Movie Description Feature Extraction for Movie Recommendation Model with Content Based Filtering Approach

Abstract

Recommendation is a method that simplifies the user's decision-making process when faced with multiple options. Recommendation might also be put into practice for the choice of movies. Prior academics have extensively proposed recommendation models employing various data and methodologies. The presence of data in the process of modeling recommendations plays a crucial role in determining the choice of approaches. The data employed in this study pertains to specific movie items that will be recommended. Therefore, the chosen method for analysis is content-based filtering. Due to the presence of descriptions in the available features, the text mining strategy is employed using vectorization methods such as TF-IDF and CountVectorizer for analysis. This approach is capable of generating effective features for the recommendation model, in addition to other movie-related features. The experimental results demonstrated that TF-IDF achieved precision@k values that were 0.6% superior to those of CountVectorizer for film type of show. Conversely, for television shows, CountVectorizer yielded 3.24% more accurate outcomes in providing relevant recommendations for the selected base movie.