I have data in this format: I have tuples of the amount of data and the processing time per function. I want to do a classification by the "class" attribute
Here is a sample:
Amount-F1 Time-F1 Amount-F2 Time-F2 [...] Class
50 10 20 10 1
20 2 100 20 3
...
How should I build the arff file? Should I use the relational attribute for the (Ammount,Time) tuples or should I use "regular" attributes?
Can you make me a sample arff file for my example please?
Thank you
WEKA can also work with csv files. But if you want to use arff-format, WEKA does support: weka.core.converters.ArffSaver / weka.core.converters.CSVLoader
You could split each tuple into 2 separate features:
@RELATION yourTable
@ATTRIBUTE Amount-F1 NUMERIC
@ATTRIBUTE Time-F1 NUMERIC
@ATTRIBUTE Amount-F2 NUMERIC
@ATTRIBUTE Time-F2 NUMERIC
@ATTRIBUTE Amount-F3 NUMERIC
@ATTRIBUTE Time-F3 NUMERIC
...
@ATTRIBUTE Class {1,2,3} % your class labels
@DATA
50, 10, 20, 10, 1
20, 2, 100, 20, 3
...
or use an Aggregation of Time and Amount:
Amount-F1 and Time-F1 as F1:
@RELATION yourTable
@ATTRIBUTE F1 NUMERIC
@ATTRIBUTE F2 NUMERIC
@ATTRIBUTE F3 NUMERIC
...
@ATTRIBUTE Class {1,2,3} % your class labels
@DATA
5, 2, 1 % 50/10, 20/10, 1
10, 5, 3 % 20/2, 100/20, 3
...
Depending on the use-case, generally I would prefer the second option.