|Title:||Evolution of Random Forest from Decision Tree and Bagging: A Bias-Variance Perspective|
|Keywords:||Supervised machine learning, Ensemble learning algorithms, Decision tree, Bagging, Random forests, Bias-variance tradeoff, Correlation.|
The ensemble methods are one of the most heavily used techniques in machine learning. The random forest arguably spearheads this army of learners. Being sprung from the decision tree in the late 90s, the benefits of a random forest have rightfully attracted practitioners to widely and successfully apply this powerful yet simple-to-understand technique to numerous applications. In this study we explain the evolution of a random forest from a decision tree in the context of bias and variance of learning theory. While doing so, we focus on the interplay between the correlation and generalization error of the random forest. This analysis is expected to enrich the literature of random forests by providing further insight into its working mechanism. These insights will assist the practitioners of the random forest implement this algorithm more wisely and in an informed way.