Emerging Patterns (EPs) as discovery knowledge from database capture emerging trends when applied in timestamped databases or capture useful contrasts between data classes when applied to datasets with classes. Moreover, EPs capture significant changes and differences between datasets are defined as itemsets whose supports (frequencies) increase significantly from one dataset to another. The changing of supports for itemsets from one dataset to another (the ratios of the two supports) is called growth rates. Furthermore, EPs use user-defined threshold in order to reduce large candidate patterns, then can be said EPs are itemsets whose growth rates are larger than a given threshold. Finally, EPs are similar to discriminant rules or evolution rules in Attribute Oriented Induction (AOI) but different since EPs do not limited by exclusiveness constraint and because the extra information of growth rate. Those EPs with very large growth rates are notable differentiating characteristic between 2 datasets and have been useful for building powerful classifiers. Thus, Those EPs with very large growth rates are frequent in one class but rare in another class. Meanwhile EPs with low to medium support such as 1% until 20% can give very useful new insights and guidance to experts, in even well understood applications. Hence, the low supports EPs such as 0.1 until 5% may be new knowledge to the dataset and discover small support EPs is interesting. The interestingness of discovery small support EPs due to reason too many EPs candidates and make naive algorithms too costly to examine all itemsets in dataset. For example if there are 350 itemsets in dataset then naive algorithm would need to process 2350 (Cartesian product) itemsets in order to find their supports in datasets D1 and D2 and then determine their growth rates. Overall, EPs algorithm can be divided become process discriminating between two datasets and classification more than two datasets.