Surface Water Quality Evaluation Based on Bayesian Network

This study aimed to establish a new water quality evaluation method based on Bayesian network owing to the shortcomings that the connections between various indicators are not considered using the traditional evaluation method. It adopted a combination of mutual information and K2 search algorithm for network structure learning and applied the maximum likelihood estimation method for parameter learning. Considering the water quality in the Hongya section of the Qingyi River as an example, the correlation and quantitative expression of the indicators were obtained. The factors that directly affected the water quality grade were found to be permanganate index, ammonia nitrogen, and total nitrogen. The water quality grade was inferred based on the quantitative relationship between indicators and water quality category. After testing, the accuracy was found to be more than 83.3%, indicating that the Bayesian network method could be used for evaluating the water quality. The process of evaluation was simple and rapid, providing a reliable basis for quickly assessing the overall water quality of the river basin. Finally, according to the method, the associated missing indicators could be predicted, providing more complete hydrological data for water environment management.