Một số cải tiến thuật toán Index-BitTableFI cho khai thác tập tin phổ biến

Abstract

Index-BitTableFI is an algorithm based on BitTable which is very effective in recent (Song& Yang, 2008). It finds out itemsets based on BitTable in vertical and horizontal, and also sets up sorting array and  equivalent  computing  method  to  fast  identify itemsets which occur concurrently with representative items.  Although  Index-BitTableFI  algorithm  reduces considerablely  cost  of  finding  out  candidate  itemsets and  computing  the  support,  but  if  number  of transactions  and  items  is  large  then  intersection computing  of  vector-bits  in  BitTable  still  costs  time. Besides, finding out frequent itemsets in depth hasnot used property of equivalent computing method yet. To resolve  this  problem,  some  improvements  for improving  more  performance  of  Index-BitTableFI algorithm are proposed in this research.