Research On Key Technologies Of High-utility Sequential Patterns Mining Based On Time Constraint

High-utility Sequential Pattern Mining(HUSPM)is a process of searching for sequential patterns with High value and high profit in quantitative sequence database.However,the traditional efficient sequential pattern mining ignores the time interval(period)when each same sequential pattern appears continuously in the quantitative sequence database.Then Periodic High-utility Sequential Pattern Mining(PHUSPM)method was proposed.In such mining method,the period value of a sequence pattern is defined as the maximum interval(period)between its occurrence in the quantitative sequence database.A sequential pattern is considered periodic if its period value is not greater than the user-defined maximum period value threshold(max Per).However,there are two problems in using max Per to judge the threshold.First,the threshold is set strictly.If only one of the cycles of a sequence pattern exceeds the threshold,then the sequence pattern will be considered as not cyclical,even if it has periodic performance in most cases.Second,when using this threshold,we can only judge whether all cycles of the sequential pattern are less than or equal to max Per,but cannot determine whether all cycles fluctuate within a certain range,resulting in the existence of large or small cycles of the excavated sequential pattern,which is unfavorable to decision-making.To solve the above problems,this thesis puts forward a method of stability.Based on this method,a new Mining method namely Stable Periodic High-utility Sequential Pattern Mining(SPHUSPM)algorithm is proposed.SPHUSPM’s contribution is as follows.Firstly,based on the concept of Stable Periodic High-utility Sequential Patterns(SPHUSPs)is first proposed.Secondly,this thesis designs a new data structure PUL-list to record the cycle information of each sequence pattern,so as to improve the mining efficiency.Third,we propose a maximum instability pruning strategy(MLPS)in sequential patterns,which can prune a large number of sequential patterns without stable periods in advance.In order to meet more requirements and improve the accuracy of decision making,this thesis introduces the concept of disutility of items on the basis of SPHUSPM algorithm.The sequential pattern containing disutility items can consider both the events that have occurred and the events that have not occurred.With the introduction of this concept,the sequential model can predict the periodic and efficient relationship between events that have occurred and events that have not occurred,thus improving the diversity of predictions.In addition,we find that in the excavated SPHUSPs,the periodicity of some sequence patterns can be derived from their subsequences,and the decision-making function of this pattern can be replaced by its subsequences.Therefore,a large number of such patterns will cause redundancy,so we need a method to solve the redundancy problem of sequence patterns.Based on the above two problems,a Non-Redundan Periodic High-utility Negative Sequential Pattern Mining(NPHNM)algorithm is proposed.The contributions of the NPHNM algorithm are as follows.First,this chapter introduces the concept of disutility items into the periodic and efficient use of serial pattern mining for the first time,adding the contents of disutility items into the periodic pattern,and the excavated pattern can reveal the periodic relationship between the occurrence and non-occurrence of events.Secondly,this thesis proposes a method to remove redundant periodic patterns in PHUSPM field for the first time,which greatly reduces the number of redundant patterns and makes the decision more accurate and purposeful.