Naive Bayes Expert Options

When the model is built, individual predictor attribute values or value pairs are ignored unless there are enough occurrences of a given value or pair in the training data. The thresholds for ignoring values are specified as fractions based on the number of records in the training data. Adjusting these thresholds may reduce noise and improve the model's ability to generalize to other datasets.

• Singleton Threshold. Specifies the threshold for a given predictor attribute value. The number of occurrences of a given value must equal or exceed the specified fraction or the value is ignored.
• Pairwise Threshold. Specifies the threshold for a given attribute and predictor value pair. The number of occurrences of a given value pair must equal or exceed the specified fraction or the pair is ignored.

Prediction probability. Allows the model to include the probability of a correct prediction for a possible outcome of the target field. To enable this feature, choose Select, click the Specify button, choose one of the possible outcomes, and then click Insert.

Use Prediction Set. Generates a table of all the possible results for all possible outcomes of the target field.