Abstarct: Prediction of stability conditions in mextal hard rock underground mines, due to safety and economic consequences of an inappropriate design is essential. The stability of such mines depends on the several parameters and accordingly, the prediction of stability condition with regard to these parameters is a complex problem. Since 1981, the stability graph method is used as the most common method for evaluating the stability condition of hard rock underground open stopes. As it is an empirical method and also absences of unique graph, the risk of interpreting the results is high. In addition, the gravity factor in this method for different surfaces has been considered the same. Therefore, to overcome these two major problems, three estimator models were developed to predict stability conditions of roof and walls of underground open stopes with the use of logistic regression, support vector machine (SVM) and fuzzy logic methods. For training and testing of these models, a databaxse including 84 and 78 case histories for walls and roof in Canada and Ghana were used respectively. These models were evaluated with use of different evaluation criterion. Finally, the performances of these models were compared with the common stability graph by using performance evaluation criteria. The results showed that the SVM, fuzzy logic and logistic regression models have better performance in predicting the open stopes stability rather than stability graph method.