基于监督学习算法的乳腺癌复发预测  

Breast cancer recurrence prediction based on supervised learning algorithm

作　　者：刘巧红 赫英臣 牛硕 张雷明 张捷[2] Liu Qiaohong;He Yingchen;Niu Shuo;Zhang Leiming;Zhang Jie(College of Medical Instruments,Shanghai University of Medicine and Health Sciences,Shanghai 201318,China;Procurement Office,Zhoupu Hospital,Shanghai University of Medicine and Health Sciences,Shanghai 200120,China)

机构地区：[1]上海健康医学院医疗器械学院,上海201318 [2]上海健康医学院附属周浦医院采购办公室,上海200120

出　　处：《现代仪器与医疗》2023年第4期7-11,共5页Modern Instruments & Medical Treatment

基　　金：国家自然科学基金(61801288);2022年上海健康医学院师资人才百人库项目(A1-2601-22-311007-12)。

摘　　要：为了更好的发掘数据分析在辅助医学诊断上的价值,对医疗数据进行精准分析,本文基于Weka平台研究了十八种常见的监督机器学习算法对乳腺癌数据的分类预测,达到准确的预测乳腺癌是否复发的目的,并从统计学和机器学习指标对所有模型的预测效果进行对比分析。结果表明统计学指标RMSE和RRSE较小的前3位算法依次为LMT、J48、AdaBoostM1,机器学习指标中性能表现较好的算法依次为BayesNet、AdaBoostM1和LMT。综合所有指标,分类预测效果较好的算法有LMT和AdaBoostM1,在乳腺癌数据集上表现出良好的分类预测,具有价高的预测价值。In order to better explore the value of data analysis in auxiliary medical diagnosis and accurately analyze medical data,this paper studies 18 common supervised machine learning algorithms to classify and predict breast cancer data based on the Weka platform,so as to accurately predict whether breast cancer recurs.Compared with statistics and machine learning indicators,the prediction effect of all models is compared and analyzed.The results show that the first three algorithms with smaller statistical indicators RMSE and RRSE are LMT,J48 and AdaBoostM1,and the algorithms with better performance in machine learning indicators are BayesNet,AdaBoostM1 and LMT.Combining all the indicators,the algorithms with better classification and prediction results are LMT and AdaBoostM1,which shows a good classification prediction in the breast cancer dataset and has high predictive value.

关 键 词：监督学习算法 数据挖掘 WEKA 乳腺癌 预测 

分 类 号：R737.9[医药卫生—肿瘤] TP391.41[医药卫生—临床医学]