Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic Piecewise Linear Regression for Predicting Student Learning Satisfaction

Haemin Choi, Gayathri Nadarajan

TL;DR

The paper addresses predicting student learning satisfaction during the pandemic using an interpretable boosting framework. It introduces Automatic Piecewise Linear Regression (APLR), which combines gradient boosting with piecewise linear bases to capture nonlinearities and interactions while preserving interpretability. Across a survey of 302 university students, APLR outperforms bagging, boosting, and a transformer-based model on four of five metrics, and provides global and local explanations linking time management, concentration, helpfulness to classmates, and offline course participation to higher satisfaction, while creative activities show no positive effect. The findings support personalized instructional design by offering interpretable, instance-level insights and suggest APLR as a practical tool for education data mining on structured, small-scale datasets.

Abstract

Although student learning satisfaction has been widely studied, modern techniques such as interpretable machine learning and neural networks have not been sufficiently explored. This study demonstrates that a recent model that combines boosting with interpretability, automatic piecewise linear regression(APLR), offers the best fit for predicting learning satisfaction among several state-of-the-art approaches. Through the analysis of APLR's numerical and visual interpretations, students' time management and concentration abilities, perceived helpfulness to classmates, and participation in offline courses have the most significant positive impact on learning satisfaction. Surprisingly, involvement in creative activities did not positively affect learning satisfaction. Moreover, the contributing factors can be interpreted on an individual level, allowing educators to customize instructions according to student profiles.

Automatic Piecewise Linear Regression for Predicting Student Learning Satisfaction

TL;DR

The paper addresses predicting student learning satisfaction during the pandemic using an interpretable boosting framework. It introduces Automatic Piecewise Linear Regression (APLR), which combines gradient boosting with piecewise linear bases to capture nonlinearities and interactions while preserving interpretability. Across a survey of 302 university students, APLR outperforms bagging, boosting, and a transformer-based model on four of five metrics, and provides global and local explanations linking time management, concentration, helpfulness to classmates, and offline course participation to higher satisfaction, while creative activities show no positive effect. The findings support personalized instructional design by offering interpretable, instance-level insights and suggest APLR as a practical tool for education data mining on structured, small-scale datasets.

Abstract

Although student learning satisfaction has been widely studied, modern techniques such as interpretable machine learning and neural networks have not been sufficiently explored. This study demonstrates that a recent model that combines boosting with interpretability, automatic piecewise linear regression(APLR), offers the best fit for predicting learning satisfaction among several state-of-the-art approaches. Through the analysis of APLR's numerical and visual interpretations, students' time management and concentration abilities, perceived helpfulness to classmates, and participation in offline courses have the most significant positive impact on learning satisfaction. Surprisingly, involvement in creative activities did not positively affect learning satisfaction. Moreover, the contributing factors can be interpreted on an individual level, allowing educators to customize instructions according to student profiles.

Paper Structure

This paper contains 20 sections, 1 equation, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Studies on Student Learning Satisfaction
  4. Interpretable ML and Neural Networks
  5. Methods
  6. Survey Dataset
  7. Task Design
  8. Data Distribution
  9. Automatic Piecewise Linear Regression (APLR)
  10. Hyperparameter Tuning
  11. Random Forest.
  12. LightGBM.
  13. APLR.
  14. Results
  15. Performance Comparison
  16. ...and 5 more sections

Figures (3)

  • Figure 1: The top fifteen contributing predictors and their global feature importance for classifying learning satisfaction using APLR.
  • Figure 2: APLR's local explanation when making a prediction for a sample from the positive class. All bars on the right indicate positive influence, with longer bars showing stronger influence. The bar on the left indicates negative influence.
  • Figure 3: APLR's local explanation when making a prediction for a sample from the negative class. All bars on the right indicate positive influence towards predicting the negative class, with longer bars showing stronger influence.