Content-based Recommender System Improvement using Hybrid Technique

Abstract

The research in recommender systems has evolved considerably over the past years; however, to date the investigation on how the emotive state of the user could be used to complement such technologies is sparse. Many systems used the emotions of the user as implicit feedbacks in the recommender systems, but limited works analyzed the emotions and using them for item profile modelling. This system extracts the affective features of the user’s face using the convolution neural network with three-dimension to build the item profile. Also, the support vector machine classifier is used for user profile building and introduces recommendations. The experiments conducted on an LDOS-PerAff-l dataset indicates that the hybrid technique contributes to improving the content-based recommender system performance using affective-automated features for item modelling instead of handcrafted features that extract using traditional techniques.