Dynamic Itemset Counting and Implication Rules for Market Basket Data-英文文献.pdf

Dynamic Items et Counting and Implication Rule s for Market Basket Data Sergey Br in Ra jeev Motwani Jerey D Ullman Shalom Tsur Department of Computer Science RD Division Hitachi Amer ica Ltd Stanford University tsurhitachi com fs ergeyra jeevullman gcsstanforde du Abstract of web page s and many more We applie d marketbasket analys i s to census data s ee s ection We cons ider the problem of analyzing marketbasket data In thi s pap er we addre ss b oth p erformance and func and pre s ent s everal imp ortant contr ibutions First we pre s ent tionality i ssue s of marketbasket analys i s We improve p er a new algor ithm for nding large items ets which us e s fewer formance over past metho ds by intro ducing a new algor ithm pass e s over the data than class ic algor ithms and yet us e s for nding large items ets an imp ortant subproblem We fewer candidate items ets than metho ds bas e d on sampling enhance functionali ty by intro ducing implication rules as an We inve stigate the idea of item reorder ing which can im alter native to asso ciation rule s s ee b elow prove the lowlevel eciency of the algor ithm Second we One very common formalization of thi s problem i s nd pre s ent a new way of generating implication rule s which ing association rules which are bas e d on support and con are normalize d bas e d on b oth the antece d