Date Added: Feb 2010
There are many existing systems which can be used to analyze user behavior, but only some of them pay attention to the change of user's behaviors in time-periods. This paper presents a new algorithm for mining frequent sequences named Weighted Time GSP (WTGSP), which uses earlier information to cut down the cost of finding sequential patterns. Generally, WTGSP increases effects of time in GSP algorithm to extract user's patterns and find patterns that are near to user's current behavior. Performance of the algorithm is examined with the log data gathered from a news website. Experimental results and performance studies demonstrate that this algorithm is efficient on real user behaviors log datasets and it can predict user's current behaviors as well.