基于MFOA-SVM算法的乳腺肿瘤识别

Breast Cancer Recognition Based on MFOA-SVM Method

针对乳腺肿瘤良恶性二值分类的特点,提出了一种基于修正的果蝇优化算法和支持向量机(MFOASVM)的乳腺肿瘤识别方法.为提高SVM分类器的泛化性能,将MFOA算法引入SVM,进而优化SVM中的惩罚参数和核函数参数.为了综合评估提出算法的有效性,在威斯康新诊断乳腺癌(Wisconsin diagnostic breast cancer,WDBC)数据集进行了实验对比分析.实验结果表明:MFOA-SVM与BP,LVQ及PSO-SVM 3种方法相比,其分类准确性和稳定性显著提高.

For the two value classification characteristics of benign and malignant breast tumors,a breast tumor recognition method MFOA-SVM was proposed based on modified fruit-fly optimization algorithm and support vector machine. The MFOA was introduced into SVM to improve the generalization performance of the SVM classifer,and then the penalty parameters and kernel function parameters in SVM were optimized. In order to comprehensively evaluate the effectiveness of the proposed algorithm,the experiment and analysis were done on the Wisconsin breast cancer dataset. The experimental results showed that the classification accuracy and stability of MFQA-SVM were improved significantly compared with BP,LVQ and PSO-SVM three methods.