Objective:To establish a prediction model for symptomatic radiation pneumonitis (SRP) after radiotherapy for thoracic cancer based on a longitudinal cohort and dose interval variations.Methods:Clinical data of 587 patients who received thoracic radiotherapy in Department of Radiotherapy of Yunnan Cancer Hospital from July 2022 to June 2023 were retrospectively analyzed. The National Cancer Institute common terminology criteria for adverse events (CTCAE) version 5.0 was used to grade radiation pneumonitis, and clinical factors, traditional independent dosimetric characteristics and dose interval variation characteristics were collected. Features used to predict the occurrence of SRP were screened using genetic algorithms and analyzed the correlation between the selected features and SRP occurrence. Predictive models for SRP occurrence were established using the selected features and evaluated, and the optimal predictive model was visualized using a column chart.Results:The incidence of SRP was 35.94%. Five clinical factors, seven independent dosimetric features and six dose interval variation features were screened out by genetic algorithms to effectively predict the occurrence of SRP. The area under ROC curve (AUC) of clinical factors combined with traditional independent dosimetric factors and dose interval variation factors was 76%. The AUC of clinical factors combined with traditional independent dosimetric factors and that of clinical factors combined with dose interval variation factors was 69% and 67%, respectively. The addition of the characteristics of dose interval variation factors significantly improved the effectiveness of the prediction model.Conclusions:The supplement of the characteristics of dose interval variation factors can significantly improve the performance of the SRP prediction model for thoracic tumors after radiotherapy. The SRP prediction model based on dose interval variations can effectively predict the occurrence of SRP.