View article

Q-matrix as a tool for mapping items to skills is of great importance in educational assessment. However, a Q-matrix is merely a way to manually define the mapping between item and skills, since the genuine relationships between them can never be observed. Therefore, the refinement of an expert-given Q-matrix, the design of the most efficient Q-matrix, and the data-driven derivation of the Q-matrix are the research topics of this thesis. With the help of machine learning, now we have more tools to tackle these problems.