Select Type of Analysis

This step allows you to select the kind of predictive analysis, and associated parameters.

Select type of analysis

Select the type of analysis to be used to generate the predictive model. The following types of analysis are currently supported:

Model Specifications

The following settings are specific to the type of analysis chosen above.

Number of Clusters (cluster analysis only)
Select either the exact or maximum number of clusters desired. The default value is two clusters exactly. If the maximum number is selected, the resulting model will contain the optimal number of clusters, limited by the specified maximum. For more information, see Cluster Analysis.
K-fold cross validation k value (decision tree analysis only)
In k-fold cross validation, k trees are generated, each based on different sub-samples of the training data (k-1 samples for learning and one sample for testing). After all trees have been analyzed, the optimal tree is selected. The default value is 3. For more information, see Decision tree analysis.
Number of time periods (time series analysis only)
For seasonal training data, this value specifies the number of time periods that constitute a single cycle (for example, 12 months or four quarters). The default value is zero, which indicates that the data is not seasonal in nature. For more information, see Time series analysis.
Number of items, confidence and support (association rules analysis only)
These values place limits on the number of rules that are generated from the training data. The default values for these settings are as follows:
- Maximum number of items per antecedent: 2
- Minimum confidence: 10
- Minimum support: 10
- Maximum consequent support: 100

Click Next to proceed with the Training Metric Wizard.

For more information, see Association rules analysis.