Confusion matrix neural network
    8 次查看(过去 30 天)
  
       显示 更早的评论
    
Hello,
The confusion matrix for my NN for classification is below. I'm struggling to understand the numbers. I know the overall correctly classified data is 81.5%.
I would be grateful if somebody had the time to answer a couple of questions.
If I take the 1st row as an example.
What does 22.2% represent?
What does 4 and 14.8% represent?
What does 60.0% (in green) represent?
What does the bottom grey row represent for example 85.7%?
Thanks for your help

0 个评论
采纳的回答
  Greg Heath
      
      
 2013-5-7
        Look at column one for class 1 targets
There were 7 class 1 targets
 6 were assigned correctly(GREEN) to output class 1
 1 was assigned incorrectly (RED) to output class 2
 0 were assigned to output class 3 (Ignore colors for 0 entries)
 100*6/7 = 85.7% (GREEN)of class 1 targets were correctly assigned
 100*1/7 = 14.3% (RED) of class 1 targets were incorrectly assigned
Look at row two for targets assigned to class 2
17 targets were assigned to output class 2
1 target from class 1 was incorrectly(RED) assigned to class 2
16 targets from class 2 were correctly(GREEN) assigned to class2
100*16/17 = 94.1%(GREEN) of assignments to class 2 were correct
100*1/17 = 5.9%(RED) of assignments to class 2 were incorrect
Look at interior square percentages
There were 6+1+16+4 = 27 targets
 100*6/27  = 22.2%
 100*1/27  =  3.7%
 100*16/27 = 59.3%
 100*4/27  = 14.8%
If it makes you feel any better, I do not like the format (e.g., I used to use the rows for target classes). However, using the column target format, I use a count confusion matrix and a percent confusion matrix:
 6   0  4  10   
 1  16  0  17 
 7  16  4  27
and
85.7    0  100   37
14.3  100    0   63
100    100  100  100
Hope this helps.
Thank you for formally accepting my answer
Greg
0 个评论
另请参阅
类别
				在 Help Center 和 File Exchange 中查找有关 Deep Learning Toolbox 的更多信息
			
	Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!



