View article

The growth of self-learning, enabled by the availability on the Internet of different forms of didactic material such as MOOCs and tutoring systems, increases in turn the relevance of personalized instructions for students in adaptive learning environment. For providing adaptive and personalized learning instructions, the assessment of student’s mastery of a topic and the estimation of when she actually knows how to answer problems correctly is recognized as paramount in the fields of learning analytics and educational data mining community. In this dissertation, I propose novel approaches for building skills and student learning models along two axes. The first axis is to recover and ensure the quality of skills sets behind problems in learning system. The second axis is on improving the predictive accuracy of students’ performance based on student ability profile on skills and considering of difficulty of the problem …