Research on Recognition and Classification of User Stealing Detection Based on Weighted Naive Bayes

With the increase of electricity consumption in various industries, there are also many losses in power grid electricity, including non-technical electricity losses of customers’ illegal consumption behavior, which has a negative impact on the company and society. In order to facilitate power theft supervisors to supervise users’ behaviors, and to reduce the number of users who conduct power theft behaviors, this paper proposes a naive Bayes method based on feature weighting to identify users of stealing electricity according to the preprocessing strategy of screening the characteristics of electricity stealing data. On the one hand, the attribute characteristics are filtered according to the calculated characteristic information value, reducing the impact of attribute correlation and unbalanced sample distribution on model performance. On the other hand, the weight coefficient of each attribute feature is further calculated by the feature information value, eliminating the influence of the inconsistency of the influence degree of each attribute on the performance of the model, and electricity theft will be judged. Finally, it was verified using the residential electricity consumption data in the Irish smart meter dataset. Experimental results show that this method effectively ensure the integrity of the electricity theft detection information. Compared with the proposed k nearest neighbor and logistic regression algorithm, the accuracy of the electricity theft detection is significantly improved, and it can be applied to the actual task of power theft detection.