Abstract: In this paper, a simple and efficient method is presented for quick clustering and trend analyzing of library large log data. First, log data is clustered into a number of subclasses based on the underlying regularity of reader's borrowing and returning books. Then a fuzzy clustering algorithm is given for clustering the subclasses. The time complexity is linear, so our method can scale to large dataset. Finally, regression analysis is performed on the each cluster in order to dis cover the trend of borrowing and returning books. The experiment shows that this approach is successful.

Key words: digital library, clustering, regression analysis, log data