Using educational data mining to identify correlations between homework effort and performance

Abstract

Homework has long been a cornerstone of education, but is it actually worthwhile for a student to put effort into homework? In this paper we present novel techniques for examining correlations between students' effort on homework and their performance in a course. Students enrolled in a Mechanical Engineering Statics course at the University of California, Riverside were given Livescribe™ digital pens with which they completed their coursework, producing an electronic, time-stamped record of all of their work. We computed numerical features from these records to estimate the effort students expended on each homework assignment. We used these features to predict student performance on a number of measures, such as homework, quiz, and exam scores, and show that these effort-based features can explain up to 39.9% of the variance in student performance (i.e., R2 = 0.399). These effort-performance correlations offer insight into the types of transfer that occurs from homework to exam problems. Additionally, these results serve as a measure of the effectiveness of homework problems, providing instructors with a principled method for improving homework assignments for future course offerings. ©American Society for Engineering education, 2013.