Research On Association Rule Based On Rough Set

Association rule mining is one of the key fields in the research of data mining. It reflects dependence and relationship between one certain item and others. R.Agrawal with Almaden Reaserch Center of IBM company firstly introduced association rule model and gave the resolved algorithm.Based on the famous Apriori algorithm, Most of later algorithms have focused on improving the time-efficiency. However, they failed to avoid a large number of invalid, useless and uninteresting rules for the items are assumed to have equal importance. Considering the different contribution to the mining result from attributes, for one thing ,a improved algorithm for association rule based on Rough Set is developed to solve this problem mentioned above.for another.a new improved weighted support definition is put forward to increase the effeciecy of the results.The main jobs of the thesis are as followed:Firstly, using relevant theory of rough sets, the theory of "multi- attribute indescernibility" is testified. Based on this theory, a new algorithm of association rule mining, which dose not scan the database frequently as traditional ones do but once, is put forward to improve the time-efficiency. A large number of experiments are carried out to validate its performance.Secondly, the fact that traditional support implicits two shortcomings is analyzed through typical examples. On the one hand , it fails to consider the impact of the weighting of items on rules; on the other hand, the impact of the amount of items on rules is not considered, either. To address this problem, a new support definition is proposed and the weighted association rule algorithm is given by analyses and experiments.