Table 1
Figure 5 .
All pattern information related to a frequent event \(e_{i}\)can be
accessed by following all the branches in WAP-tree linked by\(e_{\text{\ i}}\)queue only once. All nodes from the root of the tree
to the e i(excluded) node form a prefix series of e iand this node count
is the prefix sequence count. Let G and H be two prefix sequences of\(e_{i}\)and G is also formed by the sub-path from root that H is formed
by, H is called a super- prefix sequence of G, and G is a sub-prefix
sequence of H. The un-subsumed count is the count for sequence of\(e_{i}\) prefixes without any super-prefix sequences. For a prefix
sequence of \(e_{i}\)with some super-prefix sequences, the un-subsumed
count is the count of that sequence minus un-subsumed counts of all its
super-prefix sequences. Access patterns with same suffix are now used
for searching all web access patterns.
Mining is done as follows: For each event \(e_{i}\)in the header list,
conditional sequence base is found. A suffix event’s conditional
sequence list is obtained by following the event’s header connection and
reading the root route to each node (excluding the node). The prefix
sequence count is identical to the node count. For each prefix sequence
inserted into the conditional sequence base with count c, all its
sub-prefix sequences are inserted with –c as count, to get the
un-subsumed count. Then find the set of conditional frequency
occurrences. If it is not empty, recursive mining is done on the
conditional WAP-tree of each frequent event. When there is only one
branch in the conditional WAP-tree, all unique combinations of nodes in
that branch are generated as patterns.