Excavating Multimodal Correlation for Movie Recommendation System

Recommendation mechanisms, which aid users in navigating the deluge of information at their disposal, have become more popular as e-commerce applications have grown. In streaming services for movies, where tailored content and services are crucial, these technologies are particularly significant. But these algorithms frequently run into the issue of scant data, which makes it challenging to formulate precise suggestions. This research presents a customized deep learning-inspired multi-modal movie referral system to solve this problem. This system basically enhances the correlation between different pairs of attributes to get the highly correlated multimodal features. The correlation-based mechanism mines the hidden properties of different attributes (movie_id, movie_title, movie_description, and movie_poster) individually and the properties based on the dependency of one attribute on another. These hidden properties give a high contribution to movie score prediction. The study in this paper justifies that the multimodal features based on the correlation between different attributes enhance systems that provide suggestions. This strategy may result in many relevant suggestions and enhance user interfaces, especially for movie streaming services where customers seek much relevant products and services. The system analyzes the performance of the proposed recommendation approach using the IMDB Vision and NLP datasets and the results outperform the existing methods.