Application of artificial neural networks to classify water quality of the yellow river

Li-hua Chen

Xiao-yun Zhang

pp. 15-23

in: Bing-yuan Cao, Cheng-yi Zhang, Tai-fu Li (eds), Fuzzy information and Engineering I, Berlin, Springer, 2009

Abstract

Within the period from 2003 to 2005 (high water, normal water and low water) 63 samples are collected and the measurement of 10 chemical variables of the Yellow River of Gansu period, are carried out. These variables are dissolved oxygen (DO), chemical oxygen demand (COD), non-ion ammonia (NH_x ), volatilization Hydroxybenzene (OH), cyanide (CN), As, Hg, Cr^6 +, Pb, and Cd. For handling the results of all measurements different chemoinformatics methods are employed: (i) The basic statistical methods that uniform design is employed to determinate the data set according to the water quality standard, (ii) MLP neural network (BP) and Probabilistic neural networks (PNN) are used to classify the water quality of different sampling site and different sampling time. The correlation between the water quality classes and chemical measurements is sought. The model between the water quality classes and chemical measurements is built, and these models could quickly, completely and accurately classify the water quality of the Yellow River.