Explore advanced modeling techniques using #TidyTuesday data on board game ratings in this 48-minute screencast. Dive into custom feature engineering, xgboost tuning, and explainability methods. Learn about data overview, average distribution, preprocessing, custom tokenizing, and string squishing. Delve into regression, tuning, and result analysis. Discover how to plot function games, find the best game, and work with testing sets. Gain insights into explainability tools, including parsnip fit, model importance, and dependency partial plots. Examine min age plots and create summaries to enhance your understanding of feature engineering and interpretability in xgboost models.
Feature Engineering & Interpretability for XGBoost with Board Game Ratings