To My Parents

Pattern mining plays an essential role in many data mining tasks that attempt to find valuable patterns in massive databases. These patterns can include associations, cor-relations, sequences, episodes, classifiers and/or clusters. However, discovering such patterns is time-consuming because the extracted patterns are often too numerous and thus difficult to analyze by end users, especially when complex data structures are taken into consideration. A further serious problem is that it becomes difficult to gain insight when too many patterns are created. Furthermore, existing work mainly concentrates on discovering common knowledge in a single data set; and not much attention has been given to identifying significant differences among several data sets. This thesis mainly targets four pattern mining topics on complex data structures to provide solutions to the above problems. The first topic is probability-based frequent subtree pattern mining. In recent years, many practical applications dealing with complex data structures, such as chemical compounds, protein structures, XML and Web log databases and social networks, re