I wanted to share with everyone a research paper I wrote titled Applying Data Mining Techniques to the Dell Kace K1000 Systems Management Appliance in order to Identify Elusive Computer Objects. I wrote this paper after preliminary findings indicated that data mining could be applied to one's Dell Kace K1000 to identify objects that are similar to other objects, but fall outside of the classical smart labeling approach for some reason or another. The specific technique I used falls into the realm of supervised learning, which basically means a subject matter expert is required to seed the data mining algorithm by pre-classifying object association. My research concludes that indeed data mining is possible within the appliance, highlights some potential implications of having a data mining algorithm in place, and argues for further research in this area.
Specifically, I am going to work towards implementing an approach that uses unsupervised learning. I am envisioning, when run against one's K1000 DB (like the Twitter Bootstrap for Kace Service Desk companion project (shameless plug http://www.itninja.com/blog/view/twitter-bootstrap-for-kace-service-desk)), the algorithm would allow one to use dynamic object classification that evolves as one's environment does. For example, you could say, install this application on machines that are like a prototypical object and then adjust the likeness value. Of course this is hypothetical at this point as I haven't implemented the model yet.
Here is a link to the research paper that highlights the supervised learning approach. (feedback is welcome)
That being said, I only have access to my data model and would need access to other customer's DBs to test out my algorithm; I know, no easy feat. If you are interested in collaborating, please feel free to contact me directly. Thanks, Jay