3.13. Semi-Supervised Learning and Random Forests

Not only do semi-supervised learning and random forests also apply to classification trees, they also apply to regression trees.

3.13.1. Random Forests

In a random forest we generate a range of diverse regression trees. We then make a prediction using each regression tree and then take the average of these values and that value becomes our prediction.

3.13.2. Semi-Supervised Learning

We can add more samples to our training data by randomly generating samples and predicting their corresponding labels. While we can predict labels using just a single regression tree, we will get more accurate labels if we use our random forest.

Let’s generate 3 fictional days by picking randomly picking weather conditions (temperature and whether or not it’s raining).

Let’s make predictions based on our random forest.

Now we can add these samples to our training data.