correl hide independ mask explor diseas associ nhane data data mine survei sequenc classif earli predict time seri nearest neighbor approach paper formul problem earli classif time seri data import time sensit applic healthinformat introduc concept mpl minimum predict length develop ect earli classif time seri effect nearest neighbor classif method ect make earli predict time retain accuraci compar nn classifi full length time seri empir studi benchmark time seri data set show ect work real data set nn classif effect mine sequenc classifi earli predict supervis learn sequenc data sequenc classif recogn import data mine task signific applic tempor order import sequenc data critic applic sequenc classif medic diagnosi disast predict earli predict highli desir featur sequenc classifi earli predict sequenc classifi prefix sequenc short make accur predict knowledg earli predict sequenc data studi systemat paper identifi problem mine sequenc classifi earli predict analyz problem challeng attempt tackl problem propos interest method sequenti classif rule scr method mine set sequenti classif rule classifi call earli predict util defin select featur rule gener sequenti decis tree gsdt method adopt divid conquer strategi gener classif model conduct extens empir evalu real data set interestingli method achiev accuraci compar stateof art method typic short prefix sequenc result earli predict highli feasibl effect mine sequenc classifi earli predict supervis learn sequenc data sequenc classif recogn import data mine task signific applic tempor order import sequenc data critic applic sequenc classif medic diagnosi disast predict earli predict highli desir featur sequenc classifi earli predict sequenc classifi prefix sequenc short make accur predict knowledg earli predict sequenc data studi systemat paper identifi problem mine sequenc classifi earli predict analyz problem challeng attempt tackl problem propos interest method sequenti classif rule scr method mine set sequenti classif rule classifi call earli predict util defin select featur rule gener sequenti decis tree gsdt method adopt divid conquer strategi gener classif model conduct extens empir evalu real data set interestingli method achiev accuraci compar stateof art method typic short prefix sequenc result earli predict highli feasibl effect model multipl time unit delai gene regulatori network dynam bayesian network 