Semi-supervised feature selection method and system, medium, equipment and terminal

• Main Authors: LI DEHUI, WU QUANWANG, GONG YANLU, SUN JIANXUN, ZENG JIE
• Format: Patent
• Language: Chinese; English
• Published: 10.06.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention belongs to the technical field of machine learning and data mining, and discloses a semi-supervised feature selection method and system, a medium, equipment and a terminal.The local data structure of a data set lacking labels is evaluated through NLS, correlation between the labels and features is calculated through MIC, and a small amount of label information in the data set is utilized; and adaptively combining the NLS and the MIC according to the conflict ratio between the neighborhood and the label, and determining the NMScore of the feature, thereby evaluating the importance. According to the method, the natural Laplacian method is improved and generated by integrating the natural neighborhood into the semi-supervised Laplacian method, and higher sensitivity can be achieved for the local data structure of the data; according to the method, the MIC is used for evaluating the correlation between the label and the feature, the conflict coefficient is innovatively used for performing weighted c