XGBoost is the most popular boosting technique in recent times. Although there have been various new versions that have been developed by large corporations, XGBoost still remains the undisputed king. Let's look at a brief history of boosting.
Boosting differs from bagging in its core principles; the learning process is, in fact, sequential. Every model built in an ensemble is ideally an improved version of the previous model. To understand boosting in simple terms, imagine you are playing a game where you must remember all the objects placed on the table that you are shown just once for 30 seconds. The moderator of the game arranges around 50-100 different objects on a table, such as a bat, ball, clock, die, coins, and so on, and covers them with a large piece of cloth. When the game begins, he withdraws the cloth from the table and gives you exactly 30 seconds to see them and puts the curtain back. You now must recollect all the objects you can...