Research And Implementation Of The New Multi-dimensionality Sequence Mining Arithmetic For Intrusion Detection System

Data Mining is the result, the database technology has been investigated by many people. Recent years, Data Mining developed very fast, the technology supported by foundation arithmetic is transforming to real applications very much, and showing great vital force. In Data Mining, the core part is the arithmetic of associate rule. The arithmetic of associate rule is used to find the interesting associate or correlation among a large quantity of data. The most typical is the Apriori arithmetic and others derived from it. The Apriori arithmetic is delivered by Agrawal in 1994. The main purpose is to mining the single-dimensionality and single-layer associate rules from the customs business database.Usually, the data mining face of huge quantities of data, much basic academic arithmetic can't use effectively in fact. The big company's arithmetic are kept secret in black box. So the task for discussion is to think out a new data mining arithmetic that can be put into the Intrusion Detection System (IDS) directed by a basic data mining arithmetic's thinking.We analyzed the AprioriAll arithmetic. It's a multi-dimensionality and multi-layer sequence pattern mining arithmetic and bases on the Apriori arithmetic associate designed by Agrawal. The AprioriAll arithmetic upgraded the Apriori arithmetic from single-dimensionality and single-layer to multi-dimensionality and multi-layer. The multi-dimensionality and multi-layer arithmetic can adapt complex and huge information data better than single-dimensionality and single-layer arithmetic apparently. We can but use the multi-dimensionality and multi-layer arithmetic to show the internal and hidden relationship.In the paper, we will discuss particularly the AprioriAll arithmetic's fundamental, realization, virtue and shortcoming. The AprioriAll arithmetic used the data structure horizontal item-transaction. After the candidate itemset built, the frequent itemset would be built through scanning the huge database once at least. And the process of scanning and analysis is the time killer and the bottleneck of the arithmetic. The most important question is reducing the time in scanning and analysis.In the paper, we designed a new multi-dimensionality and multi-layer frequent sequence mining arithmetic. After scanning the database just one time, the useful data is coded and put into a planar list. The database's scan changed to a process for scanning and maintaining a list by the database described by bitmap representation. And we use the data structure in matrix to count the candidate itemset. The times in scanning is decreased largely. Actually, the new arithmetic is much better than traditional arithmetic in solving same problem.