Research On Key Technologies Of Utility-based Negative Sequential Rules Mining

Utility based sequential rule mining can mine sequential rules with high utility value,and is widely used in finance,biomedicine,manufacturing,e-commerce,social media and other fields.Compared with high utility sequential rule mining,high utility negative sequential rule mining considering the non-occurring event is able to give more complete information.At present,high utility sequential rule mining cannot be directly applied to high utility negative sequential rule mining,because there are many inherent complexity problems in the process of high utility negative sequential rule mining :(1)How to define high utility negative sequential rule mining.(2)How to calculate the local utility value and the utility of the antecedent in a high utility negative sequential rule,which is the key step to calculate the utility confidence.(3)How to mine actionable high utility negative sequential rule that can be used directly for decision-making.The study of these problems has important theoretical value and practical significance for mining more comprehensive and valuable high utility negative sequential rules.Therefore,this paper focuses on mining high utility negative sequential rules from high utility negative sequential patterns,exploring the mining methods of actionable high utility negative sequential rules,and indepth discussion on the key problems involved.The details are as follows:In view of the first two problems,an algorithm named e-HUNSR is proposed by this paper in order to mine high utility negative sequential rules.Firstly,the high utility negative sequential rule problem is formalized by proposing the concept of local utility value and utility confidence.Then a method of generating high utility negative sequential rule candidate and a pruning strategy are given.Then,a data structure is designed to store necessary information,and a simplified calculation method for calculating local utility value and utility value of the antecedent is proposed.The experimental results show that e-HUNSR algorithm can effectively extract the high utility negative sequential rule from high utility negative sequential patterns.In view of the third problem,this paper proposes an actionable high utility negative sequential rule mining algorithm A-HUNSR.Firstly,the paper explores pruning mechanisms of actionable high utility negative sequential rule mining.The first pruning is completed by judging the support of antecedent,consequent and the high utility negative sequential patterns composed of antecedent and consequent.The second pruning is completed by determining whether the correlation of the high utility negative sequential rule candidate is greater than 1.Secondly,the calculation method of support in utility environment and the calculation method of correlation of high utility negative sequential rule candidate in utility environment are proposed.Experiments show that A-HUNSR can prune a large number of rules effectively.