An Affinity Propagation Clustering Algorithm for Mixed Numeric and Categorical Datasets

Algorithm 1

Pseudocode of computing significance.

Set ;

for each numeric attribute in dataset   do

 Figure out the similarity matrix based on (10) as the input;

 Calculate the median of similarities as the shared value of preference;

 Perform the AP algorithm using (1)–(4) to obtain an classification result;

 Discretize attribute to intervals according to the clustering result;

 ;

end for

Establish a new dataset which is a pure categorical dataset composed of the discretized numeric

attributes and the original categorical attributes;

for each attribute in dataset   do

 Calculate the distance between two distinct values of any categorical attribute using (5)–(8);

 Compute the significance (weight) of each numeric attribute using (9) in which the interval is replaced by .

end for