Summary

XGBoost is a refinement of existing gradient-boosted tree algorithms with enhancements that make it faster and less prone to overfitting. The authors of the algorithm addressed problems with sparse data by having a default direction for the splitting algorithm so rows with missing values can be handled quickly. By storing the data in sorted compressed blocks, XGBoost can scale in terms of memory and compute nodes.

In the next chapter, you’ll use XGBoost to perform a prediction of house value by modifying the code you wrote in Chapter 2. You will change the code from a classification model to a prediction model. By the end of Chapter 4, you will have example code to build from for both classification and prediction models.