preprocessing:

discretize:
	bin: number of items in interval
    		equal frequency: kol el intervals ad b3d

    		equal width: el difference between el last element wel first element <= 10 w msh lazem kul el intervals tb2a ad b3d




preprocessing:
remove null values
discretize: if needed (apriori)
normalization: (KNN - Kmean - DBSCAN - Heirarchal)


apriori: data must be nominal (discretize)

KNN - Kmean - DBSCAN - Heirarchal: normalization 


strong rule - association rule
	(apriori)

dependency -> lift

frequent items count = sum of size of set



lazy classifier - neighbor - distance
	KNN (IBK) data might be nominal either numerical 



statistical classifier:
	(NAIVE BAYES) data must be nominal
ID3 (J48)   data must be nominal