I'm trying to get good accuracy using WEKA and its classification options.
by using this method I can not cover all options and this is why I'm afraid I could miss the optimal classification to get the best J48 tree solution.
I have tried using number of classifications and methods such as (NB, costSensitive, attributeClassifier, etc..) and each of this have at least few options..
My question is:
is there any option to let the software (WEKA or other software) to run (even days!!) in order to find the best classification for optimal solution? If to be more specific, could I determine the confusion matrix I would like to have and the software will tell me which classification or options to use?
You should run your classifier with different parameters and check classifier performance for each parameter set. Here you have reference for WEKA class for doing this.