Unsupervised Deep Collaborative Filtering Recommender System for E-Learning Platforms

Abstract

With the rapid development of the online learning resources, trying to respect the differences between learners in terms of cognitive ability and knowledge structure. Traditional collaborative filtering recommendation algorithms cannot identify useful learning resources which will be interesting and simple to understand. Furthermore, the redundant recommended content and the high-dimensional and nonlinear data on online learning users cannot be effectively handled, leading to inefficient resource recommendations. To enhance learning resource recommendations efficiency, this paper presents a two steps efficient resource recommendation model based on, Kohonen card unsupervised deep learning to identify the instrumental approximation of learning styles, and deep auto-encoder, whose interest is not the prediction of resource in as such, but the transformation learned by the self-encoder, which serves as an alternative representation of the input and estimate the success rate of the proposed resource to the learner. This model need deeply mines learner features course content attribute features assessment attribute features and incorporates learner platform interactions features to build Learner features vector as input for the first step and Learner-Content ratings vector to choose the more efficient learning resource to recommend.