Synergistic optimization of predictive models for water quality analysis in treatment plants using machine learning and evolutionary algorithms

• Published in: Evolutionary intelligence Vol. 18; no. 2; p. 34
• Main Authors: Ghareeb, Ahmed, Nooruldeen, Orhan, Arslan, Chelang A., Kapp, Sean, Choi, Jun-Ki
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2025; Springer Nature B.V
• Subjects: Accuracy; Applications of Mathematics; Artificial Intelligence; Bioinformatics; Chemical oxygen demand; Climate change; Control; Datasets; Desertification; Dissolved solids; Engineering; Evolutionary algorithms; Genetic algorithms; Kernel functions; Machine learning; Mathematical and Computational Engineering; Mechatronics; Optimization algorithms; Optimization techniques; Parameter estimation; Particle swarm optimization; Performance evaluation; Polynomials; Prediction models; Radial basis function; Regression; Remote sensing; Research Paper; Rivers; Robotics; Statistical Physics and Dynamical Systems; Support vector machines; Time series; Water management; Water quality; Water resources management; Water treatment plants; Iran; China; Iraq; Support vector regression; Adaptive boosting; Water quality assessment; Limited-memory Broyden–Fletcher–Goldfarb–Shanno; Particle swarm optimization; Machine learning
• ISSN: 1864-5909; 1864-5917
• DOI: 10.1007/s12065-025-01022-0

Enhancing the assessment of water quality is essential for sustainable water management, given its critical impact on environmental health and human well-being. Accurate assessment presents several difficulties due to the complexity and variability of water ecosystems. Among the various machine learning techniques employed, Support Vector Regression and Adaptive Boosting have emerged as two of the most widely used and effective methods. In recent years, many evolutionary algorithms have been developed to find approximate solutions to optimization problems. In this study, we investigate how efficient new algorithms consisting of hybrids of machine learning methods and evolutionary algorithms can be for some problems. We extensively investigate Support Vector Regression with various kernels, including linear, radial basis function, polynomial, and sigmoid. In addition, we introduce Modified Support Vector Regression and Modified Adaboost where standalone ML models are combined with the swarm-based algorithms, such as Particle Swarm Optimization, Limited-memory Broyden–Fletcher–Goldfarb–Shanno, and Distributed Evolutionary Algorithms in Python. We apply the proposed hybrid algorithms to the problem of measuring water quality. Accurate modeling and assessment of river water quality are essential endeavors with multifaceted challenges. We focus on the Tigris River modeling in Baghdad Governorate, particularly at three water treatment plants: Eastern Tigris, AL-Karkh, and AL-Wathba. The dataset comprises multiple water quality parameters, with a total dissolved solids target. Incorporating swarm-based algorithms into the adopted models has synergistically enhanced their predictive capabilities, resulting in a significant augmentation of the overall predictive prowess of the ML models. The proposed hybrid algorithms successfully enhanced SVR model performance, with the L-BFGS-B algorithm yielding mean squared error reductions of 46.23% and 28.09% on the East Tigris and Karkh plants, respectively. In comparison, the MSVR-PSO approach reduced the Wathba plant by up to 85.96%. These findings have a significant impact on advancing water treatment plants. A thorough understanding of these variations is required when deciding on the optimal machine learning model for individual applications and generalization for Tigris River water quality scale evaluation and beyond.