Naive Bayes Classifier Results - Prior Tab

Naive Bayes

Select the Prior tab of the Naive Bayes Classifier Results dialog box to access options to change some of the settings you made on the Options tab of Naive Bayes dialog box and make further predictions using these new settings.

Threshold

When the frequency of a categorical level of an independent variable is zero with respect to one of the categorical levels of a dependent variable, the likelihood of the Naive Bayes model will always yield zero, which is rather superficial. In this field, you can enter a threshold parameter that specifies a lower bound to use when such frequencies are encountered by the Naive Bayes model.

Dependent variable

Naive Bayes supports multiple classification tasks. Use this drop-down list to select the categorical variable for which you want to specify prior values.

Default prior (calculated from data)

Select this option button to assign prior values to the dependent categorical levels on the basis of their relative frequency in the training sample.

Equal (uniform) prior

Select this option button to assign equal priors to the categorical levels. From a Bayesian point of view, this option implies that you have no expectation as to which class a new data case might belong.

Custom prior

Should none of the above prior methods be suitable for your analysis, select this option button to determine custom prior values of your own. For instance, if you have the expectation that a certain class variable is more frequent to occur, you can assign a larger prior value to this particular categorical level. Note that prior values must add to unity since they are probabilities.

Note: Since the posterior distribution is proportional to the product of the prior and the likelihood, prior assignment can strongly influence predictions. Therefore, care should be taken when choosing custom priors for your analysis.

Prior display list. This spreadsheet displays class variables (categorical levels) for the selected dependent variable (see Dependent variable, above) together with their frequencies in the training data and the current value assigned to their prior probabilities.

Descriptive statistics

Click this button to produce spreadsheets of classification summary and confusion matrix.

Predictions

Click this button to create a spreadsheet of the variables and results selected in the Include group box on the Quick tab.