Modeling
Modeling Overview
The modeling step of the Rapid Process Troubleshooting project allows for the creating of various process models using the data loaded and prepared in the preceding steps. Four model types can be constructed from the data:
Accessing the Modeling view
You can access the modeling view from the troubleshooting Project Bar located to the left of the Discrete & Batch Troubleshooter project view. In order to be able to access the modeling view, the minimum requirements must be met: a loaded, valid and properly configured data source.
If the modeling view is not available (grayed out), ensure that you have a properly configured data source in step 1 (Data Preparation) and the Visualization view has been opened at least once.
The Modeling view
The modeling view has the following options for each of the modeling techniques provided:
-
Construct
Constructs a new model of the selected type. This option will not be available if a model already exists. To build a new model of the selected type, first remove the existing model by clicking the remove option (see below). -
View
Views the currently active process model of the selected type. This option will not be available if no model exists. -
Remove
Removes the currently active process model of the selected type. This option will not be available if not model exists.
PCA attributes
Once a PCA model has been constructed, the modeling view lists the following attributes for the created PCA model:
-
Number of principle components
The number of principle components that were used in PCA model. -
Data normalised
Set to "Yes" or "No" depending on wether the input data was normalised. -
Requested confidence
The percentage confidence requested before model construction. -
Actual confidence
The Actual confidence achieved after model construction. -
Data selection
Set to the data selection that was made: All data, Inside History Brushing or Outside History Brushing.
PLS attributes
Once a PLS model has been constructed, the modeling view lists the following attributes for the created PLS model:
-
Number of model components
The number of model components used in the PLS model. -
Data normalised
Set to "Yes" or "No" depending on wether the input data was normalised. -
Requested X-variance contribution
The requested X-variance contribution before PLS model construction. -
Actual X-variance contribution
The actual X-variance contribution achieved after PLS model construction. -
Actual Y-variance contribution
The actual Y-variance contribution achieved after PLS model construction.
-
Data selection
Set to the data selection that was made: All data, Inside History Brushing or Outside History Brushing.
Decision Tree attributes
Once a Decision Tree has been constructed, the modeling view lists the following attributes:
-
Number of Rules generated
The number of rules generated by the decision tree which correlates to the number of child nodes with no child nodes itself. -
Construction cases
The number of data rows used for the construction of the Decision Tree model. -
Validation cases
The number of data rows used for the validation of the Decision Tree model. -
Bad quality cases
The number of data rows that contained bad quality data. -
Target field
The field selected and the target. -
Input fields
The list of input fields that were used.
Classification model matrix
Once a Classification model has been constructed, the modeling view lists the Classification model matrix.
The model target classes are listed in the first column, and the percentage of correct classifications is listed in the rows. In a prefect classification model, there should be a diagonal line of 100% for each class to be classified correctly.
Related Topics: