Project WUM: The Web Utilization Miner

›  Project Overview ›  Download ›  Selected Publications

»  WUM Tutorial »  WUM FAQs »  Syntax of MINT

Project Overview

WUM is a sequence miner, whose primary purpose is to analyze the navigational behaviour of users visiting a Web site. Nevertheless, it is also appropriate mining sequentials pattern in any type of log file. It discovers patterns that are (i) comprised of not necessarily adjacent events and that (ii) satisfy user-specific criteria of interestingness. WUM's mining language MINT offers predicates to express interestingness criteria based on both statistical dominance or statistical rarity.

The Web Utilization Miner is an integrated environment for log preparation, querying and visualization. A data preparation tool pre-processes the data and organizes the log into sessions, according to user-specified criteria. An "aggregation service" transforms the log of sequences into a tree structure, where sequences with the same prefix are merged. The miner processes this reduced-size structure and applies further heuristics to improve performance. Its visualization mechanism displays the nodes comprising the desired pattern and the different non-frequent paths located in-between. This is quite important when examining how the Web site is really being navigated.

Visit the Myra's home page at Magdeburg University to obtain general information and to download recent publications. Do you prefer watching to reading? Henner Graubitz prepared a Flash movie that gives a glimpse of using WUM.

Download

The latest demo version WUM 7.0 Beta is available for download. Please download one of the following archives:

• WUM.v70.tar.gz [Download, 1489737 Bytes]
• WUM.v70.zip [Download, 1494921 Bytes]

Both archives have the same content. It's strongly recommended to install the miner on a system with at least 128 MB memory. Due to extensive use of the Java Collection API, WUM 7.0 Beta requires Java 2 formally known as JDK 1.2.2 or higher. If you have problems in starting WUM in a Windows environment, please download an updated version of the start script wumgui_v701.bat [Download]. There is a concise description of the installation process and the license in the archive to be downloaded or at our Web site:

• WUM.installation.txt [Download]
• WUM.copyright.txt [Download]

The WUM team would appreciate any bug reports or feature suggestions concerning the future development of the miner. Please drop us an email! You might also be interested in WUMprep, which is a collection of data pre-processing Perl scripts. They are distributed under the LPGL license and can be downloaded from the hypKNOWsys project pages at SourceForge.

Due to the usage of a 3rd party persistence engine, WUM cannot be released as open source software. However, we plan to release an open source version of our Web usage miner in the future. Note that some GNU source code of the future WUM Workbench can already be checked out from the CVS of the hypKNOWsys project at SourceForge.

Selected Publications

• Myra Spiliopoulou and Carsten Pohle. Data Mining for Measuring and Improving the Success of Web Sites. Special issue on applications of data mining to electronic commerce, Journal of Data Mining and Knowledge Discovery, 5(1-2):85-114, January-April 2001. [PDF]

• Myra Spiliopoulou, Lukas C. Faulstich and Karsten Winkler. A Data Miner Analyzing the Navigational Behaviour of Web Users. In: Proceedings of the Workshop on Machine Learning in User Modelling of the ACAI'99 International Conference, Creta, Greece, July 1999. [PDF, PS.GZ]