摘要
对基于多核函数的最小二乘支持向量机算法(MK-LSSVM)和采用支持向量机作为弱分类器的AdaBoost算法(AdaBoost-SVM)这两种新型的分类算法进行了研究。将这两种算法应用于求解心脏单光子发射计算机化断层显像(SPECT)图像数据的二分类问题和iris数据集的多分类问题,并从平均分类精度和平均运行时间两方面进行比较分析。最后通过Sammon映射给出了分类的可视化结果。试验结果验证了MK-LSSVM算法和AdaBoost-SVM算法的有效性和可行性,且MK-LSSVM算法在不损失分类精度的前提下,能够获得比AdaBoost-SVM更快的训练速度。
Two of the classification algorithms, i. e. , the least square support vector machine based on multiple kernel function （ MK-LSSVM ） and the AdaBoost algorithm using support vector machine （AdaBoost-SVM} as weak classifier are researched. These two algorithms are applied in solving problem of Binary classification of image data of cardiac single photon emission computerized tomography （ SPECT ） and multi- classification of iris data, and the comparison is carried out based on average classification accuracy and average running time, and visualized classification results are given via Sammon mapping. The results of tests verify the effectiveness and feasibility of MK-LSSVM algorithm and AdaBuost algorithm, and faster training speed is offered by MK-LSSVM algorithm than AdaBoost-SVM without loss of classification accuracy.
出处
《自动化仪表》
CAS
北大核心
2013年第5期13-15,19,共4页
Process Automation Instrumentation
关键词
多核核函数
最小二乘支持向量机
ADABOOST算法
神经网络
拉格朗日函数
分类精度
Multiple kernel function Least square support vector machine （ LS-SVM } AdaBoost algorithm Neural network Lagrangian function Classification accuracy
