Analyzing Teaching Performance of Instructors Using Data Mining Techniques

Abstract

To identify trends into instructor effectiveness, this paper utilizes data mining techniques on student evaluations of teaching quality. PCA and clustering were employed to identify distinct response patterns and underlying constructs in a set of student-faculty ratings on multiple performance dimensions. Three categories of students with different patterns in evaluation emerged using K-means clustering presenting consistent and inconsistent evaluations of teachers. PCA further reduced dimensionality by extracting a single criterion and examining the main variation in the data in a single predominant component that reflected standard evaluation biases. The results provide an evidence-based perspective on teaching quality, and identify specific strengths, and opportunities for improvement. This provides the basis for faculty development and evidence-based decision making in higher education.