# What happen with confusion matrix ?

3 views (last 30 days)
Oman Wisni on 21 Feb 2019
Commented: Oman Wisni on 22 Feb 2019
Hi, Im trying to create confusion matrix, but the result in the green color or true class is not 100%, if the range 1-10 it should be 10,0% but I get 9,1%. please help me if I wrong? or explain why the result like this ?
here the code and result :
targetsVector = ttes.'; % True classes
outputsVector = pred_tes.'; % Predicted classes
% Convert this data to a [numClasses x 55] matrix
targets = zeros(11,55);
outputs = zeros(11,55);
targetsIdx = sub2ind(size(targets), targetsVector, 1:55);
outputsIdx = sub2ind(size(outputs), outputsVector, 1:55);
targets(targetsIdx) = 1;
outputs(outputsIdx) = 1;
plotconfusion(targets,outputs) the cyclist on 22 Feb 2019
It looks like you have 55 observations. 51 of them were classified correctly (along the diagonal, indicated in green). But 4 of them were misclassified -- two observations with target class 1, but were in output class 7 and two observations with target class 7, but where in output class 5.
Classifiers are not usually perfect, so misclassifications happen. Did you expect your classifier to be perfect? Why?

#### 1 Comment

Oman Wisni on 22 Feb 2019
No, Im not need my classifier to be perfect. Yes I understand what the meaning of along the diagonal indicate in green. Just like anwer Mr Kevin Chng, can more specific explain. For exampel at the second row and second column, the value is 5 and the percentage is 9,1%, why it 9,1 % why not 100%. how to calculate it? can you give a simple example to get 9.1% results? why not 10.0%
Thank you

Kevin Chng on 22 Feb 2019
Edited: Kevin Chng on 22 Feb 2019
The result in the green color or true class is not 100%, if the range 1-10 it should be 10,0% but I get 9,1%.
You may find the detail of plotconfusion as below:
In the documentation, it stated :
The diagonal cells correspond to observations that are correctly classified. The off-diagonal cells correspond to incorrectly classified observations. Both the number of observations and the percentage of the total number of observations are shown in each cell.
for example, at the first row and first column, the value is 3 and the percentage is 5.5%.
It means that there are 3 predicted observation classified as Class1, the percentage of 3 of all the observation is 5.5%.

Oman Wisni on 22 Feb 2019
Yes sir, but how calculate it ? is it count with the result of predicted observation divide with sum total sample ?
I mean like below:
at the first row and first column, the value is = 3
sum total observation= 55
3/55 x 100 = 5,5
Is this true ?
Yup in this link https://www.mathworks.com/help/deeplearning/ref/plotconfusion.html, I ever read, it makes me corious with my result, in this link the example result 10,0% but why my result 9,1% ? Thank you
Kevin Chng on 22 Feb 2019
Yup, you are right,
3/55 = 5.5% means for the first row and first column.
in this link the example result 10,0% but why my result 9,1% ?
The example is not your example. In the exmple,
There are 5000 observation in total, at the first row and first column,
The predicted observation in this class is 499, so that
499/5000 = 10%
but why my result 9,1% ?
In your matrix, at second row and second column, the predicted observation in this class is 5.
5/55 = 9.1%
Oman Wisni on 22 Feb 2019