Article Recommendations with Item-Based Collaborative Filtering on Online News Portals

Abstract

News portals generate additional traffic or traffic visits from the article recommendation widget. However, it is unfortunate that the traffic visits obtained from the widget are still relatively small. The article recommendation widget is rarely clicked by readers because the available recommended articles are less relevant to readers, resulting in one reader only reading no more than 2 articles obtained from the article recommendation widget. The purpose of this study is to further optimize the currently available article recommendation widget feature by adding reader interest data so that the number of articles read by one user will increase and will directly have an impact on increasing traffic visits. The method used in this study is Item Based Collaborative Filtering. After using the item-based filtering method by calculating the set of items x read and the duration of the reader's time in reading item x. In this study, a simulation was given to one of the reader samples and it was found that the highest interest of the reader sample was in reading sports news with a calculation score is 0.743210. The results of this study are article recommendations that match the reader's interests. The results of the study are expected to help users find articles that match their interests and preferences, so that they can increase the level of interaction and engagement with online media.