2 years. This causes a lot of information to spread, including on e-commerce, which causes users to take longer to find the desired item. The case study conducted by the researcher found that the XYZ e-commerce website still displays product recommendations manually, so a recommendation system is needed so that it can display the product recommendations according to the user's interests. The purpose of this study was to determine the accuracy of the recommendations of the goods produced against users by implementing user collaborative filtering (UCF) using the cosine similarity algorithm. Analysis based on the results of testing conducted by researchers based on 3 scenarios, 3, 5 and all neighbours, UCF with the cosine similarity algorithm gives the best results when testing based on 3 neighbour scenarios which have a precision value of 0.758 which means accurate, with an average error level 0.9147. It can be concluded, the lower the number of neighbours chosen, the precision test results will be more effective and the value of the MAE error level will be lower.

Volume 12 | Issue 6

Pages: 2643-2650

DOI: 10.5373/JARDCS/V12I6/S20201223