learning parameters - states not appearing

[jaga]

I have followed the tutorial and mapped the nodes of my network to the columns of the data file.
I can see State1, State2 and a set of unassigned states in the Matching window against the column & values for each node that has been matched. I complete the learning process and get a message
Parameter learning is complete – network window will be activated 0:03
Log(p) = -674.177429

When I go back to the network view each node has only the State1 and State2 states. None of the mapped column values are being reflected as states. I tried dragging column values on to states in the matching window but it did not help.
Is there something else I need to do to ensure that all the states in the data file get reflected in the network ?
Does the Log(p) value have any signficance in this context?
Thanks in advance for all help!

[jaga]

I am attaching a zip with the data file (csv txt) and the xdsl for reference.

[shooltz[BayesFusion]]

Parameter learning does not create new outcomes in nodes. You need to ensure that Location node has at least 4 states before running EM, the same applies to nodes like Priority, Module and ResolutionTime. AssignedTo will work with two default states as long as the matched column has only two names.

[jaga]

Thanks for the clarification!

I am able to now create the necessary number of states and then do the mapping from the data file.
In the real life scenario that I am trying to analyze there are 100+ states. So manual creation of these states is not an option. Further when the mapping is done the states retain the default names (State0, State1, etc) and do not pick up the names from the data sheet.
Is there any way out of this?

Rgds,
Jaga

[shooltz[BayesFusion]]

In the real life scenario that I am trying to analyze there are 100+ states. So manual creation of these states is not an option.

For automated solution you'll need to look into SMILE or its Java/.Net wrappers.

Further when the mapping is done the states retain the default names (State0, State1, etc) and do not pick up the names from the data sheet.

That's by design; parameter learning only modifies the network parameters (values in the probability tables).