A Novel Autorec-Based Architecture for Recommendation System

In recent years, collaborative filtering systems based on neural networks have strides in delivering users with personalized recommendations. CI-Autorec accepts a range of content-based representations as input to enhance system efficiency. Sparse Autoencoder has been proven to synthesize and reduce computational costs while still delivering high-quality results. This paper presents a framework that leverages the content-based information utilization capabilities of the CI-Autorec and employs active units to predict of the Spare Autoencoder. Experimental trials demonstrate the advanced performance of our recommendation system against that of other combined systems.