A Proposal to Effective Long Itemstes  Association Rule Mining LIARM (Applied  on proposed stego–images dataset)

Soukaena H. Hashem; Alia K. Abdulhassan; Hala B. Abdulwhab

Volume : II, Issue : IX, September - 2013

A Proposal to Effective Long Itemstes Association Rule Mining LIARM (Applied on proposed stego–images dataset)

Soukaena H. Hashem, Alia K. Abdulhassan, Hala B. Abdulwhab

Abstract :

An attempt to improve the efficiency of the association rules technique in data mining is done by proposing an algorithm that deals with databases with extremely long itemsets. This algorithm which focuses on the long size of itemsets is called LIARM (Long Itemsets Association Rules Mining). The set of the most frequent itemsets has been found based on finding the longest frequent itemset by using a proposed traversing approach called two–phase traversing approach that aide to reduce the large search space in which it is important to speed up the process and save the storage area. The remaining frequent itemsets are searched and appended to the previous set. The complete set of frequent itemsets has been classified to normal, closed and maximal to support and strengthen the prediction in the analysis stage. And then the set of association rules for the itemset�s has been extracted using the traditional association rules procedure. The proposal applied on proposed system is to improve the usage of steganography by the following considerations: • First build a proposed database contains a huge number of stego–images with all the feature of the stego hiding and detecting operations. • Mining the proposed database using proposed LIARM technique will be effective because the itemsets which represent the stego features are very long and this will take considerable amount of time by using the traditional association rule mining. • After generating association rules the analysis stage will begin, this stage will explain the discovered knowledge, the new patterns will represent the relationships among the features of stego. These patterns will predicate a new knowledge that will be useful for administrators to improving the steganography to be immune against the steganalysis.