The Research And Application Of Association Rules Algorithm Based On Data Cube And Tree Structure

Because of the simple form,easy to understand,association rules mining has much theoretical research and application value as an important research branch of data mining.Association rules mining is the major means of extracting knowledge from large databases.It can effectively solve the "data rich,knowledge poor" status.In this paper,the related research for how to improve the algorithm efficiency of association rules mining has been done.And the results of research are applied to personalized e-commerce Recommendation System.The main research contents include:1.First of all,the related concept,basic skills and tasks of data mining were presented.Then the knowledge of association rules mining is introduced,focusing on the classic association rules mining algorithm - Apriori.The principles and frequent itemset generation process of apriori algorithm are described,and an analysis of the algorithm bottleneck is presented.Finally several existing methods to improve the efficiency of apriori are introduced.2.An association rules mining algorithm based on data cube is proposed.In the process of association rules mining,frequent I / O operations is one of bottlenecks effectting the efficiency of association rule mining.Although there are some improved algorithm is optimized now,when the transaction number increase sharply, the algorithm efficiency will be impactd more seriously.And the algorithm proposed in this paper combines the idea of attributes dividing,getting frequent items through inclusive relationship of the transaction.This Algorithm is not only un-sensitive to the increment of the transaction number,but also improved the efficiency.Experimental results show that the algorithm of mining association rules exsiting in a large number of transactions is quick and effective.3.An association rules mining algorithm based on tree structure is proposed.In the process of association rules mining,a large number of candidate items are also one of bottlenecks impacting the algorithm efficiency.This algorithm makes the frequent items storage into tree structure through association matrix,and takes advantage of the parts of the frequent items in the tree to build the tree structure including the whole frequent items.This algorithm not only improves candidate items producing efficiency,but also reduces the number of the candidate items.The experimental results show that it is faster and more efficient than other similar algorithm.4.Make scalability comparison of the two association rules mining algorithms proposed in this paper.Quantitatively analyze the different superiority of the two association rules algorithm in dealing with different characteristics mining object. Illustrating in the practical application,need to use appropriate mining algorithm according to different situations.It can more effectively improve the mining efficiency.5.Make the two association rule mining algorithms applied to personalized recommended system of one e-commerce site.According to the different logic business,fully performance respective algorithm mining properties,increase the off-line association rules mining efficiency of the recommendation system.In the course of user's purchase,according to the current user information,the recommendation system prompts the goods recommendation from different levels and improves the sense in user's operation process.Meanwhile,the recommendation system mining results have positive guiding role for business decision making.