A Classification Model for Unbalanced Power Traffic

Abstract

With the continuous development of power grid informationization, the information security of the power grid is increasingly concerned. Grid traffic classification is an important basis for ensuring information security of the grid. In the process of realizing grid traffic classification, due to the different frequency of grid services and the increasing number of new services, it leads to problems such as unbalanced grid traffic data and dynamic traffic data, etc. The unbalanced traffic data causes the prediction accuracy of small categories to be much lower than the applicable standard, and the dynamic traffic data causes the model update to take a lot of time and resource overhead The dynamic traffic data causes the model update to take a lot of time and resource overhead. To solve these problems, a classification model for unbalanced dynamic grid traffic data (UDTCM) is proposed in this paper. The model uses the statistical characteristics of the flow data to detect the prediction accuracy of the classifier in time and avoid the prediction results from significantly degrading with the change of environment. Meanwhile, a resampling algorithm is used to correct the flow data to improve the data imbalance of grid flows and improve the prediction accuracy of small classes. The experimental results show that the model improves the classification of unbalanced grid flow data and reduces the time and resource overhead of model updates due to data updates.