In this paper, a comprehensive water quality assessment model for the seagoing rivers of the Jiaozhou Bay basin was established based on a BP neural network. In the situation investigation, a list of main assessment indexes was selected, comprising COD, permanganate, DO, ammonia, volatile hydroxybenzene and mineral oil. Then Environmental Quality Standards for Surface Water was used as the training sample and comprehensive assessment was conducted for the rivers. In Comparison with results from the conventional single-factor assessment method, this model not only responded to the comprehensive river water quality status, but also improved the speed and effectiveness of training, saving time and increasing accuracy of the assessment model through a series of design optimizations.