Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Student Performance Prediction In CS1 Via In-Class Coding

Eric Hics, Vinhthuy Phan, Kriangsiri Malasri

TL;DR

The paper addresses high failure rates in CS1 by evaluating whether in-class coding exercises improve early prediction of student performance and by assessing the impact of curved grading on predictive accuracy. Using data from ~300 CS1 students across seven sections in 2019 and multiple classifiers, it shows that Random Forest models with hybrid feature sets (in-class plus traditional) provide the best predictive performance, with usable signals emerging by weeks 3–5. The findings demonstrate that in-class coding data can enhance timely interventions and that standardized grading scales yield more reliable predictions than instructor-curved grades. This work contributes to CS education by highlighting the value of in-class signals for early identification of at-risk students and guiding intervention strategies throughout the semester.

Abstract

Computer science's increased recognition as a prominent field of study has attracted students with diverse academic backgrounds. This has significantly increased the already high failure rates in introductory courses. To address this challenge, it is essential to identify struggling students early on. Incorporating in-class coding exercises in these courses not only offers additional practice opportunities to students but may also reveal their abilities and help teachers identify those in need of assistance. In this work, we seek to determine the extent to which the practice of using in-class coding exercises enhances the ability to predict student performance, especially early in the semester. Based on data obtained in a CS1 course taught at a mid-size American university, we found that in-class exercises could improve the prediction of students' eventual performance. In particular, we found relatively accurately predictions as early as academic weeks 3 through 5, making it possible to devise early intervention strategies. This work can benefit future studies on the impact of in-class exercises as well as intervention strategies throughout the semester.

Enhancing Student Performance Prediction In CS1 Via In-Class Coding

TL;DR

The paper addresses high failure rates in CS1 by evaluating whether in-class coding exercises improve early prediction of student performance and by assessing the impact of curved grading on predictive accuracy. Using data from ~300 CS1 students across seven sections in 2019 and multiple classifiers, it shows that Random Forest models with hybrid feature sets (in-class plus traditional) provide the best predictive performance, with usable signals emerging by weeks 3–5. The findings demonstrate that in-class coding data can enhance timely interventions and that standardized grading scales yield more reliable predictions than instructor-curved grades. This work contributes to CS education by highlighting the value of in-class signals for early identification of at-risk students and guiding intervention strategies throughout the semester.

Abstract

Computer science's increased recognition as a prominent field of study has attracted students with diverse academic backgrounds. This has significantly increased the already high failure rates in introductory courses. To address this challenge, it is essential to identify struggling students early on. Incorporating in-class coding exercises in these courses not only offers additional practice opportunities to students but may also reveal their abilities and help teachers identify those in need of assistance. In this work, we seek to determine the extent to which the practice of using in-class coding exercises enhances the ability to predict student performance, especially early in the semester. Based on data obtained in a CS1 course taught at a mid-size American university, we found that in-class exercises could improve the prediction of students' eventual performance. In particular, we found relatively accurately predictions as early as academic weeks 3 through 5, making it possible to devise early intervention strategies. This work can benefit future studies on the impact of in-class exercises as well as intervention strategies throughout the semester.
Paper Structure (33 sections, 4 figures)

This paper contains 33 sections, 4 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Modeling Data And Target Variables
  4. Features And Feature Engineering
  5. Curved Grading
  6. Method
  7. Participants
  8. Classroom Setting
  9. Lecture
  10. In-Class Coding Activities
  11. Lab
  12. Data Processing and Collection
  13. Modeling Student Performance in CS1
  14. Classification methods
  15. Features
  16. ...and 18 more sections

Figures (4)

  • Figure 1: Accuracy in predicting Final Course Grade, Final Exam Grade, and Lab Final Grade for various models including baselines.
  • Figure 2: Accuracy of Random Forest Classification in predicting Final Course Grade, Final Exam Grade, and Lab Final Grade, from the beginning of the semester to the end of each academic week. In "Hybrid", both in-class exercises and out-of-class assignments and exams were used to build models. In "Traditional", only out-of-class assignments and exams were used.
  • Figure 3: Accuracy of Random Forest in predicting an exact Final Course Grade versus a General Performance throughout the semester.
  • Figure 4: Standardized versus Curved Grading: Differences in Random Forest's prediction performance of final course grade using two different ways of assigning grades.