readme.bak

charm是基于垂直数据集挖掘关联规则的一个著名算法

how to run charm
----------------
input file must have XXX.data extension

        charm -i XXX -h -d -l -s <MINSUP>

        other flags
         -o output the patterns found
                output format: itemset - sup (tidset)
         -H 1 for exact closed sets (takes longer time)
                the default is to mine a superset of the closed sets
                the use of -H 1 eliminates any non-closed sets using a
                    hashing technique

        NOTE:
         -d uses diffsets instead of tidsets (from length 3 onwards)

         -l uses diffsets for pass 2 as well
                (this should NOT be used for sparse datasets, since tidset
                 size of pass 2 is smaller than diffset size for
                 sparse sets.)
         
          -h option converts from a horizontal DB XXX.data to an
             in-memory vertical DB. Thus this version should only be
             used with small DB that can fit in memory. IF you need to
             run large DB then I have other scripts that first create
             a disk-based vertical DB and then charm runs on that
             DB. I think the current version is sufficient for most
             experiments, but if you will do performance tests for
             large DB then please ask for the other scripts.


MINSUP is in fractions, i.e., specify 0.5 if you want 50% minsup or
0.01 if you want 1% support.

the input database must be in binary, with the following format per row:
        CID TID #ITEMS LIST_OF_ITEMS
e.g.    1   1   4       0 1 4 6
        2   2   3       4 7 9
items in the list must be sorted in increasing order

Finally the summary of the run is stored in the summary.out
file. The format of this file is as follows:

CHARM (other options) DB_FILENAME MINSUP NUMTRANS_IN_DB ACTUAL_SUPPORT
      [ ITER_i |Ci| |Fi| timeForIter_i avg_tidset/diffset_size ] 
      [TOT total_cands tot_freq tot_elapsed_time] 
      NumberofIntersections XXX XXX XXX XXX 
      tottime maxiters user_time sys_time