Predictive modeling in a steelmaking process using optimized relevance vector regression and support vector regression

• Published in: Annals of operations research Vol. 316; no. 2; pp. 905 - 926
• Main Authors: Acosta, Simone Massulini, Amoroso, Anderson Levati, Sant’Anna, Ângelo Márcio Oliveira, Junior, Osiris Canciglieri
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-09-2022; Springer; Springer Nature B.V
• Subjects: Adaptive algorithms; Algorithms; Artificial neural networks; Business and Management; Combinatorics; Contaminants; Evolutionary algorithms; Evolutionary computation; Global optimization; Machine learning; Neural networks; Operations research; Operations Research/Decision Theory; Phosphorus; Prediction models; Quality control; Regression models; S.i. : Claio 2018; Steel making; Support vector machines; Theory of Computation; Brazil; Steelmaking process; Differential evolution; Support vector regression; Process modeling; Relevance vector regression
• ISSN: 0254-5330; 1572-9338
• DOI: 10.1007/s10479-021-04053-9

The existence of contaminants in metal alloys products is the main problem affecting the product quality, which is an important requirement for competitiveness in industries. This paper proposes the application of a relevance vector machine for regression (RVR) and a support vector machine for regression (SVR) optimized by a self-adaptive differential evolution algorithm for regression to model the phosphorus concentration levels in a steelmaking process based on actual data. In general, the appropriate choice of learning hyperparameters is a crucial step in obtaining a well-tuned RVM and SVM. To address this issue, we apply a self-adaptive differential evolution algorithm, which is an evolutionary algorithm for global optimization. We compare the performance of the RVR and SVR models with the ridge regression, multiple linear regression, model trees, artificial neural network, and random vector functional link neural network models. RVR and SVR models have smaller RMSE values and better performance than the other models. Our study indicates that the RVR and SVR models are adequate tools for predicting the phosphorus concentration levels in the steelmaking process.