Web usage mining is a popular research area in data mining. With the extensive use of the Internet, it is essential to learn about the favorite web pages of its users and to cluster web users in order to understand the structural patterns of their usage behavior. In this paper, we propose an efficient approach to determining favorite web pages by generating large web pages, and emerging patterns of generated simple page-linked graphs. We identify the favorite web pages of each user by eliminating noise due to overall popular pages, and by clustering web users according to the generated emerging patterns. Afterwards, we label the clusters by using Term Frequency-Inverse Document Frequency (TF-IDF). In the experiments, we evaluate the parameters used in our proposed approach, discuss the effect of the parameters on generating emerging patterns, and analyze the results from clustering web users. The results of the experiments prove that the exact patterns generated in the emerging-pattern step eliminate the need to consider noise pages, and consequently, this step can improve the efficiency of subsequent mining tasks. Our proposed approach is capable of clustering web users from web log data.

Web log mining is an important area of Web Usage Mining (WUM) for discovering useful information from web log files produced by web servers. According to the mining task, user access patterns can be extracted and the result can be applied to improve websites, business intelligence systems and other areas. In this paper, we propose a novel approach for exacting user access patterns by achieving a large event set and improving the BI-Directional Extension closure checking method (BIDE). The process of getting a large event set can filter out the more frequent events and discard the infrequent events to reduce the number of data. Compared with other methods, the BIDE method with the constraint of gap can achieve a more efficient mining process and generate more accurate results. In our experiment, the effect of the mining task is discussed and the applications of this approach are presented.

The disordered region is an important protein structure which contains much information about protein function. Until now, the prediction of protein disordered region is also much popular task. In this paper, we proposed a new approach to predict protein disordered regions in a protein sequence using the gap-constraint subsequence patterns mining and association rule mining. At first, the gap-constraint frequent subsequences are generated by Gap-BIDE algorithm in two classes, disordered sequence and ordered sequence. Based on these frequent subsequences, we calculated the conditional probability of disordered subsequence patterns in both classes and classify the candidates into the class which has the higher conditional probability. Finally, we used the disordered/ordered subsequence patterns which we generate to search the disordered regions in a protein sequence. In the experiment, we used the CASP 9 and Disprot 5.7 dataset as test data and the performance is higher than other methods.