Application of Artificial Neural Network in Environmental ‎Water Quality Assessment

Water quality assessment provides a scientific basis for water resources development and management. This case study proposes a Factor analysis- Hopfield neural network model (FHNN) based on factor analysis method and Hopfield neural network method. The results showed that the factor analysis (FA) technique was introduced to identify important water quality parameters. Results revealed that biochemical oxygen demand, permanganate index, ammonia nitrogen, nitrogen, Cu, Zn and Pb were the most important parameters in assessing water quality variations of the study area. Considering these parameters, water samples of the sampling sites were classified as follows: six into Class III, eight into Class IV, and six into Class V. Afterwards, a water quality map was based on the results of water quality assessment by Factor analysis-Hopfield neural network model. It showed that the southwestern part of the study area had a generally optimum water quality, while in the northeastern part, the quality was seriously degraded. Factor Analysis-Hopfield Neural Network was much better than the Hopfield Neural Network in effectively reducing the degree of Hopfield neural network over-fitting caused by the inputs, thereby achieving more reasonable results. The comparisons with BPANN, fuzzy assessment method, and the Nemerow index method indicated that the FHNN model provided more reliable judgment and valuable information than the three other water quality classification methods.