Data Mining on Source Water Quality (Tianjin, China) for Forecasting Algae Bloom Based on Artificial Neural Network (ANN)

Harmful algae in source water become a very serious problem for water plants in China. Artificial neural networks (ANN) have been successfully used to model primary production and predict one-step weekly algae blooms in reservoir. In this study, to avoid selecting inputs randomly during the establishment of feed forward ANN forecasting algae two days later, we use correlation coefficient and index clustering to analyze source water quality parameters totally about 1744 daily measured data from 1997 to 2002 of Tianjin. So twenty-six schemes of input variables are determined and experimented for optimal inputs. Inputs of the final model are chlorophyll-a, turbidity, water temperature, ammonia, pH and alkalinity. The correlation coefficient of output values of the model and real values can reach 0.88 and the prediction accuracy is over 85%.