一种改进的随机森林在医疗诊断中的应用  

An Improved Random Forest for Medical Diagnosis

作　　者：庞泰吾 胡春燕[1] 尹钟 PANG Tai-wu;HU Chun-yan;YIN Zhong(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算工程学院,上海200093

出　　处：《软件》2020年第7期159-163,共5页Software

基　　金：国家自然科学基金项目(61703277)。

摘　　要：快速地建立预测模型并且完成准确的分类在某些特殊的医疗诊断场合下具有重要的意义。从连续特征离散化入手,本文提出了一种改进的随机森林算法。之后使用改进的算法建立了分类模型,并在三个常用的医疗数据集上进行了实验。实验结果表明改进的随机森林算法不仅运行时间显著缩减,同时预测精度也得到了提升。更进一步的,初始的连续特征经过离散化之后变得简洁明了,这可以方便研究人员的理解。The rapid building of predictive models and accurate classification is of great significance in some special medical diagnosis situations. Based on the discretization of continuous features, an improved random forest algorithm was proposed in this paper. Then the classification model was built by using the improved algorithm and experiments were carried out on three widely used medical data sets. Experimental results show that the improved random forest algorithm not only reduces the running time significantly, but also improves the prediction accuracy. Furthermore, discretization makes the initial continuous feature concise, which is convenient for researchers to understand.

关 键 词：随机森林 连续特征离散化 决策树 算法改进 医疗诊断 分类算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]