Electrical storm (ES) is a life-threatening heart condition for patients with implantable cardioverter defibrillators (ICDs). ICD patients experienced episodes are at higher risk for ES. However, predicting ES using previous episodes' parameters recorded by ICDs have never been developed. This study aims to predict ES using machine learning models based on ICD remote monitoring-summaries during episodes in the anonymized large number of patients. Episode ICD-summaries from 16,022 patients were used to construct and evaluate two models, logistic regression and random forest, for predicting the short-term risk of ES. Episode parameters in this study included the total number of sustained episodes, shocks delivered and the cycle length parameters. The models evaluated on the data sections not used for model development. Random forest performed significantly better than logistic regression (P <; 0.01), achieving a test accuracy of 0.99 and an Area Under an ROC Curve (AUC) of 0.93 (vs. an accuracy of 0.98 and an AUC of 0.90). The total number of previous sustained episodes was the most relevant variables in the both models