information gain

Information gain is an information measure used in decision tree building algorithms, notably ID3. Given a classification set C and a feature or decsion formula that creates a grouping set G
      gain(C,G)   =   Σ_c,g p_c,glog₂p_c,g   −   Σ_c p_clog₂p_c
where p_c is the frequency of items of class c in the data and p_c,g is the frequency of items of class c and in grouping g.