Research on User Behavior Classification Method Based on Self Organizing Neural Networks

The analysis of power user load characteristics based on data mining can provide more accurate demand forecasting for the power market, thereby better meeting user needs. The article quantifies the correlation between different behavioral characteristics and user load through CTGAN - Adaboost data augmentation and Pearson correlation coefficient method, and selects key factors as independent variables for constructing a load interest mining model. On this basis, an unsupervised artificial neural network called Self organizing Map (SOM) is established, which has the ability to map high-dimensional inputs to low dimensions. The clustering results generated by SOM have high visualization and interpretability. Finally, based on the selected behavior features with strong correlation, combined with the aggregated load characteristics, different user interest features are mined in a small sample environment. The numerical analysis shows that the proposed method can obtain more effective and accurate load aggregation results, and can provide reliable basis for demand response to participate in power grid supply-demand balance regulation.