A fuzzy entropy technique for dimensionality reduction in recommender systems using deep learning

Abstract

Recommenders utilize the knowledge discovery-based methods for identifying information required by the user. The recommender system faces some serious challenges in recent years to access exponentially increasing information due to high number of Web site users. Some of the challenges posed in this respect are: The system should assure high-quality recommendations and high coverage even during data sparsity and produce more recommendations per second based on million users. To improve the performance of the recommender system, selecting appropriate features from the available highly redundant information is a crucial task. The feature selection technique will bring down the dimensionality and also discard the redundant and the noise-corrupted features. The collaborative filtering-based methods will make use of the past activities or the preferences like the user ratings or content information of the products to regulate the top references. This work proposes a fuzzy entropy-based deep learning for the content features as well as a feature selection method. Deep learning-based recommender process takes extended important consideration by overwhelming difficulties of conventional models and attaining high reference excellence. A fuzzy entropy-based feature selection technique lowers the dimensionality of hyperspectral data.