Page 48 - FULL REPORT 30012024

iii. Decision Tree (DT)

A common machine learning approach used for both classification

and regression applications are the decision tree. According to Uddin

et al. (2019) article, it is a supervised learning technique that creates
a tree-like model of choices and potential outcomes. The tree is made

up of nodes and branches, where each leaf node stands in for an

outcome or prediction, each branch for a decision rule, and each
interior node for a trait or attribute. The decision tree algorithm

divides the data recursively according to the values of several features
after starting with the complete dataset at the root node. To maximise

information gain or reduce impurity at each node, the splits are
chosen. Impurity refers to the consistency of the target variable within

a certain subset of data, whereas information gain assesses how much

information a feature contributes in lowering uncertainty about the
outcome. Figure 2.11 illustrates a decision where the choice outcomes

(Class A and Class B) are depicted by rectangles, and each variable
(C1, C2, and C3) is represented by a circle. Each branch is marked

with either "True" or "False" based on the result value from the test
of its predecessor node in order to correctly assign a sample to a class.

Figure 2.11 Decision Tree
(Source: Uddin et al., 2019)

31