itemset nikolaj tatti main current challeng itemset mine discov small set high qualiti itemset paper propos gener approach measur qualiti itemset method solidli found bayesian statist decreas monoton allow effici discoveri interest itemset measur defin connect statist model collect itemset score individu itemset probabl occur random model built data concret framework exponenti model class model possess desir properti importantli occam razor bayesian model select defenc pattern explos gener exponenti model infeas practic decompos model larg class measur solvabl actual comput score sampl model posterior distribut mcmc approach experiment method demonstr measur work practic result interpret insight itemset synthet real world data 