Data mining — Confusion Matrix View

In each section of the Confusion Matrix View, the number of predicted classes, the number of correct classifications, and a confusion matrix is displayed.

The Confusion Matrix View provides the following sections:

Confusion matrix for training data
Confusion matrix for validation data
Confusion matrix as computed for the current prune level
This section is only available if you are using tree classification models.

You can display the percentage of the absolute values in the confusion matrix tables by selecting the check box Show in percent.

The following figure shows an example of a confusion matrix:

Figure 1. The Confusion Matrix View of the Classification Visualizer

This graphic shows the Confusion Matrix View of the Classification Visualizer. The section Confusion matrix for training data is expanded.

In the model that is shown in the figure above, the classes No and Yes are predicted. The correct classifications are indicated in bold font type.

You can read the confusion matrix horizontally and vertically.

Reading the table horizontally

There are 124 items classified into class No:

91 of these items are correctly classified into class No.
33 of these items are wrongly classified into class Yes.

There are 116 items in class YES:

12 of these items are wrongly classified into class No.
104 of these items are correctly classified into class Yes.

Reading the table vertically

There are 103 items classified into class No:

91 of these items are correctly classified into class No.
12 of these items are wrongly classified into class No.

There are 137 items classified into class Yes:

33 items are wrongly classified in class Yes.
104 items are correctly classified in class Yes.

The Confusion Matrix is computed by the Classification mining function. It displays the distribution of the records in terms of their actual classes and their predicted classes. This indicates the quality of the current model. A model can contain two or more predicted classes.

Generated models can already contain pruned subtrees. The mining function calculates the confusion matrix and stores this information in the classification model. This confusion matrix is shown in the section Confusion matrix as found in the model.

The tree that is contained in tree classification models might contain nodes that are pruned by the mining function. You can unprune one or more of these nodes, or prune additional nodes in the Tree View or in the Tree Node Distribution View. This changes the tree classification model and the confusion matrix of this model. The confusion matrix in the section Confusion matrix as computed for the current prune level is updated according to the pruning actions that you perform in these views.

If you do not prune the tree in the Tree View or in the Tree Node Distribution View, the values in the tables of the sections Confusion matrix as found in the model and Confusion matrix as computed for the current prune level are the same.

You can copy and paste one or more rows of the confusion matrix table into a different application of your choice. You can also copy the text strings above the confusion matrix tables.

To copy one or more table rows of the confusion matrix, follow these steps:

Click a row in the table, hold down the left mouse button, and drag the cursor to the next table rows.
Select Edit => Copy.
You can also use the keyboard shortcut Ctrl+C.
Paste the copied information into the application of your choice.