San Diego State University

1.  Introduction
2.  Related Work
3.  Student Performance 
4.  Marmoset 
5.  Modelling Techniques
6.  Features
7.  Experiments and Evaluation

Predicting student performance using data from an Auto-grading system 

Usually in programming courses students would like to have multiple attempts to submit their coding assignments and get the feedback quickly in order to improve their code.  While this is desirable for the students, it would be very difficult for teachers and their assistants to review and grade each assignment multiple times. Therefore automating system is needed and Marmoset is a tool that allows users to submit codes several times. In this paper the authors want to assess if the data from Marmoset can be used for final and midterm exam grade predictions.

Introduction (Predicting student performance using data from an Auto-grading system)

There were multiple attempts to predict student performance. Some of them focused on predicting GPA, pass/fail rates, exam grades. For predicting pass/fail rates in the final exam one of the previous works used data from PASTA  auto-grading  system. Their data contained also contained data outside of auto-grading system. In this paper the authors only use auto-grading data. 

Related Work (Predicting student performance using data from an Auto-grading system)

In the University of Vaterloo final and midterm exam contribute for 60%-90% of the final grade.  Marmoset is auto-grading system where students can submit coding assignments. Submitted code is tested on 3 different testcases: public, release and hidden. 

Student Performance and Marmoset (Predicting student performance using data from an 
Auto-grading system)

In their paper the authors used different models: decision tree, linear regression. Decision trees are built from the training set so that each split will give the maximum possible information. In the case of linear regression, they used multivariate linear regression as they had several independent variables. 

Modeling Techniques (Predicting student performance using data from an
Auto-grading system)

In their models they have used 4 distinctive features:

1.  **Passing rate for each task** - passing rate for the best submission. The total number of tasks is 28 and first 16 assignments are due before the midterm exam and the rest is due before the final exam.

2.  **Testcase Outcome** - testcase outcomes for the best submission.

3. **Submission Time Interval** - time interval between submission and the deadline.

4. **Number of Submissions**


Features (Predicting student performance using data from an Auto-grading system)

The authors constructed different models for the different tasks: classification - predict student category (GP - good performance, SP - satisfactory performance, PP - poor performance) and regression - for predicting exam grades.  The authors divided users into categories based on their performances  on the exams. The used the 4 features separately for training the model and the results were quite poor as most of students were classified as GP.  Classifying student as GP while he/she is PP is especially unsatisfactory.  In case of the regression model, they had much better performance with mean difference 0.92 when the maximum grade was 120.  Then in order to make regression models comparable with classification one, they used predicted grades to generate categories and results were better. They used only time interval for the regression model as a feature and found that there is strong correlation between exam performances and this feature. 

Learn Before

Related