1. Create an Excel Spreadsheet
3. Discretize or Combine States
6. Use the Resulting Bayes Net
You need to discretize continuous nodes, and if they have already been discretized, you may want to adjust the thresholds to better capture the Excel data. If the nodes were added automatically, then you may want to convert discrete numeric nodes to continuous ones, and then discretize them.
Automatically: To use histogram information generated from the Excel dataset, select one or more nodes and choose Modify → Discretize Node from the menu. Any of the nodes that are discrete will be converted to continuous. Netica may ask you what file you wish to use to generate the histogram, to which you would normally indicate the Excel spreadsheet file. Then Netica will ask how many states (i.e. bins) you want each node to have, and the degree of rounding (20% is usually a good amount). Finally, it will automatically do the discretization so that the bins have approximately equal amounts of data samples. If you specified 0 % for the rounding, then each bin will have as close as possible to the same number of samples, and if you chose a larger number, the bin amounts won’t be exactly equal since Netica tries to choose somewhat round numbers for the thresholds.
Manually: Alternatively, you can manually change discrete nodes to continuous with the node properties dialog, and for continuous nodes, you can discretize them or adjust their discretization thresholds with the Node Discretization setter.
Combine States: If the entries in a column are not numeric, and there are some that you don’t care about, you can delete the node’s states which correspond to them, and then add a state called other (case sensitive).
Home > Learning From Cases > Learning From an Excel File > Discretize or Combine States