A trustworthy model of recommender system using hyper-tuned restricted boltzmann machine

The rapid and ubiquitous digital revolution has led to acceleration towards a digitally connected world where accepting recommendations digitally has become a part of our e-commerce related lifestyle. Most of these recommendations are based on rating values, wherein some values are fraudulent. This is a vulnerability in the security of the system and thus hampers its credibility and trust. It is necessary to identify the fraudulent ratings in real-time and build a trustworthy black-box recommender system. This paper proposes hyper-tuned Restricted Boltzmann Machines to regenerate tabular data models using contrastive divergence learning procedures to enhance accuracy, transparency, and recommendations’ ability. Most of the existing recommender systems cannot handle the large dataset and deteriorate their performance. Experiments over the Movies Lens,Film Trust and Netflix rating dataset demonstrate that the tuned RBMs performed well compared with SVD, SVD ++, trust SVD, and Stack Auto Encoder(SAE) models with a moderate increase in computational complexity. We build a trustworthy recommender system based on hyper-tuned RBM, a deep learning-based system widely used by E-commerce companies to recommend products. Netflix recommends television shows and movies depending upon the number of customers who have seen a similar genre. Similarly, Amazon recommends an item to a customer based on whether other consumers would be interested in purchasing the product. Through rigorous evaluations, we have shown that our proposed scheme outperforms the existing recommender systems.