In order to evaluate the sensitivity of power data in business systems, the combined entropy of multi-dimensional attributes of data is used to calculate and classify data sensitivity. First, preprocess the power data, and then calculate the power data information entropy; then, filter the features through their information gain ratio and drift degree to reduce redundant and unstable features; finally, combine the availability and correlation of the data and quality, completing the calculation and grading of data sensitivity. This method fully considers the difference in sensitivity of different attribute information, can effectively and accurately measure the sensitivity of data, and improves the robustness of data sensitivity calculation.