The use of PCA and ANN to improve evaluation of the WQIclassic, development of a new index, and prediction of WQI, Coastel Constantinois, northern coast of eastern Algeria

Abstract

Abstract The objective of this research was to arrive at a better assessment of the quality of surface water in the Constantine region. The focus was on the comparison of three classical indices WQINSF (National Sanitation Foundation Water Quality Index), WQICCME (Canadian Council of Ministers of the Environment Water Quality Index) and WQIAP (weighted arithmetical Water Quality Index), the development of a new index and the prediction by ANN (artificial neural network) of WQI indices. The principal components analysis (PCA) allows the selection of 10 parameters to be used in the calculation of the classical WQI, and eight principal components to be used as input for the new proposed index (regularized WQI). However the ANN is applied for the search for prediction models of classical WQI and developed WQI. The results show that the WQIAP index assesses water quality better, and that the regularized WQI further promotes the assessment of water quality. WQIR shows that, after the pollution peak, the water quality does not return to its initial state. The modeling approach by ANN offers an effective alternative to predict the WQI, it subsequently appears that the ANN predicts the new index WQIRregularized (R2 = 0.999) better than the classic model WQIAP (R2 = 0.99).