May 2019
Intermediate to advanced
162 pages
4h 24m
English
First, we will compute the Gini impurity of the root node, which is the pre-split state. We get 0.444, as shown:

The next stage in the algorithm is to iterate each feature. There are three cases, shown as follows. Using our IG formula, we can compute which is the best split point for this feature. The first happens to be the best, in this case:

Splitting on the second case, where x1 is greater than or equal to 4, is not a good idea since the result is no different than the state at the root. Therefore, our information gain ...
Read now
Unlock full access