Creator:Ivana ?a?e
Function:Describe Naive Bayesian network of variables (classifaction)
Input Type:csv, discrete
Input from Agent:Dizzy, Data Selection Agent
Output Type:Classification advice
Output to Agent:Advice, Ceres
Short Description:

Nabby is a Naive Bayes classifier. Like Rikku and Moku, it classifies a new case using the values of input variables, and it also returns the probability of the classification. Given a patient database,  for example, Nabby can learn to calculate the probability for each diagnosis given a patient's symptoms. 

This type of classifier is relatively simple and belongs to the group of linear classifiers. This makes it possible to show the user how each input variable value contributed to the final classification.
Despite its simplicity and the underlying assumption that all variables are independent (given the classification) the Naive Bayes classifier has been shown to perform well in many different application domains. Even when the variables are not independent, the classifier remains robust. It will still find the most likely class, but the returned probability of the classification will be less accurate.

Instead of learning from data it is also possible to initialize Nabby with expert knowledge. Nabby can handle missing values.

© 2015 Alan Turing Institute Almere