Hybrid Recommendation System with Term Frequency-Inverse Document Frequency to Recommend Movies

Nowadays, the movie recommender systems have become increasingly important by helping users to navigate the vast number of movies available online. The Collaborative Filtering (CF) algorithms are the basis to provide relevant movie recommendations by capturing the user-item interactions. However, the existing models with CF algorithms facing challenges with the cold start problem where new users or new items make it difficult to generate meaningful recommendations. Hence, this research proposes a Hybrid Recommendation System (HRS) with Term Frequency-Inverse Document Frequency (TF-IDF) named as (HRS-TI) to improve recommendation accuracy and user satisfaction. The proposed HRS-TI addresses the cold start problem by calculating the similarity and allows the system to recommend new items without interaction history. Initially, the input data is collected from MovieLens 20M dataset and preprocessed with three techniques namely z-score standardization, one-hot encoding and data binarization respectively. Then, the TF-IDF is employed to calculate the similarity between users and items according to their history. Finally, the HRS which combines the CF with Content-Based Filtering (CBF) to recommend the movies based on the similarity and user interests. From the results, the proposed HRS-TI outperformed the existing Light Graph Convolutional Network (LightGCN) model in terms of precision (0.925) and recall (0.993) respectively.