Support vector machine classification prediction model based on improved chaotic differential evolution algorithm
Yaxin Hou, Xiangqian Ding, Ruichun Hou · 2017
The performance of the existing support vector machine (SVM) based diabetes classification prediction models are insufficient due to the difficulty in parameter selection. This paper proposes a SVM diabetes classification prediction model based on the improved chaotic differential evolution (ICDE) algorithm. An one-dimensional infinite fold mapping based ICDE algorithm is used to optimize the SVM penalty parameter C and kernel parameter σ. Once the optimal parameter combination (C, σ) has been found, a diabetes classification prediction model can be constructed to diagnose and predict diabetes mellitus. The proposed method has been compared with other improved SVMs, including genetic algorithm (GA) based SVM, particle swarm optimization (PSO) based SVM and differential evolution (DE) based SVM. The result shows higher classification accuracy, specificity and sensitivity.