Container throughput forecasting plays a key role for port enterprises in operation management and decision-making. Traditional forecasting methods have low forecasting accuracy. To address this issue, a SARIMA-XGBoost combination method was proposed considering seasonality and uncertainty. The seasonal autoregressive integrated moving average model (SARIMA) model was employed to capture periodic and linear features related to container throughput seasonality, while the extreme gradient boosting (XGBoost) algorithm was utilized to adaptively learn complex patterns and nonlinear features within the time series, further mitigating the impact of uncertainty factors. Through the selection of optimization index and the calculation of corresponding weights, the effective combination of linear and nonlinear features in the forecasting model was proposed, leading to improved forecasting accuracy. Empirical research and comparative analysis were conducted by using monthly data on container throughput at Shenzhen Port from 2013 to 2022. The results demonstrate that the SARIMA-XGBoost combination method exhibits superior forecasting accuracy and stability, thus validating its effectiveness in container throughput forecasting.