Enhancing predictive accuracy for Cr(VI) removal in polymer inclusion membranes: A comparative study of machine learning models

• Published in: Inorganica Chimica Acta Vol. 567; p. 122050
• Main Authors: Fetimi, Abdelhalim, Kebiche-Senhadji, Ounissa, Benguerba, Yacine, Albakri, Ghadah Shukri, Alreshidi, Maha Awjan, Abbas, Mohamed, Hamachi, Mourad, Bahita, Mohamed, Merouani, Slimane, Yadav, Krishna Kumar
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2024
• Subjects: Artificial neural network; Chromium; Particle swarm optimization; Polymer inclusion membrane; Water pollution; Chromium; Water pollution; Artificial neural network; Polymer inclusion membrane; Particle swarm optimization
• ISSN: 0020-1693
• DOI: 10.1016/j.ica.2024.122050

[Display omitted] •This study utilizes ANN-PSO algorithms to estimate the Cr(IV) extraction %.•PSO is used to optimize ANN weights and thresholds.•ANN- PSO’ coefficient of determination (R2) exceeds 0.999 after training.•The ANN-PSO model outperforms the SCG and MLP models in terms of accuracy (R2) and RMSE. In this study, three machine learning (ML) algorithms (scaled conjugate gradient – SCG, multilayer perceptron – MLP, and a novel hybrid artificial neural network algorithm – ANN-PSO) were employed to forecast the efficiency of removing heavy metals from aqueous solutions through the polymer inclusion membranes (PIMs) process, with a focus on chromium (Cr) removal efficiency. Operational parameters were varied to adjust predictive models, including time, PVC molecular weight, Aliquat 336 extractor concentration, and initial chromium (VI) content. Results indicate that the ANN-PSO model outperforms SCG and MLP models, exhibiting lower Mean Squared Error (MSE) values and remarkable R-squared (R2) values exceeding 0.99, suggesting strong predictive capabilities. Thus, the ANN-PSO model offers a robust and practical choice for optimizing the PIM process with minimal reliance on experimental work, contributing significantly to the scientific understanding of heavy metal removal methodologies.