Aiming at the shortcomings of traditional Adaptive Neural-Fuzzy Inference System (ANFIS) in water quality prediction, such as low learning efficiency and poor prediction accuracy, this paper proposed an optimal coagulant dosage prediction hybrid model based on fuzzy C-means clustering algorithm (FCM) and improved sparrow search algorithm (SSA). The hybrid prediction model is named as FCM-ISSA-ANFIS. Firstly, the water quality data of drinking water treatment plant (WTP) are statistically characterized and Pearson correlation analysis is used to determine the input variables and output variables of ANFIS. Then, the water quality data is divided into training set and test set, and the divided data sets are clustered and analyzed by FCM to determine the new fuzzy rule numbers of ANFIS. What’s more, the improved SSA is used to train the antecedent parameters and consequent parameters of ANFIS to accelerate the convergence of the algorithm and improve the ability of jumping out the local optimum. Compared with the traditional ANFIS model based on subtractive clustering, the experimental results show that the root mean square error (RMSE), mean absolute error (MAE) and standard deviation (SD) of the proposed FCM-ISSA-ANFIS for predicting the annual coagulant dosage of drinking WTP are decreased by 45.24%, 66.34% and 54.21% respectively. The proposed algorithm can not only solve the shortcomings of traditional ANFIS, but also has fast convergence and high accuracy, which can meet the real-time production demand of drinking WTP.