How many Bayes categories should I use?

Canvas can generate a naive Bayes classification model using structural fingerprints or properties as input. The dependent variable represents a class or category.

Bayes classification is most effective when the chosen number of categories does not require the accuracy of the predictions to approach the experimental uncertainty in the data being fit. For example, if the dependent variable is an activity, such as pKi or pIC50, defining more than 4 categories is not recommended.

Back to Search Results