Page 92 - Contributed Paper Session (CPS) - Volume 6

CPS1832 Nur Fazliana Rahim et al.
The data was obtained from the Bank Negara Malaysia. In this part, the
dataset of Foreign Exchange Rate (FER) data were separated into two
subsets; FER-1 used for training and FER-2 used for testing. Each dataset
comprises 30 cases. The FER data includes three characteristic; One month
previous, Two months previous and Three months previous. There are
three classification outcomes of the FER rank; Small, Intermediate and
Large.
B. Part 2: Creation of Fuzzy Rules
To generate the required fuzzy model using WSBA was introduced in
this phase. To create a system that is more readily understandable, WSBA
uses a rule generation algorithm based on fuzzy general rules or addition
of a Mamdani-type Fuzzy Rule Based Systems (FRBS). Five steps involve in
this part as follows.
First, three subgroups obtained from the training dataset based on the
classification outcomes. The measure of location, method used to

classify the outcomes as follows
−
= + ( 4 ) (1)

where = 1, 2, 3, = cumulative frequency before the class, =

total number of observations, =cumulative frequency before the

class, =frequency of the class where lies, and = size of the class

where lies.

Second, the fuzzy membership function for FER dataset was
constructed based on the measure of location; , and calculated
1
3
2
respectively from equation (1). Then, the fuzzy partition was defined from
the established fuzzy membership function. It was used to convert crisp
values into fuzzy values.
Third, for each linguistic term, fuzzy subsethood values were
calculated in each subgroup. This generated rules able to deal with
classification problems. The fuzzy subsethood value of A taking into
account , (, ) refers the degree levels to which A is subset of B [22],
[23]:

where (, )[0,1].
Fourth, using the subsethood values in Step 3, each linguistic term weights
then were obtained. In this research, weighting is restricted between 0 to 1,
which 0 is referring to the smallest weight (or less significance) and 1 the
largest weight (or more significance). The subsethood here is mean to extend

81 | I S I W S C 2 0 1 9