Hi,
I see in the project front page that the continuous data should be discretized. But:
1- When I give it continuous data I do not get any error.
2- If I want to discretize data, is it sufficient to just run for example sklearn's KBinsDiscretizer and give it the required number of bins with uniform strategy? This will not be the same as the strategy mentioned in the paper (i.e. considering mean+/-std and binning to -1, 0, and 1 accordingly) but I do not think that really matters. If the feature values are changed by a scale, why the feature selection mechanism should give us different results?
I hope to hear from you soon.
Best Regards
Shahriar
Hi,
I see in the project front page that the continuous data should be discretized. But:
1- When I give it continuous data I do not get any error.
2- If I want to discretize data, is it sufficient to just run for example sklearn's KBinsDiscretizer and give it the required number of bins with uniform strategy? This will not be the same as the strategy mentioned in the paper (i.e. considering mean+/-std and binning to -1, 0, and 1 accordingly) but I do not think that really matters. If the feature values are changed by a scale, why the feature selection mechanism should give us different results?
I hope to hear from you soon.
Best Regards
Shahriar