Accurate prediction of sensitive power customer groups can perceive customer demand and improve customer satisfaction with electricity consumption and the level of power service effectively. A power customer outage sensitivity prediction model based on Bayesian network is proposed to predict the power customer outage complaints, which defines customer power outage sensitivity data labels by customer basic information, power consumption information, smart meter energy measurement information, and user power interaction behavior, coming from 95598 customer service platform, marketing system and power information collection system.It experimentally verifies the power outage sensitivity analysis model using K-folding cross validation method, shows that the power outage sensitivity prediction model based on bayesian network has high precision in the application of power outage complaint analysis, and the experimental results demonstrate the effectiveness of the prediction model.