Energy Data Catalog Item Extraction Method Based on Semi Supervised Feature Selection

With the development of energy technology, the variety and amount of energy data increase, which brings challenges to the construction of energy data catalog system and the use of energy data. In order to extract the energy data catalog items and reduce the redundancy of the energy catalog, this paper proposes an energy data catalog item extraction method based on the semi supervised feature selection algorithm of joint mutual information. Firstly, aiming at the problem of incomplete label information of energy catalog, semi supervised learning algorithm is used to transform semi supervised data set into supervised data set. Secondly, based on the joint mutual information, the importance of all items in the energy catalog is ranked. Then, according to the different feature subsets, the classification accuracy is calculated. Finally, the final energy data catalog item is obtained through comparative analysis. The results show that the proposed method can keep a high classification accuracy even if the label information of the data sample is incomplete, which verifies the rationality and effectiveness of the method.