Page 14 - 22-0424

The International Journal of the Royal Society of Thailand
Volume XI - 2019

can be solved by any computer if those problems are transformable to classification
problems. Once the problem is transformed into the problem of classification,
some efficient algorithms to solve the problem must be developed. These

classification algorithms make the machine look intelligent from the human
perspective, which is called artificial intelligence.

There are various ways to classify a set of data according to the properties,
features, or characteristics of the data. For example, the features of a patient
may be his/her temperature, weight, height, blood pressure, size of waist, and
symptoms measured in numeric form. Each class has a specific set of feature values.
Two of the simplest approaches to classify features are by tree search (https://
en.wikipedia.org/wiki/Digital_health) and rule-based filtering (Senanarong V
et al., 2018). For tree search, the value of each feature is checked one at a time
with a predefined threshold value of each class until reaching the last feature to
predict the class. Playing the game Tic-Tac-Toe (Xs and Os) is a good example of
tree search. A computer generates all possible answers in each move and evaluates
each answer as a score. It then selects the best answer of the current move and,
from this best answer, it generates the next possible answers in the next move in
advance. This process is iterated until no more moves exist. The computer can
gain the potential to win the game by traversing this search tree. For rule-based
filtering, the approach is similar to tree search but several conditions are formed
by simultaneously considering the values of features defined in terms of logic
statements by using AND, OR, NOT as the conjunctions. A correct class is
determined if the values of all features according to a pre-defined rule of the

class is tested true. For example, suppose the goal is to determine whether a
considered object is a box or not by testing its features of orthogonality of the
dimensions, width, height, and length. The rule for this example should be:

If width > 0 and height > 0 and length > 0 and all dimensions are
orthogonal, then the object is a box.
To make the testing conditions more versatile to a real situation, some
fuzzy words, such as “very hot”, or “rather chubby”, can be included and
transformed into a numeric value by using a membership function, such as
the study of applying fuzzy rule-based classification to assess coronary artery

disease (Bovornkitti 2020).

8 Digital Health and Precision Medicine

11/7/2565 BE 13:27
_22-0424(005-010)2.indd 8 11/7/2565 BE 13:27
_22-0424(005-010)2.indd 8