Classifying Unbalanced Datasets Using Iterative Fuzzy Support Vector Machine

Author Name(s): P. Aruna Kumari, Dr. G. Jaya Suma
Author Email: arunakumarip.cse@jntukucev.ac.in

Abstract

In real world applications, training the classifier using unbalanced dataset is the major problem, as it decreases the performance of Machine Learning algorithms. Unbalanced dataset can be prominently classified based on Support Vector Machine (SVM) which uses Kernel technique to find decision boundary. High Dimensionality and uneven distribution of data has a significant impact on the decision boundary. By employing Feature selection (FS) high dimensionality of data can be solved by selecting prominent features. It is usually applied as a pre-processing step in both soft computing and machine learning tasks. FS is employed in different applications with a variety of purposes: to overcome the curse of dimensionality, to speed up the classification model construction, to help unravel and interpret the innate structure of data sets, to streamline data collection when the measurement cost of attributes are considered and to remove irrelevant and redundant features thus improving classification performance. Hence, in this paper, two different FS approaches has been proposed namely Fuzzy Rough set based FS and Fuzzy Soft set based FS. After FS the reduced dataset has been given to the proposed Iterative Fuzzy Support Vector Machine (IFSVM) for classification which has considered two different membership functions. The Experiments has been carried out on four different data sets namely Thyroid, Breast Cancer, Thoracic surgery, and Heart Disease. The results shown that the classification accuracy is better for Fuzzy Rough set based FS when compared other.

Introduction

One of the most well notorious supervised machine learning algorithms for classification or prediction is SVM [1]. By calculating a decision boundary called hyper plane, SVM performs the classification of the data points. One of the vital benefits of the SVM is that it can, with relative ease, conquer the high dimensionality problem [2]. It can handle a dataset with a few numbers of instances and with large feature space because of its data driven nature, discriminative power for classifying data points. It has been successfully employed in health care analytics which has improved prediction accuracy [3, 4]. Moreover, in the field of bioinformatics the classification performance has been greatly improved by using SVM [5, 6]. In many real world applications, the training data obtained has been often contaminated by noises. Moreover, some of the data points in the given dataset may be located away from original data space or may be migrated towards incorrect class. One of the main downside of the standard SVM is that the outliers and noisy points present in the dataset can use overfitting while training SVM. In many real world applications of classification due to presence of noise and outliers, a training point may not belongs to one of the class exactly or cannot be considered as noise point. The data point nearer to decision boundary may belong to one of the class or it may be a noisy point. However, this type of uncertainty data points plays a vital role in decision making process because which drives the training process towards overfitting problem. This uncertainty problem can be best handled by fuzzy approaches which diminishes the role of less significant data [7]. In this approach a fuzzy membership value has been calculated for each data point which will be considered as a weight for that point. Basing on these weights the significance of the data points will be evaluated. So many fuzzy approaches are developed and proposed in literature to reduce the effect of outliers. In [8] the fuzzy memberships have been calculated by adopting a similarity measure. Nevertheless, this approach has made an assumption that the outliers must exhibit some variations compares to normal data points. A triangular membership function has been employed for all the data points to eliminate the effect of the parameter C while modeling traditional binary class SVM. On the other hand, based on some assumptions this approach can be adopted [9]. Above two problems are solved by Fuzzy SVM.

Conclusion

In this work, the proposed Fuzzy Approach based machine learning technique has been generated great results on four datasets. The fuzzy rough and fuzzy soft approaches for FS have been employed which has greatly improved the classification performance. Among these FS approaches, Fuzzy soft has given better results with high time complexity. The proposed IFSVM has given promising results on different medical data sets. This can be extended for multi-classification problems where unclassifiable regions may exist if a data point belongs to more than one class or does not belong to any class. When no of classes increases time complexity also increases. This problem can be handled novel model based on support vector domain combined with kernelbased fuzzy clustering.

1,284 total views, no views today

Download PDF File

About the author: admin