BACKGROUND: Given the absence of a predictive random survival forest (RSF) model for retinoblastoma (RB) patient prognosis, this study intends to build an RSF model. This study aimed to investigate the prognostic factors of patients with RB and provide experience for the diagnosis and treatment of RB patients in clinical practice. METHODS: The Surveillance, Epidemiology, and End Results (SEER) Stat Version 8.4.3 software was employed to download data from the SEER database. The relevant data of 577 patients diagnosed with RB from January 1, 2000 to December 30, 2019 were collected. The follow-up period for each patient began at the diagnosis of RB and ended at the time of death. The entire study cohort was randomly allocated to a training set and a validation set in a 7:3 proportion. Potential predictive factors and feature selection were appraised through the calculation of feature importance under the random forest framework. The optimal mtry and nodesize tuning parameters (mtry =10, nodesize =25) for the random forest model were found by means of the out-of-sample error. The optimal ntree (ntree =160) was selected by the learning curve. Based on the above selected parameters, a random survival model was established. The predictive performance was validated and evaluated by internal validation and consistency index (C-index), calibration curves, and area under the receiver operating characteristic curve (AUC). RESULTS: In this study, 577 patients were totally included, of whom 17 died. There were 279 males (48%) and 298 females (52%). There were 13 features included in the model, including stage, T-stage, M-stage, surgery radiation sequence, radiation, systemic therapy surgery sequence, age, chemotherapy, surgery, race, residence, sex, and primary sequence. The C-index value of the training cohort was 0.9803, and the C-index value of the validation cohort was 0.9122. The AUCs of the model for predicting mortality at 3, 5, and 10 years were 0.983, 0.986, and 0.996 in the training set, and 0.892, 0.910, and 0.904 in the validation cohort. CONCLUSIONS: We have established an RSF model with superior predictive performance based on simple variables in the SEER database. The most crucial variable in the model is Stage, followed by M stage and T stage, which may help evaluate the prognosis of high-risk RB patients.