Abstract:Based on the basic data of 11 factors that influenced the agricultural water consumption of Qinghai Province from 2000 to 2009, a partial least squares regression model was established. Considering the practicality and accuracy of the model, the backward method was used to improve the partial least squares method. Five unnecessary variables were excluded, and a higher prediction accuracy was obtained. Data from 2010 to 2013 were selected to verify the model. The results show that the prediction using the partial least squares method agreed with the actual situation, and the improved model showed even greater agreement. Therefore, the partial least squares method is highly applicable to the prediction of the agricultural water consumption of Qinghai Province. The improved partial least squares method greatly simplifies the original model, improves the prediction accuracy, and provides a basis for the prediction of the agricultural water consumption of Qinghai Province.