In view of the uncertainty of indexes for evaluating water quality, in order to test the rationality of the criteria in evaluating water quality, resolve the incompatibility of the results of evaluating water quality indexes，and raise the distinguishment of the model of evaluating water quality，an effective and general model Projection Pursuit Classification model was proposed.The density window breadth R of the model was improved throretically, making the model more scientific and stable. Real Coding based on Accelerating Genetic was used Algorithm to find the best projective direction，and the datum of the best projective direction was used to research the level of the influence of each factor to water quality of the Yellow River.The classification results which accord with the fact were gained, providing the decision proof for protecting the water of the Yellow River reasonably.
