I have already used cross validation techniques but facing some over-fitting problem. I have used Random forest algorithm and got 100% accuracy. I have used re-sampling techniques, used randomized and re-sample functions in weka and it worked. But i want to know whats the general procedure to overcome over-fitting before training the data.
Overfitting means that you made your model "memorize" the training set by giving "bigger" training set. Or your testing set very much resembles to your training set.
Resampling basically selects different sets for training and testing.
if you gradually decrease the size of your training set less than the first experiment you increase the change to generalize your model.