Random Forest Models

Random Forest is a method for improving predictions by asking multiple simple decision trees the same question and combining their outputs.

A decision tree works like a flowchart.

If this is true, go left.

If not, go right.

A single tree can be wrong because it learns biases in the data.

Random Forest fixes that.

It builds many trees, each trained on a different random sample of the data. Each tree also considers a random subset of features, thereby forcing trees to think independently rather than copying one another.

When it is time to predict, all trees vote. For categories, the forest picks the most common answer. For numbers, it averages the results.

This averaging cancels out many mistakes. The result is higher accuracy and lower overfitting than that of a single tree.

Simple mental model.

One tree is one opinion.

A random forest is a crowd. Crowds are usually smarter than individuals when problems get noisy.