Classification and Grading for Power Sensitive Data Based on Hybrid Data Distribution Learning
Xiuli Huang, Yujia Zhai, Congcong Shi, Yiwen Jiang, Shijun Zhang · 2023
Power industry data contain various sensitive information, and the leakage of such data will cause critical damage to both the company and the country. However, the unbalanced and various data types make the traditional data classification and grading method fail to achieve enough performance. To better classify and grade the power sensitive data, we propose a Hybrid Data Distribution Learning Network, which simultaneously learns the distribution of multiple data types, including Character-based, Text-based, and Numerical data. Furthermore, we use an optimized nonlinear Support Vector Machine model to detect the grade of power data based on the learned data distribution. The experimental results show that our method outperforms other baselines and achieves a precision of 0.994, a recall of 0.985, and an F1of 0.989. The visualization results also show that our method effectively learns the distribution of multiple types of power data.