Incremental session based collaborative filtering with forgetting mechanisms

Most of research works in music recommendation systems use Collaborative Filtering (CF) for generating personalized recommendations based on user´s previous song ratings or static usage history data. But those researches adapting CF do not consider behavior of listening to songs and are not able to maintain the systems to sensitive to recent user´s preferences. Behavior of music listening is continuous and repetitive process, especially, the latest song listening can infer to the favorite song at that moment. In this work, we present Incremental Session based Collaborative Filtering with forgetting mechanism or ISSCF by adapting Session-based Collaborative Filtering (SSCF), which considers music listened continuously and maintains the recent session. In order to avoid unnecessary memory usage and processing time, we use forgetting mechanism: sliding windows and fading factors incorporating with SSCF. We evaluate our purposed framework by measuring the HitRatio. From experimental results, it shows that performance of our purposed approach increases the accuracy of recommendation and low computational time and space when comparing with than SSCF.