| Astrophysics (Index) | About |
Random forest is a machine learning/decision technique that makes use of multiple decision trees. Decision trees are designed to classify data and can be built automatically based upon a set of sample data along with associated classifications (decision tree learning), basically a type of machine learning by example. Techniques for building and using a single decision tree from such data can produce an overly-detailed tree, with a lot of logic contributing little. The random forest strategy is to produce a set of such decision trees, but each limited in detail, then applying them all, then taking a vote to produce the ultimate classification. The individual decision trees are built normally, but each built from a subset of the data sample. The random forest technique can be used on survey data, and has been used, for example, for identifying galaxy mergers.