Research On Analysis Methods Of Negative Sequential Patterns And Rules Based On Influence

Negative sequential pattern(NSP)mining can capture frequently occurring and non-occurring behavior information,and plays an irreplaceable role in education,medical,biological and other fields.Most traditional NSP mining algorithms adopt a support measure to discover interesting patterns.However,the support measure does not truly reflect the interestingness of patterns in some cases.In particular,it ignores the effect of the support of each element and the order characteristics among these elements.Hence,an influence measure was proposed to truly reflect the interestingness of patterns.However,its candidate sequence generation and calculation method is only used for positive sequential pattern(PSP)mining and cannot be applied to NSPs.In addition,the traditional negative sequential rule(NSR)mining algorithm aims to discover all possible rules and cannot meet the needs of users who are interested in specific target(outcome)rules.Impact-oriented NSR mining algorithms are used to mine specific target rules.However,existing research is limited to mining NSRs from PSPs without considering non-occurring elements or non-occurring items and ignoring rules from NSPs.To address the above problems,this thesis takes the negative sequences with non-occurring items as the research object,and studies the NSP mining method based on influence and the impact-oriented NSR analysis method.The main innovative work of this thesis is as follows:(1)An influence-based and interesting NSPs mining algorithm Inf I-NSP is proposed.Firstly,the constraint of sc-NSP(Better Structure and Constraint to Mine Negative Sequential Patterns,sc-NSP)is relaxed to enable sc-NSP to mine NSPs of non-occurring items.Secondly,based on the candidate sequence generation strategy in modified sc-NSP,a new influence calculation method(Inf-Calculation)is proposed,which could calculate the influence of negative candidates quickly through a new pattern processing method.Finally,an algorithm named Inf I-NSP is proposed to mine interesting NSPs based on influence.Experiments on real-life and synthetic datasets show that Inf I-NSP can remove valueless patterns and discover truly interesting NSPs in short time.(2)An impact-oriented NSR analysis method Im NSRule is proposed.Firstly,a generation method of impact-oriented negative candidate sequential rule is proposed and combined with Inf I-NSP to mining impact-oriented NSRs from NSPs.Thus,the problem of ignoring non-occurring elements or non-occurring items is solved.Secondly,the method is extended on the basis of the existing methods for calculating the impact of non-occurring items in rule antecedent on the outcome in rule postecedent.Thus,the impact of non-occurring items on the outcome is measured from multiple perspectives.Finally,a large number of experimental results show that the Im NSRule algorithm can mine impact-oriented interesting NSRs in short time and can effectively calculate the impact of non-occurring item(s)in rule antecedent on outcomes in rule postecedent.Through the application of Im NSRule to students’ campus sequence data,it is found that the impact of students’ breakfast meals at different times on their academic performance is different,among which the impact of breakfast meals on academic performance is greater on Sundays and during examination weeks.