A study on intelligent diagnosis model of shortwave receiving system based on improved KFCM and LapSVM

Aiming at the difficulty of obtaining a large number of labeled samples of the shortwave receiving system, an intelligent diagnosis method for the shortwave receiving system based on the improved Laplacian SVM algorithm is proposed. By introducing the idea of neighborhood density into the adjacency graph construction of Laplacian SVM, the manifold structure information of samples is more fully mined, thus improving the performance of Laplacian SVM classifier and realizing the optimization of traditional Laplacian SVM. KFCM clustering algorithm was used to select unlabeled boundary samples and labeled samples to form the reduction training set. The method of the KFCM pre-selection sample was combined with the improved Laplacian SVM algorithm to enhance the learning efficiency. The simulation results using the UCI data set and the experimental verification results of shortwave receiving system sample data indicate that the proposed algorithm could more fully mine the manifold structure information of samples and improve the performance of the Laplacian SVM classifier.