We present a pioneering comparison between an expert-driven clustering technique called Facet Theory with the data-driven q-matrix technique for educational data mining. Both facets and q-matrices were created in order to assist instructors with diagnosing and correcting student errors, and each have been used to augment computer-assisted instructional systems with diagnostic information. However, facets are very specific aspects of knowledge, and the decomposition of a topic into facets can be overwhelming to teachers who need this diagnostic help. We present a set of four experiments, demonstrating that the q-matrix educational data mining technique reflects expert-identified conceptual ideas, but does so at a higher level than facets, indicating that a combination of expert-derived and data-derived conceptualizations of student knowledge may be most beneficial.
展开▼
