Objective To construct an optimal prediction model of chickenpox in Urumqi, and to provide reference for formulating the prevention and control strategies of chickenpox. Methods The multivariate autoregressive moving average model (ARIMAX) and random forest model (RF) were established based on the monthly incidence of chickenpox in Urumqi from 2014 to 2018, and the monthly incidence of chickenpox in 2019 was used to test the models and evaluate their prediction effect. The prediction performance of the two models was compared, and the best model was selected to predict the incidence of chickenpox in Urumqi. Results The incidence of chickenpox in Urumqi showed a regular bimodal distribution with obvious seasonality, and it showed a slow upward trend from July 2014 to December 2019. The fitting model was ARIMA(0,1,0)(0,1,1)₁₂, the root mean square error (RMSE) and mean absolute error (MAE) of ARIMAX model training set were 1.29 and 0.95, respectively, and the RMSE and MAE of the test set were 1.88 and 1.44, respectively. The training set RMSE and MAE of RF model were 1.56 and 1.56, respectively, and the test set RMSE and MAE were 4.83 and 3.96, respectively. Conclusion The performance of ARIMAX model is better than that of RF model, which can better predict the incidence trend of chickenpox in Urumqi. It is necessary to optimize the prediction model according to the actual situation and provide scientific guidance for the prevention and control of chickenpox.