Splitting data set using information gain

How find maximunm information gain and using this maximum information gain how can I split each row of my data set.

[KSSV]

Can you elaborate? What is maximum information gain? What exactly you want to do with the data?

[Silpa K]

Or if it possible to do with information entropy for splitting.

clc

clear

b=zeros(36,1);

ts = xlsread('ArrowHead_TRAIN.xlsx');

l=length(ts);

for i = 1:36

p=ts(i,:);

fa = movstd(p,20,1);

secarray=movstd(fa,20,1);

k=maxk(secarray,10);

[~,ii] = min(abs(p(:) - k(:)'));

out = p(unique(ii));

end

Here I have some points in out.I need to find the information entropy based on that I need to spilt each row of my data set and checking each row that contain any of the points in out if any of the number is present I need to add only that splitted series in an excel series.

Answer 1

[Shishir Singhal]

Hi,

I seems like you want to split your data into two sets in the basis of information gain.

Please refer to the file exchange link: https://in.mathworks.com/matlabcentral/fileexchange/14996-entropy

I only need to decide which column of your dataset is represented as a "class" i,e. target variable and which columns are represented as features. Use function "entropyF" to calculate the entropy of each feature variable with respect to "class" variable. Function "getBestEnt" return the index of the feature having highest information gain.

Hope it helps. !!!

Thanks