Introducing decision trees with XGBoost
XGBoost is an ensemble method, meaning that it is composed of different machine learning models that combine to work together. The individual models that make up the ensemble in XGBoost are called base learners.
Decision trees, the most commonly used XGBoost base learners, are unique in the machine learning landscape. Instead of multiplying column values by numeric weights, as in linear regression and logistic regression (Chapter 1, Machine Learning Landscape), decision trees split the data by asking questions about the columns. In fact, building decision trees is like playing a game of 20 Questions.
For instance, a decision tree may have a temperature column, and that column could branch into two groups, one with temperatures above 70 degrees, and one with temperatures below 70 degrees. The next split could be based on the seasons, following one branch if it's summer and another branch otherwise. Now the data has been split into four...
