custom data mining algorithm

I have been using 'A Tutorial for Constructing a Managed Plug-In Algorithm'
to develop my custom algorithm. I have successfully developed the code
needed to train the algorithm but I am little confused on what to do for the
prediction part of the algorithm. The tutorial uses min,max, and variance
for attributes in a cluster to set data distribution. What should I be
setting or programming for prediction to work with my custom algorithm? It
would also be helpful if anyone knows any additional documents that might
help me further understand prediction in data mining algorithms.