Research And Application Of Time Series Association Rules Based On Fuzzy Set

Time-serial association rules is so effective in the application of the time-varyingfield that we can better describe objective reality according to the regular patternvarying over time with the rules, better recognize the reproducible characteristics ofdynamic systems, timely adjust the strategy of business, predict a particular eventoccurrence in the future, and also make possible as an important research direction indata mining, have important significance in practical applications.For the shortcomings of original time-serial association rules in dealing withhigh-density mass data: spending too much time while processing algorithm, producinga huge frequent candidate set while repeatedly scanning the database, the low efficiencyin the implementation and take full advantage of the framework of FP-Growth in hightemporal and spatial efficiency, time-serial tree optimization algorithm OFP-Growthbased on the redefined support degree metrics method has been proposed. With theeffective frequent item sets generating, frequency vector can be stored and calculated,this algorithm can be greatly improved in the efficiency of execution.For the uncertain and inaccurate situation in solving commodity itemsclassification hierarchical by time-serial association rules, a fuzzy extension ofhierarchical association rules has been introduced, that is based on the fuzzymembership degree of hierarchical classification structure. It overcomes that theoriginal classification structure only considering the semantic difference between thedifferent levels, without further considering the lack of semantic distinction withdifferent membership project. In the process of building a fuzzy hierarchicalclassification structure, through to the form of directed acyclic graph to redefine thedistance between commodity items, item sets and get the finally measurement methodof the distance between association rules. Experiments verify the validity and goodperformance of this method. The similarity between the rules can be obtained by time-serial association rule results of cluster analysis, and can be displayed visually. Thesituation of commodity items classification can be vividly observed by comparing thesimilarity in the class and between the classes, convenient for enterprises managebusiness.