# Subtractive Clustering

7 views (last 30 days)
zheng on 30 Mar 2011
Hi,all I hope someone can answer questions about the subtractive clustering which is provided by matlab.
If my data is in following format ========================================================================================= x(m) y(m) accuracy speed direction id year time timestamp ========================================================================================= I did three tests to use subclust find clusters by using different parameters of radii and options. unfortunately, it seems like that those corresponding information for each point (x and y) cannot be involved. In addition, after calculation, it is found that S can give the range of influence of cluster center in meters. It gives me two constant values for each cluster center, one for x and one for y. Am I correct about this?
second, what I can do if I want that the output cluster centers also have corresponding information, such as accuracy, speed, direction, etc. Since subtractive clustering never computes new points except treat every point as a potential cluster center based on the density of its neighborhood, I will use the computed S for x and y (called as Sx and Sy) to find MAX/MIN of x and y in each cluster. For example, [C,S] gives a number of cluster centers, and first cluster is used here. MAX_x in cluster 1 = C_X1+(Sx/2); MIN_x in cluster 1 = C_X1+(Sx/2); MAX_y in cluster 1 = C_y1+(Sy/2); MIN_y in cluster 1 = C_y1+(Sy/2); Through comparing input data points with above max/min values, I can decide which and how many points are belong to cluster 1. Does it make sense?