Research On Incremental Mining Of Multisource Data Association Rules

In the big data era,data is normally saved in several data sources.The original data mining method requires data combination and fusing before analysis.But there will be the following problems: large data size,varied data structures and privacy protection.Therefore effective data mining without data combination and fusing becomes a popular topic.Association rule is one of the most important research topics of data mining.Based on association rule,relationship and connection of data or attributes can be found out,which will support further work like intelligent prediction or intelligent recommendation.Current researchers have achieved success on association rule mining of static centralized data.Although there are some researches on the mining of dynamic data association rules,it is not perfect.The association rules incremental mining is an important direction of dynamic association rule mining.Existing algorithms for efficient mining of association rules mining do not generate candidate sets,and they are implemented in a space-for-time way.Therefore,how to optimize the space on the premise of maintaining the time efficiency basically unchanged is a new problem of association rule updating mining.Moreover,with the diversification of data,association rules mining for different types of data is also a hot issue.This thesis is mainly about association rule incremental mining and multisource data association rule incremental mining.The summary shows as following.1.Research of association rule incremental mining.First of all,a Can tree algorithm,which has better time efficiency,is found out by analysing of current association rule incremental mining algorithm.Then,it is found that the algorithm has the disadvantage of high space occupancy rate and this shortcoming is related to the preorder order.Meanwhile a new ordering data method,based on data size,is introduced for overcome this weakness.Finally,test result is shown to verify this method.2.Research of multisource data association rule incremental mining.Firstly,the feature of multisource data,variety,dispersivity and asynchronism,and the problems that should be paid attention to in multisource data mining are analyzed.Then based on their dispersion and synchronism,a new method is introduced,which combine distribute association rule mining and association rule incremental mining.In the end,it is proved that the algorithm can solve the dispersivity and asynchronism of multisource data through a number of experiments,and compares the incremental mining algorithms of multiple association rules for the second mining,and selects one of the better two mining algorithms as the algorithm.