Figure 2
From: BicSPAM: flexible biclustering using sequential patterns
Mining order-preserving biclusters in real or itemizespaced matrices. To discover order-preserving biclusters the first step is to order the column indexes according to real or discretized values and map them to itemset sequences based on the observed ordering (precedences and co-occurrences). In particular, when targeting the or behavior, co-occurrences are propagated n times, being n the number of items co-occuring. Illustrating, x2= {y1=0, y2=2, y3= 0} is mapped as (y1y3)y2 sequence according to the xor behavior and as (y1y3)(y1y3)y2 under the or behavior. Second, a SPM method is applied over the set of sequences to extract the set of sequential patterns. Finally, biclusters are derived from the set of items and supporting transactions for each sequential pattern.