Educational data mining: Collection and analysis of score matrices for outcomes-based assessment.

摘要

In this dissertation we provide an overview of the nascent state of Educational Data Mining (EDM). EDM is poised to leverage an enormous amount of research from the data mining community and apply that research to educational problems in learning, cognition, and assessment. Similar problems have been researched in the educational community for over a century, but the enormous computing power and algorithmic maturity brought to bear by data mining has proven to be more successful at many of these educational statistics problems. The timing of these developments could not be better, given the current rising importance of assessment throughout education, particularly in the United States. After determining the structural commonalities between EDM projects, I detail my own EDM assessment project, covering tools for collecting, strategies for storing and archiving, and new techniques for analyzing matrices of student scores. I also detail issues of real-world deployment and adoption of this assessment system. After examining the state of EDM, both in the abstract and with my own implementation and deployment, I predict near-term trends in EDM and assessment, and conclude with thoughts on the implications of this work, both for pedagogy and for the data mining community as a whole.