RapidMiner

Append rows with diffrent number of atribiuts

Contributor

Append rows with diffrent number of atribiuts

Hi guys!
First, im a beginner with RapidMiner so please, be patient.
I have sets of data that describes movement. I was able to create a single row form every sample (with windowing), but the thing is, every movement had diffrent lenght, so number of rows varies greatly (1k-2k of difference), and so now, I would like to append all of that data to create one nice training set, but i can't due to the diffrence. I know that I can create empty atributes columns, but doing that by hand sounds imposible. Can I do it in some "smart way"? I dont want to retransfome my date to uniform size, as lack of atribute is a great information about the movement lenght and dynamic.

2 REPLIES
Regular Contributor

Re: Append rows with diffrent number of atribiuts

Hi,

I have always appended data sets that have the same number of columns/attributes. Have you tried using the generate attributes operator to create additional attributes to ensure each data set has the same number of attributes? To denote lack of attributes, you you could use binary values; 1 for presence of an attribute and 0 if an attribute doesn't exist. I hope this helps.

Highlighted
RMStaff

Re: Append rows with diffrent number of atribiuts

Have you tried 'append' or 'union'?

If all the samples have the same attributes, 'append' operator can be used to build a merged ExampleSet from two or more compatible ExampleSets by adding all examples into a combined set.

http://docs.rapidminer.com/studio/operators/blending/table/joins/append.html

Union operator builds the superset of features of both input ExampleSets such that all regular attributes of both ExampleSets are part of the superset. If there is any column not available in one sample, it will create missing values in the merged ExampleSet.

http://docs.rapidminer.com/studio/operators/blending/table/joins/union.html