package learners
Type Members
-
case class
RandomForest(numTrees: Int = -1, useJackknife: Boolean = true, biasLearner: Option[Learner] = None, leafLearner: Option[Learner] = None, subsetStrategy: Any = "auto", minLeafInstances: Int = 1, maxDepth: Int = Integer.MAX_VALUE, uncertaintyCalibration: Boolean = false, randomizePivotLocation: Boolean = false, randomlyRotateFeatures: Boolean = false) extends Learner with Product with Serializable
Standard random forest as a wrapper around bagged decision trees Created by maxhutch on 1/9/17.
Standard random forest as a wrapper around bagged decision trees Created by maxhutch on 1/9/17.
- numTrees
number of trees to use (-1 => number of training instances)
- useJackknife
whether to use jackknife based variance estimates
- biasLearner
learner to model bias (absolute residual)
- leafLearner
learner to use at the leaves of the trees
- subsetStrategy
for random feature selection at each split (auto => all fetures for regression, sqrt for classification)
- minLeafInstances
minimum number of instances per leave in each tree
- maxDepth
maximum depth of each tree in the forest (default: unlimited)
- uncertaintyCalibration
whether to empirically recalibrate the predicted uncertainties (default: false)
- randomizePivotLocation
whether generate splits randomly between the data points (default: false)
- randomlyRotateFeatures
whether to randomly rotate real features for each tree in the forest (default: false)