Topics related to Bayes Nets
There are some topics related to Bayes Nets, such as Data mining, Machine learning, and more.
Data mining
Elder Research Data Mining Consulting
Bing Liu (Liu,
Bing)'s Home Page
Machine Learning
Neural networks, neuroscience and
relevant mathematics - Wlodek Duch
links
Pattern Recognition Information
RBP Network & RULEX Rule Extraction Downloads
Correlation for continuous variables
In probability theory and
statistics, correlation, also called correlation coefficient, indicates the
strength and direction of a linear relationship between two random variables.
The correlation coefficient 
between two
random variables X and Y with expected values 
and 
and standard
deviations 
and 
is defined as:
where 
is the function
for the expected value and 
is the function for
covariance value. The maximum of the absolute correlation coefficient value is
1. If the correlation coefficient is +1, it means that two variables change
linearly in the same directions, and if the correlation coefficient is -1, it
means that two variables change linearly in the opposite directions.
When two variables are independent, their correlation should be 0. However, when the correlation is 0, it does not mean that two variables are independent, since correlation only measures the linear dependency between two variables.
Chi-square test for discrete variables
The chi-square value between two variables is defined as
Where 
is the number of
values of variable 1, 
is the number of values
of variable 2, 
is the number of
instances with i-th value for variable 1 and j-th value for variable 2, 
is the number of
instances with i-th value for variable 1, 
is the number of instances
with j-th value for variable 2, 
is the number of
the total instances, and 
is the expected
frequency of. The chi-square value measures the difference of the expected
frequencies and the actual frequencies in different categories.
Mutual information for discrete variables
Mutual Information (MI) is an entropy-based measure of the dependency between two variables. It is the difference between the prior entropy of variable C and the posterior entropy of variable C given values of another variable F:
Back to home last updated: 09/18/2007 © Copyright 2007