3.1.2 Construction of WAP-Tree and Mining of WAP-Tree
Only frequent 1-sequences are considered for constructing WAP-tree, as
they only are useful in generating k-sequences where k>1.
Common prefixes are shared in the WAP-tree to save space. WAP-tree
records, labels and counts two parts of data. The tree’s root is a
virtual special node with an empty label and count 0.
The construction of WAP-tree is done as follows: for each series of
accesses, frequent subsequence is entered into the tree, starting from
the root node. While inserting the first event e, if the current node
has already a child e, then Increase the child node count with event e
by 1, otherwise a new child with label e and count 1 will be created.
Then insert the remaining subsequence into the sub-tree rooted in the
present e node recursively. All nodes with same label e are linked by
shared label linkages into a queue, called event-queue. Event queue for\(e_{i}\)is called \(e_{i}\)-queue. Head of each event-queue is
registered in a header table H.
A simple web access sequence database obtained after preprocessing, with
the set of access events E = {a, b, c, d, e, f} is shown in Table 1
and WAP-tree with linkage header for the WASD is given in Fig. 5.