USeeTV TVOD Platform Recommendation System Optimization Using Singular Value Decomposition

The most important thing here is finding out why the learner is quiet. If it's simply because they're a quiet person who only speaks their mind when needed, then we can provide activities that require everyone to say something. However, if the person is quiet because they don't feel comfortable expressing their mind, we need to assure them that our learning session is a safe environment without any judgment. UseeTV's TVOD platform is developing very rapidly. Due to these developments, the TVOD UseeTV platform has 172 channels with 41,131 content. Due to the large amount of content available, it is necessary to implement a recommendation system on the platform. This study used the content-based filtering method with TF-IDF and cosine similarity to create a recommendation system. However, TF-IDF still has problems, where the amount of feature space generated is still vast and sparse. To overcome this, we optimized TF-IDF usingSVD. Based on the study results, it has been proven that using SVD can improve the recommendation system's performance by increasing computing speed, which is two times faster. In addition, the average similarity score of the recommendation results is up to 4 times better when compared to the Recommendation System, which SVD does not optimize.