[Comment]
Don't categorizing continuous variables if possible. My personal case shows that it's usually true since when you want to predict/use real values, categorizing usually lose accuracy.
But how about input feature? For example, temperature, we usually only cares about certain values (25, 75 etc) there won't be any interpolation.
Link: https://discourse.datamethods.org/t/categorizing-continuous-variables/3402 Original Link: http://biostat.mc.vanderbilt.edu/wiki/Main/CatContinuous
Ref: https://stats.stackexchange.com/questions/68834/what-is-the-benefit-of-breaking-up-a-continuous-predictor-variable
[Comment] Don't categorizing continuous variables if possible. My personal case shows that it's usually true since when you want to predict/use real values, categorizing usually lose accuracy. But how about input feature? For example, temperature, we usually only cares about certain values (25, 75 etc) there won't be any interpolation.