使用 function randomForest 时，分类变量的类型应该是什么？

Question

This is just a general theory question, I was asked this question in the college mock interviews for data science, I tried to search for this answer but was unable to get it elsewhere. Hope someone helps me with this. Also I dont have much hands on randomforest

Answer 1

In terms of general theory , random forests can work with both numeric and categorical data. The function randomForest ( documentation here ) supports categorical data coded as factors, so that would be your type.

Answer 2

Machine learning algorithms require features to be encoded in numerical form. You can either one hot encode (0 or 1s) for each level of a feature to indicate its presence or you can label encode such that each level within a feature will then have a numerical value (1,2,3). Typically one-hot encoding is used as label encoding may appear to give an order to the feature. A risk with one-hot encoding is that if you have too many features the feature space will expand too much resulting in a high-dimensional feature set which can be a challenge if not enough data is present. Hence, some approaches only feature encode the most common levels of a feature.

Sources: AceAI Interview Prep, Kaggle, An Introduction To Statistical Learning With Applications in R