Development of a hybrid information recommendation system considering serendipity
Haruto Domoto, Takahiro Uchiya, Ichi Takumi · 2023
In conventional recommendation systems, once an item is recommended, only similar items are recommended. The user becomes bored with the recommendation results. To resolve this shortcoming, we specifically examine "serendipity", an indicator of unexpectedness. This study was conducted to improve serendipity using the popularity rankings of items to find rare items and to assign priority to recommendation of those items using User-based Collaborative Filtering. Furthermore, to make recommendations suitable for user preferences, Item-based Collaborative Filtering or SlopeOne were introduced to create hybrid recommendations. The results underscore the effectiveness of our proposed system.