Real-time traffic flow forecasting is of vital significance to the intelligent transportation system (ITS). Providing real-time and efficient information for travellers, the forecast aims to reduce travel times, control traffic pressure, and reduce pollution. The time-series autoregressive integrated moving average (ARIMA) model has been widely used in advanced traffic management systems. Compared to the traditional ARIMA model, the autoregressive integrated moving average with exogenous inputs (ARIMAX) model can take the impact of covariates on the forecasting into account to improve the comprehensiveness and accuracy of the prediction. Based on data accessibility and correlation analysis, the upstream traffic flow is taken as the covariate X of the model. Data was collected from Interstate Highway 280 in California, with a sampling period of 5 minutes. The results showed that the ARIMAX model outperforms the ARIMA model during morning peak hours in terms of accuracy and comprehensiveness.