Frequency sorted binary encoding. Gives number of binarg digits based on the frequency of its apperance in the data. So high occuring data gets represented with less digits and less frequent data gets represented by long digits.
This can be adopted to classification problem encoding but by replacing the frequency with the confusion matrix.
So the class with less confusion gets less digits and class with most confusion gets long digits.
Frequency sorted binary encoding. Gives number of binarg digits based on the frequency of its apperance in the data. So high occuring data gets represented with less digits and less frequent data gets represented by long digits.
This can be adopted to classification problem encoding but by replacing the frequency with the confusion matrix.
So the class with less confusion gets less digits and class with most confusion gets long digits.
Or maybe viceverse