Methods and reliability study of moral education assessment in universities: A machine learning-based approach

Abstract

The research aims to assess the effectiveness of machine learning (ML) techniques in evaluating moral education programs at university institutions. The objective is to employ data-driven methodologies to enhance ethical assessment frameworks through improved objectivity, scalability, and consistency. This analysis utilizes Principal Component Analysis (PCA) alongside the k-Nearest Neighbor (k-NN) method, Support Vector Regression (SVR), and Artificial Neural Networks (ANN) to study student performance indices, enabling the prediction of ethical reasoning capabilities for standardized evaluation. The study demonstrates how machine learning efficiently assesses student moral education performance by leveraging PCA to identify patterns and using ML models to make accurate predictions. Findings reveal a strong correlation between subject proficiency in mathematics, reading, and writing and moral reasoning abilities, highlighting the role of academic competencies in ethical decision-making. Additionally, gender-based analysis indicates that female students tend to achieve better results in moral skills assessments than their male counterparts. Among the models tested, SVR exhibits the highest predictive accuracy, whereas k-NN returns the widest prediction errors. The study recommends the deployment of AI-based moral assessment systems in universities to ensure consistent and objective evaluation processes for policy development.