Power data contains a large amount of sensitive information

and the leakage of such sensitive data may have aserious impact on society.In view of the characteristics of various types of power data

complex data features

and significantimbalance in data distribution

this paper innovatively proposes a multi-type data classification and grading method.Firstly

it maps the original power data to a high-dimensional space through a context-aware power multi-type data distributionfeature learning network

so that the extracted high-dimensional vectors contain distribution information between varioustypes of data

improving the model's ability to perceive the distribution of power data at different security levels. Then

combined with the sensitive information gain change rate of noise-added features

it optimizes the high-dimensionalsensitive feature vectors containing different levels of data distribution information

and finally classifies the feature sensitiveattention score vectors through a lightweight support vector machine classification model to achieve classification and gradingof power data. Experimental results show that this method can effectively learn the distribution information of multi-classdata

improve the accuracy of classification and grading

and strongly support asset inventory and refined security protectionof power data.