• Used in C4.5 (a successor of ID3) to overcome bias towards attributes with many values
    • Normalises information gain 标准化信息增益

    image.png
    splitInforepresents the potential i_nformation generated by splitting D into v partitions, corresponding to the v outcomes of a test on A. 表示通过将D分割成v个分区而产生的潜在信息，对应于a上测试的v个结果。
    Now we define
    _GainRatio(A) = Gain(A)/SplitInfo(A)

    Example (continued from previous):

    image.png
    gain_ratio(income) = 0.029/1.557 = 0.019

    The attribute with the maximum gain ratio is selected as the splitting attribute.