Cross domain and adversarial learning based deep learning approach for web recommendation

The web has become a massive source of knowledge in the internet age. This extra information makes it hard to choose items based on individual needs. Today, choosing suitable products takes time and effort. Daily uploads and downloads from YouTube, Instagram, Facebook, and others generate massive volumes of data. Keep up with the internet's wealth of information. Recommender systems can help users find useful data in vast datasets. User-interested recommender systems provide personalised and non-personalised recommendations. Real-time applications need recommender systems, but conventional methods have problems. In this work, we identified the issues and developed a cross-domain web recommendation system using a deep learning-based scheme. A joint reconstruction loss model reduces learning error with an autoencoder and adversarial learning technique. An open-source cross-domain dataset tests the proposed approach. For the Movie dataset, average HR, NDCG, and MRR are 0.8951, 0.5911, and 0.6121. The book dataset averages 0.8358, 0.6824, and 0.5575.