Semi-naive Bayes Classifiers

Semi-naive Bayes classifierIt is a classification method that takes into account the interdependence between some attributes. It is a relaxation strategy when the mutual independence of the features of the naive Bayes classifier is difficult to satisfy.

The most commonly used strategy of the semi-naive Bayes classifier is to assume that each attribute depends on at most one other attribute, and the attribute it depends on is called its super-parent attribute. This relationship is called: unique dependency estimate (ODE).

Changes in mathematical form

The sample prediction probability of Naive Bayes is:

The sample prediction probability of semi-naive Bayes is:

It can be seen that the class conditional probability P(x_i | c) Modified to x_i Depends on a category c and a dependency property pa_i .

Related words: Naive Bayes classifier

Sub-words: unique dependency estimation

References

【1】https://blog.csdn.net/xo3ylAF9kGs/article/details/78643424

【2】https://github.com/familyld/Machine_Learning/blob/master/07Bayes_classifier.md