Unique Dependence Estimation(ODE) is the most commonly used strategy for semi-naive Bayes classifiers. The so-called unique dependency is to assume that each attribute depends on at most one other attribute outside the category.

Naive Bayes class conditional probability:

Class conditional probability under independent dependency estimation:

Among them,_i For attribute x_i The attribute on which it depends is called x_i For each attribute, if its parent attribute is known, a mathematical method can be used to estimate the probability value P ( x_i | c, pa_i ) .

Classification of unique dependency estimates

There are three main methods for implementing independent dependency estimation:

1) SPODE (Super-parent Dependency Estimation), assumes that all attributes depend on the same attribute, called the "super-parent", and then determines the super-parent attribute through model methods such as cross-validation.

2) TAN (Tree Augmented naive Bayes), builds dependencies based on the maximum weighted spanning tree algorithm.

3) AODE (Average Independent Dependence Estimation), attempts to construct SPOE with each attribute as a super-parent, integrates the results, and is similar to Naive Bayes, without the need for model selection, and counts the samples that meet the conditions.

Related terms: semi-naive Bayes classifier, super-parent attribute.