An Exploration of Intentional Behavior in Web Usage Mining

• Main Authors: Yu Min Su, 蘇育民
• Other Authors: Yu Hui Tao
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/46243348505306292654

碩士 === 義守大學 === 資訊工程學系 === 90 === Data mining is a popular research area in recent years, whose goal is to mining useful rules from a large data set. It’s named web mining when data mining is applied over the Internet, and web usage mining, one of its stems, which aims at analyzing user browsing behavior for useful patterns or rules. Currently, the major source for web mining is the web log files, which can’t promote its usability in practical applications despite the progresses or innovations on the algorithmic or programming techniques. Experiences indicate that the innovations in techniques demonstrating far less effects than the variety of the data sources Browsing behaviors outside the log files, such as ‘copy’, ’using scrollbar ’ or ‘save as’, although show no obvious motives or purposes, many useful information is hidden for web mining. We call these “Intentional Behavior”, which reflects a web user’s internal intend on their external behavior and becomes a valuable mining source. Intentional behavior presents the effects of complementary and enhancement in web mining. However, the literature lacks the prerequisite of such applications, an online collection mechanism of browsing behaviors. According to the potential benefits intentional behavior could have brought into web mining research and applications, the study conducts a basic research of browsing behavior, and probes into the potential effect of complementary and enhancement in web mining. Within the basic research, we begin by defining intentional behavior and classifying the browsing behavior, and bringing up an open framework for collecting online browsing behavior as the basis in related applications. Within probing the effects of the web usage mining, we illustrate by examples of WTM Algorithm (Web Transaction Mining) and FWMA (Fuzzy Web Mining Algorithm). The IWTM (Intentional-Based WTM) is further divided into Purchase (IWTMP) and No-Purchase (IWTMNP)：extending from WTM,. IWTMP integrates intentional behavior for explaining the potential benefits of enhancements while IWTMNP addresses what the WTM cant not handle via intentional behavior for its complementary. The intentional-based FWMA (IFWMA) is improving the accuracy issue caused by the unreasonable long browsing time within the algorithm of FWMA while data pre-processing can be used before applying to the algorithm itself. Both the IFWMA and the data pre-processing generate enhancements on the data accuracy. At the end, this research setup a virtual website, simulate and collect browsing behavior from users, and run the four algorithms with these simulated data sets, and discuss their application in commercial applications accordingly.