| ***** Some mathematical symbols involved in this paper ***** |
| % : stock trading data matrix after preprocessing. |
| % : the number of columns in . |
| % : the row threshold of bicluster. |
| % : the column threshold of bicluster. |
| % BS: the set of bicluster seeds. |
| % BIC_Set: the set of biclusters. |
| Input |
| Output BIC_Set |
| Let BS = and BIC_Set = |
| Description: |
| (1) ********* Obtain all biclusters with one column ********* |
| (2) for |
| (3) Apply the -means algorithm for the th column. |
| (4) Then we obtain a set of clusters in the column, denoted as where . |
| (5) % Update BIC_Set. |
| (6) if the number of rows in |
| (7) Add it into BIC_Set. |
| (8) end if |
| (9) end for |
| (10) BS = BIC_Set. |
| (11) ******************* Expand BS ****************** |
| (12) % The following steps are implemented in BS. |
| (13) % Merge all the biclusters of one column to get the biclusters of two columns. |
| (14) for |
| (15) for |
| (16) for |
| (17) is jointed with . % that is, intersect the row sets and join the column sets. |
| (18) Update BIC_Set similar to the process mentioned above. |
| (19) end for |
| (20) end for |
| (21) end for |
| (22) The newly obtained biclusters of two columns which satisfy the row threshold are put into BS. |
| (23) % Continue merging to get biclusters of more columns. |
| (24) while and the number of columns in biclusters is less than |
| (25) Merge every two length- biclusters (biclusters whose number of columns are ) in BS, |
| (26) to get length- biclusters whose number of columns are . |
| (27) Add the biclusters that meet the threshold into BIC_Set. |
| (28) Put the satisfying length- biclusters into BS |
| (29) end while |
| (30) ******************** Output BIC_Set ******************** |
| (31) Filter out the duplicate biclusters and delete the biclusters whose column number is less than . |
| (32) Output BIC_Set |
