The forecast of typhoon intensity, especially the rapid intensification (RI) forecast, is still a very challenging difficulty in current typhoon forecasting. Based on the XGBoost model, this article uses the NCEP GFS analysis and forecast data in 2015-2020, and IBTrACS data to construct RI forecast model (FM) and forecast correction model (FCM) for typhoons in the Northwest Pacific 24 h in advance. Through predictor contribution analysis of the FM, we have found that the five factors that have the greatest impact on model forecasts are typhoon abundance, average temperature at 200 hPa, intensity changes over the past 6 h, potential intensity, and average divergence at 200 hPa. The model is independently tested with the data in 2021-2022, and the results show that the FM has higher accuracy when tested by analytical data, with false negative rate (FNR), false positive rate (FPR) and threat score (TS) being 0.25, 0.24 and 0.32, respectively. However, due to the influence of forecast errors caused by forecast factors, the performance of FM in real-time forecasting decreases (FNR, FPR and TS are 0.32, 0.26 and 0.27, respectively). The FCM constructed based on forecast data can effectively correct the forecast errors by learning them, thereby reducing the impact of forecast errors. The FNR, FPR and TS of the FCM in real-time forecasting tests are 0.28, 0.25 and 0.30, respectively; compared with the FM, the FNR and FPR are reduced by 0.04 and 0.01, but the TS rises by 0.03. Thus, the FCM is convenient and easy to use, and can provide reference for real-time forecasting of typhoon intensity and typhoon RI.