Research on abnormal electricity rate data processing based on DC-XGBoost

The abnormal situation of electricity rate data has a significant impact on both power supply enterprises and electricity customers, which not only touches on the fundamental interests, but also may hinder the effective recovery of electricity rate. The previous technology has difficulties in obtaining the label of training data, and faces challenges such as uneven sample distribution, insufficient detection coverage of a few categories of samples, and complex operation processes that are difficult to be widely applied. To address these challenges, this paper deeply analyzes the structural characteristics of power consumption data and applies machine learning methods to achieve intelligent detection. The research content mainly includes the following aspects: Firstly, in order to deal with the problem of sample imbalance, an algorithm based on KUS-KSOS is proposed to improve the detection coverage and accuracy of abnormal power consumption data. Secondly, considering the correlation between power consumption data features, this paper proposes a DC-XGBoost algorithm based on XGBoost algorithm, combined with Pearson correlation coefficient and distance correlation coefficient, to improve the efficiency and accuracy of detection. The experimental results show that compared with the original model, the new model has improved recall ratio and AUC value, which verifies its effectiveness and practicability in practical application.