Dynamic Nonparametric Random Forest Using Covariance

Table 1

Characteristics of various RF algorithms.

	Goal	Learning Characteristics	Drawback
PERT [14]	Accuracy Improvement	To set the cut point between randomly selected two learning instances	Accuracy varies depending on learning data
Rodriguez et al.’s algorithm [15]	Accuracy Improvement	To find key attributes using PCA before learning	Accuracy varies depending on learning data
Robnik-Sikonja’s algorithm [16]	Accuracy Improvement	To get the target value using weighted voting	Accuracy varies depending on a well-formed tree
Bernard et al.’s algorithm [17]	Accuracy Improvement	To assign weights to learning instances	Much calculation time and the possibility of overfitting
A. Cuzzocrea et al.’s algorithm [18]	Optimal Number of Trees	To select the number of trees through many tests	Waste of memory and high computation costs
P. Latinne et al.’s algorithm [19]	Optimal Number of Trees	To use as an input parameter	The number of trees varies depending on