Editing Random forest (section)

===From bagging to random forests===
{{main|Random subspace method}}
The above procedure describes the original bagging algorithm for trees. Random forests also include another type of bagging scheme: they use a modified tree learning algorithm that selects, at each candidate split in the learning process, a [[Random subspace method|random subset of the features]]. This process is sometimes called "feature bagging". The reason for doing this is the correlation of the trees in an ordinary bootstrap sample: if one or a few [[Feature (machine learning)|features]] are very strong predictors for the response variable (target output), these features will be selected in many of the {{mvar|B}} trees, causing them to become correlated. An analysis of how bagging and random subspace projection contribute to accuracy gains under different conditions is given by Ho.<ref name="ho2002">{{cite journal | first = Tin Kam | last = Ho | title = A Data Complexity Analysis of Comparative Advantages of Decision Forest Constructors | journal = Pattern Analysis and Applications | volume = 5 | issue = 2 | year = 2002 | pages = 102–112 | url = http://ect.bell-labs.com/who/tkh/publications/papers/compare.pdf | doi = 10.1007/s100440200009 | s2cid = 7415435 | access-date = 2015-11-13 | archive-date = 2016-04-17 | archive-url = https://web.archive.org/web/20160417091232/http://ect.bell-labs.com/who/tkh/publications/papers/compare.pdf | url-status = dead }}</ref>

Typically, for a classification problem with {{mvar|p}} features, {{sqrt|{{mvar|p}}}} (rounded down) features are used in each split.<ref name="elemstatlearn"/>{{rp|p=592}}  For regression problems the inventors recommend {{math|''p''/3}} (rounded down) with a minimum node size of 5 as the default.<ref name="elemstatlearn"/>{{rp|p=592}} In practice, the best values for these parameters should be tuned on a case-to-case basis for every problem.<ref name="elemstatlearn"/>{{rp|592}}