Research On Association Rule Mining Based On Adaptive Algorithm And Parallel Computing

With the progress of science and technology and the continuous improvement of the degree of civilization of human society,the amount of data that people need to deal with is more and more big.How to dig out the knowledge and information we need from such a vast amount of data is a very realistic and important issue in the information age of such a vast amount of data.Association rule mining is an important branch of data mining.In real life,we need to consider whether there is a corresponding relationship between the association rules and classification,which is a kind of special association rules.On the one hand,in real life,the attributes of data elements may be changed at any time,the attributes of the changes will increase a lot of useful information,but also increase the difficulty and complexity of mining.On the other hand,kind of existing association rule mining algorithms are sequential algorithm,namely the use of a processor from A to Z follow the prescribed order execution,this algorithm is widely used in multiprocessor system conditions in today's computer is misfits,low efficiency,is not conducive to the practical application.In this paper,based on the above two points,the adaptive mining algorithm and parallel ideas into the class association rules mining,to improve the traditional CAR-Miner algorithm.The main research contents are as follows:?This paper briefly introduces the concept and definition of data mining,association rules and related rules,as well as the domestic and international research status of the problem of class association rules mining.?Traditional association rule mining algorithm only guarantees the completeness of the mining results,which is not considered in the mining process.In the actual situation,the number of attributes used to describe the data elements of the data set may be dynamically increased,when the traditional algorithm to re run over time to spend too much time.Therefore,according to the attribute increases,we propose a new idea of the class association rule mining algorithm can attribute to the changes made in the rapid response,improve the mining efficiency at the same time,to ensure the completeness of the result of data mining.?In order to improve the efficiency of the traditional sequential algorithm,this paper uses the parallel theory to improve the efficiency of the algorithm.We adopt two kinds of parallel strategies,which are independent class and share class,to do parallel processing,and verify the effectiveness of the two strategies through experiments.After that,we have improved the shared class to make it operate on a smaller particle size,and we compared the efficiency of the traditional algorithm,the shared class and the new algorithm through simulation experiments.By analyzing the results of simulation experiments,the improved algorithm is feasible and effective,which greatly improves the efficiency of mining association rules.But our study is not perfect,there are still many shortcomings and need improvement,such as the improvement of our proposed algorithm is only applicable to the old attribute and its value remains unchanged and the value of new properties and the corresponding join data set,which makes our proposed adaptive algorithm application range are restricted;we are not given to impose restrictions on mining results,selected users need rules etc..These are the direction we need further research in the future.