Tuesday, March 20, 2012

Using Decision Trees for Predicting Student Academic Performance

Using Decision Trees for Predicting Student Academic Performance

In this study, authors Anupama Kumar and Dr. Vijayalakshmi use decision trees to predict student academic performance. Data will be used to help tutors make better decisions regarding student success and academic needs.

The authors look to find a way use a combined method of decision trees and rule mining to predict student performance in upcoming tests, quizzes, etc for the use of tutors to create a better way to teach these students. In this study, the authors use a decision tree algorithm called C4.5 to predict whether students will pass or fail said exam. The outcome of the decision tree analysis was then used in a comparative analysis which stated the prediction helped students who were struggling, and challenged excelling students to reach higher levels of success.

Decision trees are a successful way to predict student outcome and improve their results based on prior educational data. Decision tree algorithms can be better rated for efficiency based on their accuracy and time taken to derive the tree.


Kumar, S. Anupama, Vijayalakshmi, Dr., (2009). Efficiency of decision trees in predicting student’s academic performance. Retrieved from http://airccj.org/CSCP/vol1/csit1230.pdf


  1. I also read about the C4.5 but did not find any additional research on its application in the course of my research. Even though, it seems like decision trees' greatest utility come in the form of computer programming. The possible combinations of leafs and nodes can be mind-numbing for some applications, as I'm sure you've found.

  2. It'd be interesting to see if this research could be taken a step further to see if customized curriculum could be developed for each student to fit individual style to obtain a higher degree of academic success.

  3. Did the decision tree method result in higher test scores? I would imagine if a method like this could be employed in classrooms, teachers would have an easier time deciding how best to utilize class time and resources.

  4. Would you be interested in implementing this for graduate intelligence students. I wouldn't mind having my grades improved using a decision tree algorithm.

    Is it applicable for all students? Or is it only useful for those using the wrong learning techniques?

  5. @ Dave - I'm not sure that it would be as applicable in public schools, as there is already a lot of individualized intruction happening, but perhaps if the data were gathered from teachers and given to a third party?
    @ Leslie - Yes, there were higher test scores for all students, and those who did not experience academic difficulty to begin with still ended up excelling, if only slightly.
    @ Dean - According to this study, applicable to all students.

  6. I'd be interested to know more about C4.5. Did you get any sense of how useful this algorithm would be?