Does NaN used in confusion matrix affects results?

2 次查看（过去 30 天）

Ekta Sharma 2016-6-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/290135-does-nan-used-in-confusion-matrix-affects-results

评论： Chad Greene 2016-6-17

I have used confusion matrix and for making two matrices of equal dimensions I had used NaN but is it wise to use that? for example: A= [2 3 7 ]; B= [3 4 NaN]; C= A+B; C= 5 7 NaN I am getting where result should have been [5 7 7] according to me.Can anyone help? Thanx in advance.

3 个评论
显示 1更早的评论隐藏 1更早的评论

Ekta Sharma 2016-6-15

Sir actually 0 is one the value in my matrix as I am using chain code,in that we have directions from 0 to 7.

Ekta Sharma 2016-6-15

Is there anything else we can do for making two matrices equal in length?

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Chad Greene 2016-6-16

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/290135-does-nan-used-in-confusion-matrix-affects-results#answer_225766

在 MATLAB Online 中打开

Here's one way:

C = nansum([A; B])
C =
   5     7     7

2 个评论
显示无隐藏无

Ekta Sharma 2016-6-17

Sir may I know if I will keep using NaN in one of the matrices for making two matrices of equal dimensions.Will confusion matrix results such as true positive,false positive,true negative,false negatives values will get affected or not?

Chad Greene 2016-6-17

Using NaN as a place filler is often the best way to do it. The reason 1+NaN=NaN is because Matlab doesn't what the answer is if you add a number to something that is not a number. But the nansum function and its sisters (nanmean, nanmedian, etc) are available to simply ignore NaNs the way you want.

请先登录，再进行评论。