Objective To compare five time series models and predict the monthly incidence of bacillary dysentery in Qinghai Province in 2024, and provide reference for the prevention and control. Methods The epidemic characteristics of bacterial dysentery in Qinghai Province from 2014 to 2023 were analyzed. R4.3.1 software was used for establishing seasonal autoregressive integrated moving average (SARIMA) model, Holt-Winters triple exponential smoothing (Holt Winters) model, exponential smoothing (ETS) model, neural network autoregression (NNAR) model, and trigonometric seasonality, Box-Cox transformation, ARMA errors, trend and seasonal components (TBATS) model. Fitting effect of the models was analyzed and accuracy was compared. Results From 2014 to 2023, a total of 5 833 cases of bacterial dysentery were reported in Qinghai Province, without deaths, male to female ratio being 1.23 ∶1. The highest incidence was reported in 2016 (15.45 per 100 000 people), and the lowest incidence was reported in 2023 (3.68 per 100 000 people). Incidence increased from 2014 to 2016, then decreased, showing an obvious overall downward trend. Case number in < 5 years age group was the highest, accounting for 29.76% of the total cases (n=1 736). Regarding population distribution, the top three were children in childcare institutions and scattered children (35.56%), farmers (24.65%), and students (12.62%). Except the additive Holt-Winters model, the predicted trends of the other four models were consistent with actuality. The ETS model had the best fitting effect, with a relatively balanced overall performance (training set: MAE=0.13, RMSE=0.21, MAPE=19.55%; testing set: MAE=0.11, RMSE=0.16, MAPE=28.66%). It is recommended to predict the incidence of bacillary dysentery in Qinghai Province based on ETS model. Conclusion From 2014 to 2023, bacterial dysentery in Qinghai Province showed a downward trend, with the peak of the epidemic from June to August. Preschool and scattered children were high-risk groups. Among the five prediction models, ETS model has the best fitting effect, and can be used to predict the incidence of bacillary dysentery.