Abstract: To study the key water quality indexes affecting the water quality of the basin, this paper selects Yihe River as the research area.The annual water quality monitoring data and laboratory sampling data of the Yihe River from 2006 to 2019 are used to evaluate and model the river water quality by using the water quality index method. Water quality index（WQI） can transform a large number of complex water quality data into a single index. This single index can reflect the overall state of water quality, so the water quality index is often used to evaluate water quality at present. A total of 10 water quality indexes including total phosphorus（TP）, pH, water temperature（WT）, dissolved oxygen（DO）, nitrate nitrogen（NO₃-N）, 5-day biochemical oxygen demand（BOD₅）, fluoride（F^-）, chemical oxygen demand（COD）, sulfate(SO₄^2-), and ammonia nitrogen（NH₃-N） are analyzed. Based on the multiple linear regression analysis, the key water quality index evaluation model WQImin of the Yihe River is established. The indexes involved in the evaluation of the Yihe River water quality are reduced.The results of this paper are as follows. When the water quality index is not weighted, the fitting degree and prediction accuracy of the fourindex water quality assessment model and the six-index water quality assessment model do not reach the highest; when the water quality index is weighted, the fitting degree and prediction accuracy of the four-index water quality assessment model and the six-index water quality assessment model do not reach the highest, too. Neither of these two simplified index models is the optimal critical water quality evaluation model in this study. Through model training and testing, the weighted five-index model WQI_min+WT^w has good water quality evaluation performance, R~2=0.972,MSE=0.51,PE=2.07%,P＜0.05, and is the optimal key water quality index model in this study. The WQI_min+WT^w model is a weighted five-index water quality evaluation model, including five water quality indexes: NH₃-N, BOD₅, DO, SO₄^2-, and WT, which shows a significant positive correlation with the WQI model（P＜0.001）. The weighted five-index model not only maintains the accuracy of water quality evaluation, but also effectively reduces the cost of water quality index detection, improves the efficiency of water resources evaluation, and can effectively replace the WQI model for water quality evaluation in the basin. In addition, the artificial neural network model is developed based on the same sample data, which can be effectively applied to the evaluation and prediction of water quality in the Yihe River. On the one hand, the artificial neural network model can provide a reference for the future change trend of water quality in the Yihe River.On the other hand, the artificial neural network model can provide a new technical way for the intelligent simulation of water environment.