Power grid service classification method based on improved EM algorithm

Abstract

The invention relates to a power grid service classification method based on an improved EM algorithm. According to the power grid service classification method, acquired data are divided into a training set and a testing set and preprocessed; an appropriate attribute set is selected through an SFS algorithm; the class prior information of the training data is imported to carry out the EM algorithm to obtain a Gaussian mixed model and a class information related matrix; the matrix can be used for limiting that one certain Gaussian model is only capable of representing several specific classes; and moreover, the matrix and a parameter q based deterministic annealing EM algorithm q-DAEM are utilized to obtain an accurate clustering model. The power grid service classification method is capable of generating a Z matrix by use of the EM algorithm and the class prior information of the training set; as the Z matrix is imported into the q-DAEM algorithm, the convergence rate can be higher during iteration and an appropriate clustering model can be generated to process the power grid acquired data; as a result, the service identification capability can be effectively improved.

Description

technical field [0001] The invention relates to the field of power grid business classification, in particular to a power grid business classification method based on an improved EM algorithm. Background technique [0002] With the deepening of the construction of smart grid and "three collections and five majors", the types of business carried by the grid are increasing and tend to be complex and changeable. In order to better manage and control services, optimize network resource allocation, and customize individual requirements for different services, services must be classified. [0003] Expectation Maximization (EM, ExpectationMaximization) algorithm is an iterative algorithm that calculates maximum likelihood estimation or posterior distribution in the case of incomplete data. The EM algorithm alternately executes the E step and the M step in each iteration process, and the E step (Expectation step) calculates the conditional expectation of the logarithmic likelihood ...

AI Technical Summary

Problems solved by technology

[0004] However, the main problem with the EM algorithm is its sensitivity to the initial value, causing the result to converge to a local maximum

Images