C4.5 with feature selection Particle swarm Optimization

Iwan

Hi, I'm thingking about doing some classification mining using C4.5 and feature selection with PSO (Particle Swarm Optimization).

Is it possible doing it ?

I know that we can use Weka-J48 extension as replacement of C4.5. How about PSO ? Can we use Optimize weight (PSO) as feature selection ? Am i doing wrong ?

Big thanks,

Iwan

MartinLiebig

Dear Iwan,

how do you want to use PSO for FS? PSO is a minimization method which can be used at various places, not a FS technique per se.

~Martin

sigitplus

same question , please help

rubyiftikhar052

one of the research paper "Prediction of heart disease using Ensemble Learning and Particle Swarm Optimization" used pso for feature selection but i dont know how it is work... can any one help me regarding pso for feature selection in rapindminer

IngoRM

Hi,

To be honest, I would not necessarily think about PSO as a good approach for feature selection. I think genetic algorithms are a better approach for FS. Especially if you use multi-objective optimization.

Check out these links here:

• //www.turtlecreekpls.com/resource/webinar-better-machine-learning-models-multi-objective-optimization/
• //www.turtlecreekpls.com/resource/multi-objective-optimization-ebook/
  

Starting with version 9.2, we also offer a new operator which takes away a lot of the complexities of setting this up. It is called "Automatic Feature Engineering" and is also the one which is used by Auto Model. I would recommend to check out processes generated by Auto Model to see how it works. Another good resource is the tutorial process which is available in the help of this operator.

Hope this helps,

Ingo