classification model - problem with date column

LeMarc

Hi,

I  trained a model on an example set which includes a column date (DD.MM.YYYY) e.g. June 2020.

However the model cant be applied to another data set [with date e.g. July 2020] with exactly the same Features as in the data set to build the model. Theres always a an error which says : "mismatch of date".

Is it possible that the model cant be applied because of differen dates in the data set?

MartinLiebig

Hi,

rule of a thumb: you don't want to use dates in a model. If you want to use date information than its DayOfMonth or MonthOfYear.

Because otherwise you run into excaclty the problem you run into. If the model found a rule "if 5th of november 1605 => remember!" then it only works for 1605, not for 2020.

Cheers,

Martin

LeMarc

Thank you @mschmitz for the hint!

Is it possible that the prediction model can check if the order of the attribute data [Format DateofMonth] is correct?

e.g. 1 1 1 2 7 2 3 3 4 4 9 4 5 5 5

MartinLiebig

Hi @LeMarc ,

what do you mean by order here? RapidMiner identifies attributes by name, not by index.

Cheers,

Martin

LeMarc

I jsut wanted to check if the values of an attribute date is within the right order. e.g. instead of Month July 2020 for all instances, theres an instances with date of xx January 2009. Outlier detection seems a better approach for that though.