Random forest is a technique for building decision trees (essentially, algorithms to classify data points in the form of a flowchart) through algorithmic processing of a prepared set of already-classified data points. As such, it is a technique of decision tree learning, a class of machine learning. The random forest technique can be used on survey data, and has been used, for example, for identifying galaxy mergers.
The general strategy of random forest is to build decision trees on numerous (overlapping) subsets of the data, resulting in many trees, then using a technique to produce an "averaged" decision tree. Compared to other methods, the result has the advantage of producing fairly efficient decision trees and the disadvantage of producing a tree with some inherent imperfection.