In recent years, the surge in telecommunications fraud has become a global problem, and scholars and experts have conducted in-depth studies involving psychology, sociology and machine learning fraud detection. Such fraud seriously threatens people’s property and destabilizes society, and needs to be addressed urgently. Such events as telecommunication fraud not only bring serious threats to the property safety of the general public, but also invisibly undermine the harmony and stability of society [1]. Fraudsters rely on the convenience of modern communication technology, through the phone, SMS, the network and other channels, cleverly designed a variety of fraudulent words, they pose as public prosecutors and law enforcement agencies, bank employees or friends and relatives, luring the victim to make transfers and remittances, so as to easily cheat money. This type of fraud has a very high degree of covertness and speed of dissemination, fraudulent means are endless, so that the victim can not be defended. Once deceived, the victim’s personal property will suffer significant losses, and may even be in psychological distress, further triggering social instability.As the fight against constantly updated and iterative fraudulent techniques and the improvement of relevant laws and regulations always occur after the completion of the fraud, the people’s property safety has been violated. Taking the initiative is the right choice, identifying and predicting vulnerable groups before the fraud occurs, building a machine learning-based telecom fraud warning model, and launching intensive and targeted anti-fraud publicity and education for different vulnerable groups.