Using classic method to obtain fatigue properties of materials is costly and time consuming. In recent years, Self-heating method has been proposed as a faster and cheaper alternative to classic fatigue tests. This method, which is based on temperature measurements, is able to predict the fatigue properties of materials in a short time by subjecting the specimens to a low number of loading cycles. Since fatigue loading time is short, this method does not damage the specimen; thus the specimen is reusable. Consequently, by using this method, one can save a lot in testing costs. In the present work, by using a probabilistic two-scale model, relations of Shape Memory Alloys (SMAs) are implemented in Self-heating method to study fatigue of these alloys. This model is capable of predicting the S/N curve of the specimen as well as its scatter by using probability relations. Moreover, the presented approach can predict the S/N curve of a specimen for any failure probability in a much shorter time compared to the classic method. To investigate validity of the numerical results, the theoretical predictions were compared with experimental data. This comparison showed a good agreement between the results produced by the model and the experimental ones which indicates the reliability of the model. Keywords: Fatigue, High Cycle Fatigue, Shape Memory Alloys, Self-heating Method, Thermography