Mining frequent sequential patterns and association rules on campus map system

Search log provides valuable insight into user behavior and interest. By analyzing search log, website manager could have a better understanding on user behavior and interest, and the system design and user experience could be improved. This paper analyzes campus map search log in Tsinghua University on term level, query level and session level. To find out interests of users, generalized sequential pattern algorithm is adopted to mine the frequent sequences of terms. From the frequent sequences, suggestions for improving the design of the campus map system is offered. On query level, the trend of query amount varies with the occurrence of real world event is analyzed. On session level, the search log is divided into sessions using sliding window method. Most users stop searching in 290.87 seconds with no more than 5 queries. The correlations between frequent sequences and locations are indicated using association rules. With the analysis of campus map search log, system performance and user experience is improved.