any other plugin algorithm developed?

hi,

as we know we get clustering algorithm with managed plugin algorithm API

does anyone have developed any other plugin algorithm as i want to check what are the things that needs to be modified. i am not data mining algorithm developer but i just want to check where we have to make changes. i would be better if i get source code for algorithm other than clustering

ANOTHER PLUGIN ALGORITHM REQURIED?

thanks in advanced

We haven't made any other samples available as of yet. Angoss and KXEN, among others, both have implemented plug-in algorthms that they sell commercially. I don't know of any parties that have made source code available for their plug-ins - maybe someone will start?|||

is it possible to get detailed class diagram of DMPluginwrapper?

because i want to understand it in detail, as i want to implement framework in c# (totally) but before that i want to understand current architecture in detail.

|||

All the classes in the DMPluginWrapper are fully documented in the CHM help file that is included in the sample. Is there something missing from that?

A diagram of the interactions (in the COM system) for DM plug-in algorithms can be found here: http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnsql90/html/ssdmpia.asp

I think I mentioned before, I am not sure what kind of benefits can be obtained from re-implementing all the framework in C#. The COM interop layer must be there, anyway, and whether it is developed in C# or managed C++, the differences are just syntax elements.

Also, related to your initial question: I have developed a few different (incomplete) algorithms using the framework. Typically, a predictive algorithm (such as a classification or regression solution) requires a subset of the features included in the sample for the clustering algorithm. The cluster specific features are:

- CaseLikelihood and Clustermembership (in the algorithm class)

- the "*Cluster*" functions in the list of standard supported functions of the Metadata class

- the Cluster viewer in the Metadata class

If you remove these, your algorithm is not a clustering algorithm anymore. However, the remaining code should be enough for classification, regression or association types of algorithms. Of course, you will still need to implement your own training and prediction mechanism (specific to your algorithm)

Hope this helps