论文部分内容阅读
近年来功能磁共振成像(functional Magnetic Resonance Imaging,f MRI)技术在脑科学领域迅速发展,它可以将人脑在不同状态下的活动准确地反映出来,而且不会对人体造成任何创伤。与此同时,模式识别方法大量地用于医学数据分析研究中,它可以从模式分类的角度诠释生理和病理结果,因此具有广阔的应用前景。本文建立了基于支持向量机(Support Vector Model,SVM)的预测模型对一组听觉相关的f MRI数据进行分类。通过选择适当的SVM核函数类型以及优化其核参数可以提高SVM分类准确率,本文比较随机核参数的SVM和网格优化(GS)后的SVM的分类准确率。
In recent years, functional magnetic resonance imaging (functional magnetic resonance imaging, f MRI) technology in the rapid development of brain science, it can be the human brain in different states accurately reflect the activities, and will not cause any trauma to the human body. At the same time, pattern recognition methods are widely used in medical data analysis and research. They can interpret physiological and pathological findings from the perspective of pattern classification, and thus have broad application prospects. In this paper, a predictive model based on Support Vector Model (SVM) is established to classify a set of auditory related f MRI data. The SVM classification accuracy can be improved by selecting the appropriate SVM kernel function type and optimizing its kernel parameters. In this paper, the classification accuracy of SVM with random kernel parameters and SV with grid optimization (GS) is compared.