matlab神经网络3：模式分类

1.模式分类与数据集

Pattern recognition is the process of training a neural network to assign the correct target classes to a set of input patterns. Once trained the network can be used to classify patterns it has not seen before.  This dataset can be used to demonstrate how a neural network can be trained to classify data using a set of examples.

cancerInputs - a 9x699 matrix defining nine attributes of 699 biopsies.  

 1. Clump thickness 

 2. Uniformity of cell size 

 3. Uniformity of cell shape 

 4. Marginal Adhesion 

 5. Single epithelial cell size 

 6. Bare nuclei 

 7. Bland chomatin 

 8. Normal nucleoli 

 9. Mitoses  

cancerTargets - a 2x966 matrix where each column indicates a correct category with a one in either element 1 or element 2.  

 1. Benign 

 2. Malignant

2.神经网络用于模式分类

2.1 GUI方法

输入：9个关联指标

输出：正常/异常二分类

CE：Minimizing Cross-Entropy results in good classification. Lower values are better. Zero means no error.(最小化交叉熵结果，值越小性能越好)

Percent Error: 错误分类的比例. 值越小效果越好.

评价指标1：confusion

The next figure shows the confusion matrices for training, testing, and validation, and the three kinds of data combined. The network outputs are very accurate, as you can see by the high numbers of correct responses in the green squares and the low numbers of incorrect responses in the red squares. The lower right blue squares illustrate the overall accuracies.

评价指标2：ROC

受试者工作特征曲线 （receiver operating characteristic curve，简称ROC曲线），又称为感受性曲线（sensitivity curve）。得此名的原因在于曲线上各点反映着相同的感受性，它们都是对同一信号刺激的反应，只不过是在几种不同的判定标准下所得的结果而已。接受者操作特性曲线就是以假阳性概率（False positive rate）为横轴，击中概率为纵轴所组成的坐标图，和被试在特定刺激条件下由于采用不同的判断标准得出的不同结果画出的曲线。
ROC曲线是根据一系列不同的二分类方式（分界值或决定阈），以真阳性率（灵敏度）为纵坐标，假阳性率（1-特异度）为横坐标绘制的曲线。传统的诊断试验评价方法有一个共同的特点，必须将试验结果分为两类，再进行统计分析。ROC曲线的评价方法与传统的评价方法不同，无须此限制，而是根据实际情况，允许有中间状态，可以把试验结果划分为多个有序分类，如正常、大致正常、可疑、大致异常和异常五个等级再进行统计分析。因此，ROC曲线评价方法适用的范围更为广泛。