Learning From Cases Using Excel

1. Create an Excel Spreadsheet

2. Add Nodes to the Net

3. Discretize or Combine States

4. Add Link Structure

5. Learn CPTs

6. Use the Resulting Bayes Net

 

Once the CPTs have been learned, you can use the resulting Bayes net.  If necessary, compile the net to do probabilistic inference by choosing Network Compile, or click the image\LighteningTOOL.gif toolbar button.  You can do what-if analysis by entering findings for some nodes (by clicking on the state names), and observing how all the probabilities change.

If you have a particular new case you wish to analyze, then you can enter everything you know about it as findings, and then observe the resulting probabilities of the target node to have Netica do classification, prediction or diagnosis.

You can open each node’s CPT dialog box to view the conditional probabilities of each node given its parent nodes.

If there is one variable of particular interest (the “target variable”), then you can see how strongly each of the other variables are related to it by selecting it and choosing Network Sensitivity to Findings from the menu.  To understand the report generated, contact Norsys for the Sensitivity document.

You can copy and paste parts of the network into another network to combine knowledge learned from this Excel spreadsheet with the knowledge from the other network (which may have been learned from other data, or created by hand by an expert).  If you do this, you may want to use Modify Reverse Links and delete some extraneous links before you copy and paste, to get the link directions in the right (i.e. causal) direction first.

Prev step <<