itemset background knowledg rank itemset decompos famili itemset find good itemset pack data maximum entropi base signific itemset activ topic mine binari data find interest itemset tradit approach search frequent itemset set item occur frequent definit import nice theoret practic properti drawback frequent itemset ab interest individu attribut frequent hand ab infrequ frequent itemset ab interest commonli approach defin import itemset compar frequenc independ assumpt itemset deviat independ assumpt interest altern approach predict frequenc itemset individu attribut expand idea itemset predict compar itemset predict base famili itemset predict maximum entropi popular method estim distribut comparison kullback leibler diverg point rank method normal normal statist test test method real world dataset experi found surprisingli larg portion itemset import independ model unimport larger itemset predict case itemset produc rank interest phenomenon type overlearn occur itemset mislead predict model nois data figur remedi behavior suggest greedi algorithm automat pick lowest rank itemset nr maximum entropi base signific itemset activ topic mine binari data find interest itemset tradit approach search frequent itemset set item occur frequent definit import nice theoret practic properti drawback frequent itemset ab interest individu attribut frequent hand ab infrequ frequent itemset ab interest commonli approach defin import itemset compar frequenc independ assumpt itemset deviat independ assumpt interest altern approach predict frequenc itemset individu attribut expand idea itemset predict compar itemset predict base famili itemset predict maximum entropi popular method estim distribut comparison kullback leibler diverg point rank method normal normal statist test test method real world dataset experi found surprisingli larg portion itemset import independ model unimport larger itemset predict case itemset produc rank interest phenomenon type overlearn occur itemset mislead predict model nois data figur remedi behavior suggest greedi algorithm automat pick lowest rank itemset nr dimens binari data dataset larg number variabl spars depend structur variabl simpler number variabl suggest defin effect dimension dataset nontrivi problem problem defin robust measur dimens dataset show basic idea fractal dimens adapt binari data fractal dimens difficult interpret introduc concept normal fractal dimens dataset normal fractal dimens count number independ column need achiev unnorm fractal dimens normal fractal dimens measur degre depend structur data studi properti normal fractal dimens discuss comput give empir result normal fractal dimens compar pca distanc data set base summari statist concept similar distanc crucial data mine problem defin distanc data set compar summari statist comput data set initi definit distanc base geometr notion set distribut show distanc comput cubic time intuit properti show distanc uniqu mahalanobi distanc satisfi assumpt demonstr deal binari data set distanc repres natur pariti function evalu linear time empir test real world data show distanc work dimens binari data dataset larg number variabl spars depend structur variabl simpler number variabl suggest defin effect dimension dataset nontrivi problem problem defin robust measur dimens dataset show basic idea fractal dimens adapt binari data fractal dimens difficult interpret introduc concept normal fractal dimens dataset normal fractal dimens count number independ column need achiev unnorm fractal dimens normal fractal dimens measur degre depend structur data studi properti normal fractal dimens discuss comput give empir result normal fractal dimens compar pca safe project binari data set activ topic mine binari data find interest itemset tradit approach search frequent itemset set item occur frequent definit import nice theoret practic properti drawback frequent itemset ab interest individu attribut frequent hand ab infrequ frequent itemset ab interest commonli approach defin import itemset compar frequenc independ assumpt itemset deviat independ assumpt interest altern approach predict frequenc itemset individu attribut expand idea itemset predict compar itemset predict base famili itemset predict maximum entropi popular method estim distribut comparison kullback leibler diverg point rank method normal normal statist test test method real world dataset experi found surprisingli larg portion itemset import independ model unimport larger itemset predict case itemset produc rank interest phenomenon type overlearn occur itemset mislead predict model nois data figur remedi behavior suggest greedi algorithm automat pick lowest rank itemset nr 