Pre-order linked WAP-tree mining of sequential patterns

Web usage mining applies data mining techniques to the discovery of usage patterns of web data. Web usage mining mines the secondary data which are recorded users' behavior generally kept in the web log. Web usage mining can be widely used to improve the system and site design, leading to better market decisions. A navigation pattern on the web is considered a sequence of web page accesses. A sequence is an ordered list of events, and sequential mining is used to find the correlation between events. WAP-tree (Web Access Pattern tree) mining is a sequential pattern mining technique for web log access sequences. The WAP-tree technique is based on a prefix tree, which first stores the original web access sequence database, and the frequent sequences are then mined from this tree by recursively re-constructing intermediate trees.; This thesis proposes a WAP-tree based algorithm for finding frequent access sequences, which eliminates the need to reconstruct intermediate trees. In order to avoid reconstructing intermediate WAP-trees, the proposed algorithm builds the frequent header node links of the original tree in a pre-ordered fashion. It also uses position codes to identify the ancestor/descendant relationships between nodes of the tree, and finds common prefix subsequences of mined sequential patterns through a condition prefix sequence search. This results in much better response time as time for reconstructing and traversing several huge trees is saved.