ELECTRICITY THEFT DETECTION BASED ON SPARSE AUTOENCODER AND OPTIMIZED RUSBOOST

In order to improve the detection accuracy and reduce the computational complexity, a electricity theft detection based on sparse autoencoder (SAE) and optimized RUSBoost is proposed. According to the three aspects of the relationship between users, temperature and power consumption, the electricity users were marked as benign or malicious users. After assigning labels to the data, features were extracted from the data by introducing reconstruction based independent component analysis and SAE. Differential evolution random under sampling enhanced RUSBoost and Java optimized RUSBoost were used for classification. The experimental results of the last two data sets show that the proposed method can achieve low complexity and high-precision electricity theft detection.