Enhancing Personalization in Movie Recommendation Using Android Application: A Machine Learning Approach

With the rapid and tremendous growth of digital contents, personalized movie recommendations have become essential for enhancing the experience as well as interaction of users. However, conventional movie recommendation systems appear to face challenges such as low scalability, cold start issues, and limited dataset attributes that hinder their effectiveness. This paper reports a study that attempted to improve movie recommendations by using an Android application that adopted a machine learning algorithm. The Android application developed in this study employed the hybrid-based filtering algorithm that consisted of the following two algorithms: (1) a collaborative filtering algorithm with singular value decomposition (SVD); and (2) a content-based algorithm with cosine similarity. The MovieLens dataset was utilized for training and evaluation of the algorithm. The development of the Android application involved the Android Studio and the Firebase database for real-time database management. The experimental results of the study showed that the hybrid-based filtering algorithm was able to achieve recall@5 of 69.83%, recall@10 of 79.00%, mean absolute error (MAE) of 2.77, and root mean squared error (RMSE) of 2.96. The effectiveness of the Android application was also assessed by using the feedback collected from the users through surveys. Results of the surveys showed that most of the users were satisfied with the Android application developed in the study, particularly in terms of its design, functionality, and usability. The study further suggested that future research should specifically emphasize on enhancing the backend integration and graphical user interface for the purpose of broader adoption.