Power data classification method based on selective ensemble learning

The power data implies a large number of user's characteristic attributes and the user's power consumption rules. If these potential behavior attributes of the user can be mined in this way, the precise power supply on the power supply side will provide strong support. In this paper, based on the user's electricity information data, the improved TF-IDF is used to preprocess the data. The whole two-layer ensemble learning framework is adopted, and the word vector is introduced to expand the characteristics of the text. Finally, the result of the first layer is obtained. After the feature splicing with the word vector, the classification prediction is performed through the CNN network, and the final prediction model is obtained to predict and classify the user's power usage behavior. Compared with the traditional CNN model, the classification effect of this paper has been significantly improved. © 2020 Computer Society of the Republic of China. All rights reserved.