jeudi 28 avril 2016

WEKA IP Address arff file error

I have the following ARFF Training file

@relation cowrie

@attribute SourceIP {64.31.57.18,121.34.178.154}
@attribute SourcePort numeric
@attribute Username {nano,root}
@attribute Password {'G0na,.',root}
@attribute Status {failed,succeeded}
@attribute Malicious {No,Yes}

@data

64.31.57.18,42001,nano,'G0na,.',failed,No
64.31.57.18,42131,root,'G0na,.',failed,No
121.34.178.154,43064,root,root,succeeded,Yes

I want the source ip, username and password to remain as train attributes somehow without confusing WEKA so that I am able to test an entry against this train file.

I read that WEKA only sees things as yes/no 1/0 for machine learning but I also read that some attributes can be made to be considered as nominal.

Is there a way to keep the mentioned attributes in the train file and make WEKA work with them?

Aucun commentaire:

Enregistrer un commentaire