June 23, 2013
June 23, 2013
June 26, 2013
Educational Research and Methods
23.1311.1 - 23.1311.14
Knowledge Discovery and Pattern Finding in Students’ Solution SequencesIn this paper we apply machine learning techniques to automatically identifypatterns in the way in which students solve problems in an undergraduateMechanical Engineering course. Such patterns can provide valuable insight intostudents’ cognitive processes and, when correlated with performance in the class,can provide insight as to which behaviors may contribute to or impede success inthe classroom.We provided 150 students enrolled in a Mechanical Engineering statics course withLivescribe digital pens with which they completed all of their coursework. Thesepens serve the same purpose as traditional ink pens, but additionally digitize theink, producing a digital, time-stamped copy of the students’ coursework.This digital representation of student work provides an unprecedented view intothe sequence of steps students take to solve problems. For example, using thetiming information, we can answer the question, “How often do students completeall their free body diagrams before beginning to solve equations?” While manuallyinspecting this data set for interesting patterns of problem-solving behaviors isprohibitively time consuming, a digital corpus of student work allows us to usedata mining techniques to automatically identify such patterns.We encode a student’s solution to an assignment as a sequence of characters, usingan alphabet in which each letter represents a specific kind of action taken by thestudent. For example, a simple alphabet might contain two letters, “A”, indicatingthat a student drew a free body diagram, and “B”, indicating that a student wrote anequation. Using this alphabet, the sequence “ABA” would indicate that a particularstudent began by drawing a free body diagram, then wrote equations, and thenrevisited his/her free body diagram. We consider several different alphabetsdescribing a range of problem-solving activities. We then mine the resultingcharacter sequences using several popular data mining methods, e.g., FisherKernels and motif discovery. Through these methods, we identify commonpatterns of problem-solving behavior which correlate with both successful andunsuccessful course performance.
Herold, J., & Stahovich, T., & Rawson, K. (2013, June), Using Educational Data Mining to Identify Correlations Between Homework Effort and Performance Paper presented at 2013 ASEE Annual Conference & Exposition, Atlanta, Georgia. 10.18260/1-2--22697
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 2013 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015