Feature: %string to numerical value conversion

[vukosim]

You have some datasets that have % values strings e.g. '95%',''82%' etc.

It would be great if this could be automatically dealt with. On Pandas dataframe this can be done with

df = df.replace('%','',regex=True).astype('float')

[rhiever]

I like this idea. Before we implement it I want to consider how this might affect other input values that aren't percentages.

For example, what if the user passes a DataFrame with class labels that are, say, ">50%" and "<=50%"? We obviously don't want to parse that into numerical percentages, nor do we want to remove the percentages.

Perhaps one way to accomplish this is:

1. Check if the column is of type 'object'. If not, then it won't contain a '%' anyway.

2. Check if any entry in the column contains a '%'. If not, skip the column.

3. Make a copy of the column and apply the transformation you suggested. If it doesn't crash, then it very likely was a string encoding of a percentage. If it does crash, then it probably was some other string(s) that contained %s.

4. In the non-crashing case, apply the change to the column.

Are there any cases that such a procedure would miss and incorrectly encode?

[MagnetonBora]

Will try to implement this feature request.

[rhiever]

Looking forward to it! 👍

[vukosim]

Had forgotten about this, would be great 👍