Hybrid Prediction Model of Electricity Default Risk

• 1、School of Mathematics and Statistics, Chongqing University,Chongqing 401331
• 2、School of Mathematics and Statistics, Yili Normal University, Yili, Xinjiang 853000

Abstract： The operation mode of power grid companies is "use electricity first -- pay later". High-voltage customers consume a lot of electricity, which accounts for more than 70% of the electricity generated every month. Therefore, there is a great potential risk of electricity charge recovery. If the payment defaults, the power company will be difficult to make up for the loss. This article first screening out the seven main factors influencing the electricity arrears, and then use the neural networks to predict high-pressure customer risk score values. After that, we combine the risk score values and seven main influence factors to establish a logistic regression, predicting high-pressure customer risk. By doing so, we change the risk management model from a labor intensive model to a technical one, timely warning high-risk power customers. The experiment proves that the mixed prediction model of neural network and logistic regression established in this paper has high prediction accuracy and good consistency with the actual situation, so as to avoid the power company\'s losses caused by high voltage users\' breach of contract.

Keywords： Risk Prediction, Neural Network, Logistic Regression, High Voltage Customers, Power Grid.