Abstract—Classification of unbalanced information is a
problem of machine learning frameworks and learning of
existing basic algorithms which will be more complicated if the
data has more than two classes. Therefore, this research
proposes a concept to improve the classification of unbalanced
data with more than two classes with a model called
One-Versus-One Classification Technique based on Data
Synthesis with Appropriate Distant Neighbors (OVO-SynDN).
This OVO-SynDN model will divide the problem of
classification of multiclass data into binary class data
classification with the learning technique “One-Versus-One”.
Then it will adjust the information imbalance by synthesizing
the data which selects nearest neighbors with Euclidean
distance techniques. For the amount of data to be synthesized of
each data set, it will be selected from the number of nearest
neighbors that are in the opposite group. The features of the
new synthesized data, it depends on the characteristics of the
original data and the nearest neighbors. Then combine the
existing imbalanced data sets and synthetic data sets to
construct the learning model. The standard algorithm, Support
Vector Machine (SVM) with polynomial kernel function, will be
selected to learn these data sets. Some parameters are adjusted
so that the algorithm is suitable for learning each data set. The
results show that the OVO-SynDN model has satisfactory
performance and reliability with high MAvA and MFA values.
In addition, the OVO-SynDN model can still classify
imbalanced data better than the four techniques that are
compared. That means that the proposed method can be
applied to the classification of unbalanced data that has more
than two classes.
Index Terms—SVM with polynomial kernel function,
synthesizing imbalance data, multiclass imbalanced datasets
classification, one-versus-one, binary decomposition.
P. Chujai is with the Electrical Technology Education Department,
Faculty of Industrial Education and Technology, King Mongkut’s University
of Technology Thonburi, Bangkok, Thailand (e-mail:
pasapitchchujai@gmail.com).
K. Chaiyakhan with Computer Engineering Department, Rajamangala
University of Technology Isan, Muang, Nakhon Ratchasima, Thailand
(e-mail: kedkarnc@hotmail.com).
N. Kerdprasop and K. Kerdprasop are with the School of Computer
Engineering, Suranaree University of Technology, Nakhon Ratchasima,
Thailand (e-mail: nittaya.k@gmail.com, kittisakthailand@gmail.com).
[PDF]
Cite:Pasapitch Chujai, Kedkarn Chaiyakhan, Nittaya Kerdprasop, and Kittisak Kerdprasop, "The One-Versus-One Classification Technique Based on Data Synthesis with Appropriate Distant Neighbors," International Journal of Information and Electronics Engineering vol. 9, no. 2, pp. 43-49, 2019.