论文部分内容阅读
马田系统是由日本著名质量工程学家田口玄一提出的一种模式识别方法,它将正交试验设计、信噪比与马氏距离进行集成,筛选有效特征变量,对待测群体进行诊断、评价和预测.马田系统利用正交表和信噪比筛选特征变量可能存在不足之处,而粗糙集是处理不完善、不确定数据等不完全信息并能进行属性约简的有效方法,引入粗糙集筛选有效特征变量以改进马田系统.癌细胞的及早发现有助于乳腺癌的早期预防和及时治疗,以乳腺癌细胞的分类检测为背景,选取UCI数据库中600个细胞作为研究样本,使用改进马田系统方法区分正常细胞和乳腺癌细胞,并将其分类效果与经典马田系统相比较.结果表明,基于粗糙集的改进马田系统对乳腺癌细胞的分类正确率高于经典马田系统,粗糙集方法大大减少了特征变量个数,可简化数据的收集工作,为医疗上乳腺癌疾病的早期诊断及其他实际分类工作提供技术参考.
Martin system is a famous Japanese quality engineer Taguchi Xuan Yi proposed a pattern recognition method, it orthogonal design, signal to noise ratio and Mahalanobis distance to integrate the effective feature variables, the test group diagnosis, Evaluation and prediction.However, there are some shortcomings in Martin’s system to filter eigenvalues using orthogonal tables and signal-to-noise ratio. Rough set is an effective method to deal with imperfect information such as imperfect and uncertain data, Rough set to screen effective characteristic variables to improve the Martin system.Immunity detection of early cancer cells is helpful to the early prevention and timely treatment of breast cancer.Under the background of breast cancer cell classification detection, 600 cells in the UCI database were selected as the research samples, The improved method was used to distinguish normal cells from breast cancer cells, and its classification effect was compared with that of classical Martin system.The results showed that the modified Martin system based on rough set improved the accuracy of classification of breast cancer cells compared with classical horse Field system, rough set method greatly reduces the number of characteristic variables, which can simplify the data collection, for the early diagnosis of breast cancer and medical diagnosis and Other actual classification work to provide technical reference.