Sie befinden Sich nicht im Netzwerk der Universität Paderborn. Der Zugriff auf elektronische Ressourcen ist gegebenenfalls nur via VPN oder Shibboleth (DFN-AAI) möglich. mehr Informationen...
Development of AI-based hybrid soft computing models for prediction of critical river water quality indicators
Ist Teil von
Environmental science and pollution research international, 2024-04, Vol.31 (19), p.27829-27845
Ort / Verlag
Berlin/Heidelberg: Springer Berlin Heidelberg
Erscheinungsjahr
2024
Link zum Volltext
Quelle
MEDLINE
Beschreibungen/Notizen
Prediction of river water quality indicators (RWQIs) using artificial intelligence (AI)–based hybrid soft computing modeling techniques could provide essential predictions required for efficient river health planning and management. The study described the development of a novel AI-based relative weighted ensemble (AIRWE) hybrid model for predicting critical RWQIs, i.e., biochemical oxygen demand (BOD) and total coliform (TC). The study involved comprehensive water quality (WQ) monitoring from 30 locations along the Damodar River to establish the baseline data and delineate the WQ. The representative input features showing a strong association with BOD and TC were identified using Spearman’s rank-coupled orthogonal linear transformation (SOT). The relative weighted ensemble (RWE) method was applied to determine the relative weights for base learners in the AIRWE model. The statistical analysis of the developed model revealed that it was most efficient and accurate for predicting BOD (
R
2
, 0.97; RMSE, 0.06; MAE, 0.04) and TC (
R
2
, 0.98; RMSE, 0.06; MAE, 0.05) over the traditional techniques. The
t
stat
(BOD 0.02 and TC 0.47) was lesser than
t
crit
(1.672), confirming its unbiased predictions. The SOT technique removed the data noise and multicollinearity, whereas RWE curtailed the individual model’s limitations and predicted more reliable results. The model resulted 97% accuracy with high precision (96%) in classifying the river water quality for various end uses. The study describes a novel approach for researchers, scientists, and decision-makers for modeling and predicting various environmental attributes.