YouTube Video Review: XGBoost in Python (Hyperparameter Tuning)

I have recently written a blog post on how to use XGBoost and machine learning to solve a solve a univariate time series problem, which I found to be somewhat successful. When I wrote the code for the time series dataset that I had made predictions on, I did not attempt to fine tune the parameters.

I found a YouTube video that discussed parameter tuning and decided to watch it because I am not very familiar with hyperparameter tuning of XGBoost, although this library is said to be very good at making accurate predictions and winning awards. The review of the contents of the video is as follows:-

Overfitting occurs when the model learns too intensely on the training data. When this occurs, the training and fitting of the data is not generalised and the accuracy of the predictions on the test set will be poor. An optimum model is the goal of model fitting.

The hyperparameter, max_depth, describes how much depth the model is allowed to grow. When the max_depth is increased, this can lead to overfitting.

The hyperparameter, subsample, is the percentage of the data that is being built.

The hyperparameter, colsample_bytree, is the subsample of columns when constructing each tree. Subsampling occurs once every tree is constructed.