More recent approaches to web log data representation aim to capture the user navigational patterns with respect to the overall structure of the web site. One such representation is tree-structured log files which is the focus of this work. Most existing methods for analyzing such data are based on the use of frequent sub tree mining techniques to extract frequent user activity and navigational paths. In this paper we evaluate the use of other standard data mining techniques enabled by a recently proposed structure preserving flat data representation for tree-structured data. The initially proposed framework was adjusted to better suit the web log mining task. Experimental evaluation is performed on two real world web log datasets and comparisons are made with an existing state-of-the-art classifier for tree-structured data. The results show the great potential of the method in enabling the application of a wider range of data mining/analysis techniques to tree-structured web log data.