MATLAB的ROC曲线绘制比较研究

基于非计算机专业人员的计算机应用能力稍差，而又要经常使用受试者工作特征（ROC）曲线作为判断的评价方法，介绍了ROC曲线的概念、原理和优点，以及如何在MATLAB下绘制ROC曲线。以评价分离信号与噪声数据为例，分别采用两种方法来绘制ROC曲线。一是基于ROC基本概念和Excel的计算功能，分别取不同的阈值得到相应的敏感度和特异度，然后绘制出ROC曲线；二是依据MATLAB工具下的直方图和ROC的概念，编程完成绘制ROC曲线。第一种方法容易理解、简单，但是必须要多次打开Excel才能完成选择不同阈值得到多组敏感度和特异度的目的；第二种方法应用MATLAB软件编写程序，程序短小、简单，易使用，而且可随意设置修改阈值，得到相应的敏感度和特异度以及绘制出ROC曲线。

Abstract:

 Based on non-computer professionals with poor ability of using computer，but frequently used ROC curve as the evaluation method， this paper introduced the concepts，principles and advantages of the ROC curve，and how to draw a ROC curve with MATLAB software. As an example of evaluating separate signal and noise data，two methods were used to draw the ROC curve. First，based on the basic concepts of ROC and Excel tools，it got the corresponding sensitivity and specificity at a different threshold，then draw the ROC curve. Second，based on histograms of MATLAB and concepts of ROC curve，it programed to draw the ROC curve. So the first method was simple and easy to understand，but it must to open Excel repeatedly for choosing different thresholds to get multiple sensitivity and specificity. The second method applied MATLAB software to program，the program was short，simple and easy to use，and also modified the threshold value getting the corresponding sensitivity and specificity，as well as drawing the ROC curve.