Breast cancer diagnosis and prediction model based on improved PSO-SVM based on gray relational analysis

Abstract

Early breast cancer diagnosis and prediction models use image data as input, which is likely to cause a large possibility of error in the conversion process of image data. Therefore, this paper proposes a PSO-SVM diagnostic prediction model called GP-SVM based on gray relational analysis (GRA) of a data set consisting of conventional sign data and blood analysis data. First of all, the original data set is optimized by gray relational analysis (GRA) to obtain a new data set. Secondly, the GP-SVM model composed of improved PSO and SVM, and uses the obtained data set as its input. The improvement point of its PSO algorithm is to dynamically adjust the inertial weights and learning factors to make the improved PSO The algorithm optimizes the parameters of SVM and balances the globality and speed of PSO algorithm convergence. On the breast cancer Coimbra data set in UCI, compared with other prediction models, the performance of the GP-SVM prediction model has better.