Deep learning based semantic personalized recommendation system

Abstract

The past decade has seen significant development in the number of personalized recommendation applications on the World Wide Web. It aims to assist users to retrieve relevant items from a large repository of contents by providing items or services of likely interest based on examined evidence of the users’ preferences and desires. However, this vision is complex due to the huge amount of information aka media-rich information available on the web. Most of the systems formulated so far use the metadata linked with the digital contents, but such systems fail to generate significant recommendations results. In these circumstances, a semantic personalized recommendation system (SPRS) plays an important role to take away the semantic gap between high-level semantic contents and low-level media features. The proposed system recommends personalized sets of videos to users depending on their previous activity on the site and exploits a domain ontology and user items content to the domain concepts. To evaluate the performance of the framework, items’ prediction is executed by utilizing the proposed framework, and performance is determined by comparing the predicted and actual ratings of the items in terms of Predictive Accuracy Metrics, precision, and recall.