EVALUATING THE EFFECTIVENESS OF NEURAL COLLABORATIVE FILTERING, MULTILAYER PERCEPTRON AND MATRIX FACTORIZATION TECHNIQUES FOR RECOMMENDER SYSTEMS

Abstract

In the age of digital connectivity, where a vast array of content is easily accessible, it has become increasingly difficult to find captivating and personally appealing material. Moreover, as the internet continues to expand at a rapid pace, more and more people are turning to OTT platforms and digital content providers, like e-books, to enhance their leisure activities. In response to this challenge, recommendation systems have emerged as indispensable tools that offer users customized content suggestions based on their individual preferences and interests. In this project, we aim to assess the effectiveness of three distinct methods: Neural Collaborative Filtering, Multilayer Perceptron, and Matrix Factorization. Our objective is to evaluate the performance and efficacy of these approaches in the context of our study. Neural Collaborative Filtering (NCF) is a powerful approach that combines the capabilities of neural networks with collaborative filtering techniques, enabling the generation of highly personalized recommendations. By leveraging the strengths of both neural networks and collaborative filtering, NCF excels at capturing complex patterns and relationships in user-item interactions, resulting in accurate and tailored recommendations for individual users. Matrix Factorization is a technique commonly employed in collaborative filtering-based recommendation systems. Its primary objective is to decompose a user-item interaction matrix into lower-dimensional representations. This approach assumes that the observed interactions between users and items can be effectively explained by a set of latent factors or features. By discovering these underlying factors, matrix factorization enables the system to make accurate predictions and provide recommendations based on users’ preferences and historical behaviour. The Multilayer Perceptron (MLP) is an Artificial Neural Network (ANN) architecture with interconnected layers of artificial neurons. It processes information in a one-way flow from input to output, without loops. Neurons apply nonlinear functions and pass outputs to the next layer. MLPs effectively capture complex data relationships for tasks like classification, regression, and pattern recognition.