Scattered and polluting enterprise research and judgment method based on a clustering feature tree and outlier quantification

The invention relates to a scattered and polluting enterprise research and judgment method based on a clustering feature tree and outlier quantification. The method comprises the following steps of: collecting historical power consumption data including daily power consumption, valley power consumption, peak power consumption and usual power consumption of an enterprise within a month, preprocessing the historical power consumption data, and constructing power consumption features of the enterprise; constructing a clustering feature tree model by calculating clustering feature statistics according to the power consumption features of the enterprise; calculating the outlier degree through the density of clustering features and node depth, and adjusting the data outlier degree in a pruning and reconstruction process; and performing clustering by using leaf nodes of the clustering feature tree, and determining an updated data outlier degree through infection propagation of the outlier degree to make the updated