Abstract: The exist of residents electric larceny not only damages the economic benefits of power supply enterprises, but also affects the security of electric grid. With the rapid development of China’s digital economy and the improvement of electric acquisition system, electric larceny detected methods which are based on big data are updating constantly. The paper puts forward a new method which combines Pearson correlation coefficient, SMOTE algorithm and SVM. Firstly, the paper uses Pearson correlation coefficient to collect abnormal users’ effective power stealing data which are recorded, then SMOTE algorithm is used to enrich effective power stealing database. At last the paper builds a detection mathematical model based on the data base by SVM. By comparing the results of detection mathematical model and the reality situation of users, the validity and feasibility of the method is proved. The new method not only gives a new thought for electricity anti-stealing in power enterprises, but also improves the efficient of working team.

Key words: electric larceny detection, digital economy, Pearson correlation coefficient, SMOTE algorithm, support vector machine(SVM)