Improving Recommender Systems Performances Using User Dimension Expansion by Movies’ Genres and Voting-Based Ensemble Machine Learning Technique

Abstract

The recommender system's performance needs to be improved more than ever by increasing computer systems' usage in various applications. Recommender systems are a valuable tool in e-commerce websites. Their primary purpose is to generate accurate forecasts to access information in less time and energy for end-users. Classification optimizes information retrieval activity in these systems and reduces user search time. Besides, clustering tries to insert the new object in the best similar class, like using the k-nearest neighbor algorithm as a classifier. The proposing approach focuses on modeling categories by averaging rates of movie genres.Moreover, the user clustering will be improved by voting machine learning classifiers on multilayer perceptron (MLP) neural networks and k-nearest neighbors (kNN) algorithms. The experiments performed on the MovieLens dataset show that the proposed method is more successful than other previous methods in predicting user clusters with 93.81% accuracy, 94.45% precision, and 92.81% recall. Also, Davies Bouldin metrics indicates better clustering result using dimension expansion of movies' genres.