Based on chaos theory local-region method is more advanced forecasting technology, for the prediction of urban daily water supply system with strong nonlinear and non-stationary characteristics, usually chaotic local method is more applicable. for chaos local-region method of daily water supply capacity prediction, most important is the use historical information, namely on adjacent points selection and use, but few people consider the adjacent points selection in the urban daily water supply forecast by using the method of local-region method, or only consider the number of adjacent points, or a position adjacent, no researcher combined the two cases. Accordingly, the combination method of the BIC information criterion and the attenuation coefficient mining data for phase space reconstruction, optimize adjacent points of the local-region method, and the model was used to predict the actual daily water supply of the water plant, the results show that the method can significantly improve the prediction accuracy of daily water supply, the average absolute error of the prediction is only 1.064%，it shows that the combination method is feasible.