Predicting User Behaviour to Facilitate Efficient Provision of Health Applications

Practical analysis of user behavior patterns in social network and community software is one of the many application of data mining tools. Using data mining techniques, over 100 users of the Vividesk private online network (´desktop´) for Canadian healthcare professionals was examined over a four year period for user behavior trends using decision trees (DTs) mining. Vividesk provides users with an online community of research and of practice, enabling clients to use Web 2.0 and social networking principles to enhance their medical research and practice. Our interest rests primarily in usage patterns related to usergroups and classes of applications on these desktops. Some applications link to licensed information resources which can cost thousands of dollars per year to license. As a result, examining application usage data can help generate information on the relative cost effectiveness of the resources for which clients pay. This study presents an initial analysis of data by grouped resource applications as a means of determining whether deeper analysis is warranted. Data was warehoused and mined using a Microsoft Business intelligence tool (cube), using predetermined dimensions. By warehousing and mining desktop application usage through DTs, previously hidden usage patterns were uncovered. The DT experiments revealed various application usage patterns both within and between the desktops. Usergroups within each environment also demonstrated different access patterns at different times. Based on this analysis, further drilling down is warranted to uncover patterns of particular resource use. Exercises such as these will help predict future user behavior and facilitate the planning of desktop resource provision.