Post-Processing of Data Mining Results

Abstract

In the information-theoretic network (see Chapter 3 above), each connection between a terminal (unsplit / final layer) node and a target node represents a probabilistic rule between a conjunction of input attribute-values and a target value. An information-theoretic weight of each rule is given by:

$$ w_z^{ij} = P({V_{ij;}}z) \bullet \log \frac{{P({V_{ij}}/z)}}{{P({V_{ij}})}} $$