Matlab中混淆矩阵的最小值和最大值

Question

In Matlab I have a cell of 100 confusion matrices.

For example confusionMatrix{1} =

16     1     0
 0    12     2
 0    10     9

How do I find the two confusion matrices out of these that have the best and worst prediction rates (sum of non-diagonal elements)?

Answer 1

You can discriminate the non-diagonal elements of a matrix A by using triu displaced up for both the matrix and its transpose: triu(A,1) and triu(A',1):

sum(sum(triu(a,1)+triu(a',1)'))

If you want to do so for all cells in confusionMatrix{:} you can use cellfun, and then sort:

prediction_rate=cellfun(@(a)  sum(sum(triu(a,1)+triu(a',1)')), confusionMatrix);
[r,idx]=sort(prediction_rate);
best_rate=r(end);
best_rated=idx(end);
worst_rate=r(1);
worst_rated=idx(1);