Study on the portrait of electricity bill-sensitive users in urban villages based on three channels

Urban villages face multiple challenges in electrical transformation from technical, management, and user perspectives as a special area in urbanization. Accurately identifying electricity-sensitive users is a crucial step. Most existing studies focus on general scenarios and lack targeted analysis of urban villages' complex electricity usage patterns. This paper proposes a three-channel method for identifying electricity-sensitive users based on feature fusion and XGBoost algorithms to address this. Integrating data from 95598 work orders, user call information, and billed electricity fees, combined with feature importance evaluation using XGBoost and Bagging ensemble algorithms, a predictive model is constructed for different channels to create power user profiles for urban villages. The experimental results show that the three-channel model significantly outperforms the traditional single-channel method regarding F1 score, AP value, and operational efficiency, with an identification accuracy rate of 99.4%. It effectively reduces the cost of manual collection and optimizes resource allocation. The method provides a data-driven solution for user management in urban village electrical transformations, offering significant practical value.