Data mining analysis of digital library database usage patterns as a tool facilitating efficient user navigation

World Wide Web usage data mining techniques were used to investigate the feasibility of developing navigation recommendation methods suitable for a large-scale online clinical digital library. Using custom-designed methods, server log data recorded over a nine-month period were cleaned in order to facilitate data mining of characteristic user navigation patterns.; Three main issues were investigated: A method of data cleaning was developed to enable server log records generated by library user navigation to be mined. This method may be of general applicability to many web sites composed of pages containing HTML framesets. Second, criteria to establish “interesting” user-navigation phenomena were translated into appropriate data mining queries. The results of these queries showed that several contexts suitable for navigational recommendations relevant to the user could be found, based on usage within particular medical library instances and across clinical sub-topics. Third, a particular navigational context was explored to examine and develop a feasible, simple outline implementation to deliver sophisticated navigational recommendations based on stepwise sequentially increasing statistical confidence.; In addition, several tactics, for website organization, data cleaning, and for the computing of mined results within a UNIX environment were adduced. User session time threshold settings for server log analyses were also investigated.; Suggestions for the practical application of navigation recommendations were described in the context of proposals for further research concerning the human users' response to the presentation of such information in a library context.; This research represents a detailed case study of the precise requirements for the mining of data from server logs of the subject clinical digital libraries and the utilization of an extension and modification of existing web usage mining methods as developed for e-commerce web site personalization and optimization.; The results of the study indicate that data mining based on aggregate usage of an online clinical digital library environment can be a viable informational enhancement offered to the users of these services.*; *This dissertation includes a CD that is compound (contains both a paper copy and a CD as part of the dissertation). The CD requires the following applications: Windows 95/98; Adobe Acrobat; Microsoft Office; Notepad; Internet Browser; Corel WordPerfect.