Hi All.
How can we implement INTERACT(INT) feature selection process in WEKA ? I think SymmetricalUncertainityAttributeEvaluation (SU) is not enough. Right?
It also says after SU, do ''c-contribution''. But I couldn`t find it.
When I apply SymmetricalUncertainityAttributeEvaluation, it takes all the features(by default).. I have to do one more thing but I could`t figured it out.
Sincerely, I do not know. The WEKA has a friendly interface . I think u . should revolve your problem in a layer below. Doing an analysis of prior data, using for example , some commands R or RS to treat their data before entering WEKA . Look on the internet this kind of data analysis has a lot of thing to be ready utilizado.veja eg https://mineracaodedados.wordpress.com/tag/algoritmo
I don't know this "Interact" method specifically. But I'm assuming that it does not exist by default on WEKA and you want to implement it.
Before start you should answer to yourself these questions:
1) Are you confortable with Java programming language and object-oriented programming?
2) Is there other programming language in which your method is already implemented and is easy to use?
If you answer "NO" to question 1) or "YES" to question 2), I would suggest that you implement/use your method in other programming language and generate the modified database in an .arff or .csv file for posterior use in WEKA.
If you answered "YES" to question 1) and "NO" to question 2), you can implement your method as one of WEKA classes (maybe you can also share your code with the community).
Look in WEKA javadocs ... I guess you will have to implement a class using the "ASEvaluation" interface, then you can execute the feature selection through the use of a filter. For an example of class doing this, look at "ChiSquaredAttributeEval".
Regards,
Leno.
http://weka.sourceforge.net/doc.stable/
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software
- System Software
- Operating Systems