A new variable selection approach using Random Forests

摘要: Random Forests are frequently applied as they achieve a high prediction accuracy and have the ability to identify informative variables. Several approaches for variable selection been proposed combine intensify these qualities. An extensive review of corresponding literature led development new approach that is based on theoretical framework permutation tests meets important statistical properties. A comparison another eight popular methods in three simulation studies four real data applications indicated that: can also be used control test-wise family-wise error rate, provides higher power distinguish relevant from irrelevant variables leads models which located among very best performing ones. In addition, it equally applicable regression classification problems.