Another method to create trees in a Random Forest ensemble is Extra Trees (extremely randomized trees). The main difference with the previous method is that the feature and split point combination does not have to be the optimal. Instead, a number of split points are randomly generated, one for each available feature. The best split point of those generated is selected. The algorithm constructs a tree as follows:

1. Select the number of features m that will be considered at each node and the minimum number of samples n in order to split a node
2. For each base learner, do the following:
   1. Create a bootstrap train sample
   2. Select the node to split (the node must have at least n samples)
   3. Select m features randomly
   4. Randomly generate m split points, with values between the minimum and maximum value of each feature
   5. Select the best of these split points
   6. Split the node into two nodes and repeat from step 2-2 until there are no available nodes