Editing Random forest (section)

== Unsupervised learning ==
As part of their construction, random forest predictors naturally lead to a dissimilarity measure among observations. One can analogously define dissimilarity between unlabeled data, by training a forest to distinguish original "observed" data from suitably generated synthetic data drawn from a reference distribution.<ref name=breiman2001/><ref>{{cite journal |author=Shi, T. |author2=Horvath, S. |year=2006 |title=Unsupervised Learning with Random Forest Predictors |journal=Journal of Computational and Graphical Statistics |volume=15 |issue=1 |pages=118–138 |doi=10.1198/106186006X94072 |jstor=27594168|citeseerx=10.1.1.698.2365 |s2cid=245216 }}</ref> A random forest dissimilarity is attractive because it handles mixed variable types very well, is invariant to monotonic transformations of the input variables, and is robust to outlying observations. Random forest dissimilarity easily deals with a large number of semi-continuous variables due to its intrinsic variable selection; for example, the "Addcl 1" random forest dissimilarity weighs the contribution of each variable according to how dependent it is on other variables. Random forest dissimilarity has been used in a variety of applications, e.g. to find clusters of patients based on tissue marker data.<ref>{{cite journal | vauthors = Shi T, Seligson D, Belldegrun AS, Palotie A, Horvath S | title = Tumor classification by tissue microarray profiling: random forest clustering applied to renal cell carcinoma | journal = Modern Pathology | volume = 18 | issue = 4 | pages = 547–57 | date = April 2005 | pmid = 15529185 | doi = 10.1038/modpathol.3800322 | doi-access = free }}</ref>