Ahmad A. Kardan - faculty member of Department of computer Engineering and information technology, AmirKabir University of Technology, Tehran, Iran
Omid R. B Speily - Advanced E Learning Technologies Lab, AmirKabir University of Technology

Recommender systems play an important role in learning process by predicting user preferences. Learning process needs dynamic interactions between the learner and the learning system to recognize learner abilities, behaviors or other learner characteristics. Recommender systems have become increasingly popular in entertainment and e-commerce domains, but they have a little success in the elearning domains. Recommender systems learn about user preference over time, automatically finding things of similar interest. It reduces the burden of creating explicit queries during the learning process. Recommender systems use some techniques to recognize learners' preferences, such as filtering, machine learning techniques or hybrid techniques. In e-learning, some of these techniques can cause some problems or may be impossible to implement. This paper investigates a technique for recommender systems suitable for the learning environments to recognizing learners' preferences in the learning process. This technique predicts user preferences in order to identify a useful set of items and to be recommended in response to the learners specific information need. We propose a hybrid technique based on machine learning to recognize learner preferences and predict theirs required contents with high accuracy.

machine learning, recommender systems, reinforcement learning, learner model, Q learning