The Assessment of Water Quality Forecasting Using AI-Based ML Algorithms

Water quality matters to people, animals, plants, ecosystems, and entities. Recent environmental damage and contamination have harmed water purity. Because they indicate water authenticity, the Water Quality Index (WQI) and Water Quality Classification (WQC) are difficult to predict. In this paper, KNN imputers improve many MLalgorithms for water quality prediction. The precision of current methods is not good enough. Additionally, there are missing values in the dataset that are currently accessible for study, and these missing values significantly impact the classifiers' performance. This paper proposes an automated water-quality prediction system that effectively handles missing data while achieving high forecast accuracy. Furthermore, the accuracy of the suggested approach is assessed in relation to that of four machine learning methods. BPNN, RBFNN, SVM, and K-Nearest Neighbours are these approaches. These approaches model and predict water quality parameters such as DO, pH, NH3, NO3, and NO2. For the purpose of evaluating the accuracy of the various approaches to prediction, published data were utilized and for DO prediction, BPNN, RBFNN, SVM, and KNN had Pearson correlation values of 0.60, 0.99, 0.99, and 0.99. BPNN, RBFNN, SVM, and LSSVM had Pearson correlation coefficients of 0.56, 0.84, 0.99, and 0.57 for pH prediction. For NH3-N forecasting, BPNN, RBFNN, SVM, and LSSVM had Pearson correlation coefficients of 0.28, 0.88, 0.99, and 0.25. For NO3-N prediction, BPNN, RBFNN, SVM, and LSSVM obtained coefficients of correlation of 0.96, 0.87, 0.99, and 0.87. With correlation ratings of 0.87, 0.08, 0.99, and 0.75, BPNN, RBFNN, SVM, and LSSVM projected NO2-Ncorrelated coefficients. SVM was used to forecast water quality for groundwater-based industrial aquaculture systems. Most precise and reliable predictions were made with SVM. Both published and commercial aqua farming system data showed that the support vector machine (SVM) had the highest prediction accuracy, with 99% accuracy. For commercial farming water quality modeling and forecasting, utilize the SVM model.est.