n a tree pruning process is completed, a leaf node may not be pure

lass of data points. In this case, a probability that data points

g to one class in the corresponding subspace is calculated for the

. For instance, if m < n of n data points belong to the class A in

sponding subspace denoted by a leaf, the probability of the class

mmonly denoted by m/n and the probability of the class B is

y denoted by (nെm)/n for the leaf.

ning process may produce multiple simpler trees. Therefore, an

tree is selected among multiple pruned trees for the optimal

ation capability in the last step of CART.

a constructed CART tree or model for making predictions on

ta or interpreting the decision-making process using the model

a tree from its root node gradually downwards to a leaf. The root

mines a variable’s value against a threshold to determine where

i.e., to its left branch or its right branch. The same examination

o all the following branch nodes of a tree until reaching a leaf.

eaf node is reached, the prediction is made through maximising

abilities, i.e., whether m/n > (nെm)/n or m/n ൑ (nെm)/n.

T still has a wide application in biological/medical pattern

For instance, it has been used to study how β-Lactam

odynamics plays a role in critical illness due to the infection

y Gram-negative blood stream [Wong, et al., 2020]. The study

vered that a β-lactam fCmin/MIC >1.3 can be a critical indicator

l illness among patients with Gram-negative BSI. Therefore, they

mmended to consider the possibility of using this as an antibiotic

rget. In a study of early detection of blood stream infection among

ients, CART has been used to discover the break point of some

e parameters [Walker, et al., 2020]. The study has discovered

nificant predictive values for the management of BSI among older

main R function for CART is in the library tree. Its syntax is

elow, where x is a matrix (data frame) and formula is used to

e relationship between a dependent variable and independent