A Recommender System for Professors and Course Coordinators Using Failure Prediction in Distance Learning

This paper proposes an educational recommendation system model based on prediction of students flunking in e-learning courses. RECD is proposed, a model of Educational Recommendation Systems which receives from some prediction system, among them MD-PREAD, statistical data on the possibility of learners' failure in a discipline and, based on these data, recommends to the target audience to reduce the number of apprentices who can fail. Techniques such as classification of user profiles, context awareness and Custom-er Relationship Manager were used to provide learners an opportunity not to fail in a discipline, or to improve the coeffi-cient of performance, to reduce the time enrolled in the course and to accelerate the certification Through pedagogical inter-ventions A prototype was designed to be experimented at the Federal Institute of Education, Science and Technology of Amazonas, in the program Open University of Brazil, in the Philosophy of Education course, in discipline of Brazilian Sign Language, in the second semester of 2015. We collected 30 teachers profiles, allowing the classification of the teacher profile using decision tree with RapidMiner. The prototype was also presented to 12 teachers so that they could make an evaluation of perceived ease of use and utility perception through the Technology Acceptance Model. It was concluded that RECD is a computational tool that can help teachers and course coordinators to rescue apprentices before the failure to culminate, in the discipline in progress.