Data Mining to Predict Student Success in a Large Class (578) Home : Annual Conference 2014

Data Mining to Predict Student Success in a Large Class (578)

Public Group Active 7 years, 3 months ago

Mon, Sep 1 2014, 3:10pm – 4:10pm

Description

New classroom technologies enable the collection of unprecedented data on student participation. This paper builds on previous data mining research (Romero & Ventura, 2007; Siemens & Baker, 2012) to consider to what degree the collection of intrinsic and extrinsic data can be related to student outcomes. This interactive session invites attendees to participate in a mock class session with tools that collect student notes, questions and answers in class to see what can be produced in real-time. The presentation will include examples of lessons learned from previous classes on how student note-taking, question responses and even emotional state can be used as predictive measures of student outcomes.

Data collected over multiple courses and disciplines illustrates trends between student note taking and student outcomes with the greatest number of words per students associated with those getting both the best grades and the worst. Also, student outcomes appear to be well correlated with how well students answer questions during class time. Finally, student outcomes seem to be associated with students’ emotional and physical state. Surveys done daily were clustered and it was found that the best grades were associated with those consistently reporting the best emotional state while those reporting low and variable emotional state fared poorest.  The results presented are for one introductory science course but the methods are extensible to other courses using in-class technologies.

[http://www.sageonstage.com/2014/04/predicting-student-success-in-a-large-class/]