The main of this study is to analyze recent detection methodologies of unintentional islanding in distributed generators of electric energy. This research focus the use of Unsupervised Learning Techniques and demonstrates that the use of these methodologies can achieve a performance superior to the conventional passive method of islanding detection due to non-detection relays zones. A theoretical approach is taken on the international and national standards that prohibit the islanding occurrence, and the risk and damages due its. Next, techniques for islanding detection are discussed, and comments about advantages and disadvantages of them are presented. A Simulation of Islanding and Non-Islanding operating states in a distribution system with Distributed Generation using the Simulink software is performed. The data obtained are analyzed through data grouping techniques to identify the operating states of islanding and non-islanding with a higher set rate than the conventional passive relays.