如何从表矩阵计算精度(How to calculate accuracy from table matrix)

如何从表矩阵计算精度(How to calculate accuracy from table matrix)

我正在使用table来显示kmeans集群的结果与实际的类值。

如何根据该表计算％准确度。 我知道如何手动完成。

Iris-setosa在群集2中全部为50，而Iris-versicolor在另一群集中有两个。

有没有办法计算％像Incorrectly classified instances: 52%

我想按类和集群打印混淆矩阵。 有点像这样：

0 1 <-- assigned to cluster 380 120 | 1 135 133 | 0 Cluster 0 <-- 1 Cluster 1 <-- 0 Incorrectly clustered instances : 255.0 33.2031 %

I'm using table to show results from the kmeans cluster vs. the actual class values.

How can I calculate the % accuracy based on that table. I know how to do it manually.

Iris-setosa had all 50 in cluster 2 while Iris-versicolor had two in the other cluster.

Is there a way to calculate the % like Incorrectly classified instances: 52%

I would like to print the confusion matrix by classes and clusters. Something lke this:

0 1 <-- assigned to cluster 380 120 | 1 135 133 | 0 Cluster 0 <-- 1 Cluster 1 <-- 0 Incorrectly clustered instances : 255.0 33.2031 %

最满意答案

您可以使用diag()在对角线上选择案例，并使用它来计算（in）准确度，如下所示：

sum(diag(d))/sum(d) #overall accuracy 1-sum(diag(d))/sum(d) #incorrect classification

您还可以使用它来计算正确分类的案例数（in）：

sum(diag(d)) #N cases correctly classified sum(d)-sum(diag(d)) #N cases incorrectly classified

其中 d是你的混淆矩阵

You can use diag() to select the cases on the diagonal and use that to calculate (in)accuracy as shown below:

sum(diag(d))/sum(d) #overall accuracy 1-sum(diag(d))/sum(d) #incorrect classification

You can also use this to calculate the number of cases (in)correctly classified:

sum(diag(d)) #N cases correctly classified sum(d)-sum(diag(d)) #N cases incorrectly classified

where d is your confusion matrix