An analytical approach to deriving usage patterns in a Web-based information system

With the growing popularity of the World Wide Web, Web-based information systems have become one of the primary means for people to access information. Though usage patterns in traditional information systems such as public access library catalogs (OPACs) have been studied for more than three decades, investigations of similar information systems in a hypertext environment remain scarce. This study presents an analytical approach to deriving usage patterns in a Web-based information system.; First, system users are divided into groups with similar use of the system by employing multivariate statistical analysis techniques. Second, a continuous-time stochastic model (a semi-Markov chain model) is developed for each user group. The transition rates as well as transition probabilities of the Markov model (called intrasession usage patterns) are used to characterize user behavior probabilistically. Third, a generic algorithm (called GREEDY) is developed to discover both time-invariant and time-dependent sequential usage patterns (called intersession usage patterns) that are common to the members of a group. The intersession usage patterns provide a causal interpretation (cause and effect) of user behavior from a logical/timing perspective.; The proposed methodology was demonstrated and tested for validity using two independent samples of user sessions drawn from the transaction logs of the University of California's MELVYLRTM online library catalog system (www.melvyl.ucop.edu). The results indicate that there are five nonsearcher groups and six searcher groups in the MELVYL system. The majority of user groups have 3rd-order sequential dependency in transitions, and the remaining user groups follow 4th-order sequential dependency in transitions. User session length (duration of stay) can be approximated by a lognormal distribution. The differences in derived usage patterns between user groups were tested statistically. The test results show that users of different groups have distinct patterns of use of the system, which justifies the methodology employed in this study.; The acquired knowledge of usage patterns can aid the design of an advanced online help system that provides situational learning and customized help, depending on the context the user is in. This study provides a background for further analysis of user behavior on the Web, which has been recognized as the key to the success of electronic commerce.