Decision Trees

Question 1

What is ensembles?

Answer 1

Ensembles are techniques that use multiple machine learning methods to create more powerful models. There are two categories for these: random forests and gradient boosted decision trees.

Question 2

What is random forests?

Answer 2

Random forests are an ensemble method in order to overcome the overhitting disadvantage of decision trees. Many different decision trees are produced which over-fit in their unique way, Combining the above you have an overall model that avoids in some degree ovefitting.

Question 3

Why random trees are called random?

Answer 3

The aim of this ensemble method is to produce UNIQUE trees. Somehow, we should provide a randomness propriety among them. This takes place by choosing a different portion of the dataset or selecting a different feature to split

Question 4

When you need to classify a point, how the classification takes place in the random forest trees?

Answer 4

For random forests the following procedure takes place when it comes to predict values in classification cases.
Each test data point is traversed on the whole tree and a soft probability is assigned for each class in the dataset.
This is averaged across the probabilities of all the trees.