Is it possible to configure the algorithms to create rules such as feature_x>bin_value_y?
Today I often get rules that choose to condition on adjacent bins. For example, if we have bin1=1-2, bin2=2-4
Then I have separate rules, that filter on both.
The more interpretable way would be to "combine" them into a single bin (this could be more expensive computationally) or perhaps only consider the triplet (value equals bin, a value larger than a bin, a value smaller than bin) for every bin. This perhaps is more plausible - and shouldn't increase computational complexity significantly (that's my intuition at least.)
Is it possible to configure the algorithms to create rules such as feature_x>bin_value_y? Today I often get rules that choose to condition on adjacent bins. For example, if we have bin1=1-2, bin2=2-4 Then I have separate rules, that filter on both.
The more interpretable way would be to "combine" them into a single bin (this could be more expensive computationally) or perhaps only consider the triplet (value equals bin, a value larger than a bin, a value smaller than bin) for every bin. This perhaps is more plausible - and shouldn't increase computational complexity significantly (that's my intuition at least.)
Thank you, Itay from Tel Aviv