Artificial intelligence grace: On the empirical performance of ML approaches with applications in english education and the radiation-oriented sector

This study aims to explore the performance of Machine Learning (ML) algorithms on two different application areas: English education and radiation-oriented sectors, in the context of the Artificial Intelligence Grace. The selected operational environments are intentionally different case studies, aimed at assessing the scalability and generalizability of the proposed framework in fundamentally different operational environments. One domain is a learning and interaction domain, and the other is a safety-critical high-stakes technical decision-making domain. The study assesses the ability of these ML models to model the variety of these contexts by leveraging structured data sets that provide meaningful relationships between their features and balanced class distributions for reliable model development and assessment. The exploratory data analysis indicated that there were correlations among the features, with possible multicollinearity, which could affect the stability of the model and the estimation of the parameters. In order to overcome this problem, L2 regularization was added to Logistic Regression and Support Vector Machine (SVM) models, which led to greater stability and less variance. Three classification models, namely Logistic Regression, Support Vector Machine (SVM), and Random Forest, were tested and compared by means of accuracy and F1-score. The results show that the SVM model has better predictive performance in both domains compared to other models, particularly in the domain where there are high-dimensional and correlated feature spaces. Overall, regularization methods and balanced data representations were found to improve generalization and also reduce the classification bias in cross-domain ML applications.