Better handle categorical features

[louiskirsch]

[nikriek]

uuh, there actually is another error. We are already catching FileNotFoundError. Problem is that it throws ValueError: could not convert string to float: '09:04:45'. Did you upload a dataset with timestamps, @ArvindKumarS?

[nikriek]

My idea would be to introduce some new models. For each we would have different attributes types (float, string) and different statistical values like mode, median..

class Feature(AbstractModel):
  pass

class CategoricalFeature(Feature):
  pass
    
class ContinuousFeature(Feature):
  pass 

[nikriek]

Are you ok with the abstraction above @timediv?

[louiskirsch]

I mean, the details aren't filled in yet. But sounds pretty reasonable 👍