Mining Temporal Web Interesting Patterns

Previous work on mining web associations focus primar- ily on finding frequent access patterns in the data. However, they ignore an important relationship that web frequent ac- cess patterns have the dynamic characteristic of time vary- ing. It is also important that in database, some items which are infrequent in whole dataset but those depend on the present of a mediator itemset may be frequent in a particu- lar time period, which induce some interesting patterns may not be discover. In this study, our focus is to apply a new mining technique called indirect association onto tempo- ral web data and propose the TIFP-mine algorithm based on a new model WM-graph, which are both capable of ex- tracting all temporal indirect frequent patterns and its tem- poral extended patterns. Experimental results confirm that TIFP-mine algorithm is efficient and effective. Our analysis shows very promising results, especially in terms of identi- fying Web users with distinct interests.