Volume 21, Issue 2, April 2016, Pages 387–396
P. Velvizhy1, S. Abayambigai2, and A. Kannan3
1 Department of Computer Science and Engineering, Anna University, Chennai, India, India
2 Department of Computer Science and Engineering, Anna University, Chennai, India, India
3 Department of Information Science and Technology, Anna University, Chennai, India, India
Original language: English
Copyright © 2016 ISSR Journals. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Boosting is the general method which converts any weak learning algorithm into strong learner in order to improve the accuracy. The limitations in boosting is overfitting on the training data and filtering out the correct data in the subsequent function since boosting concentrates on regions not predicted well by other learners. So, the cluster based boosting (CBB) approach is used to address limitations in boosting. In this paper, initially X-Means algorithm is used to cluster the data and the clusters are selectively boosted based on the additional structure information provided by clusters and previous function accuracy on the member data. To apply Cluster Based Boosting to the high dimensional data, dimensionality reduction technique is performed. In this paper, we apply Global Redundancy Minimization frame work which considers the redundancy of the feature with all other features. The selected features will contribute more mutual information for prediction. This frame work can be used with any other feature selection technique. We provide experimental results on various dataset. These results demonstrate the effectiveness of Global redundancy framework and also effectiveness of Cluster Based Boosting with Global redundancy minimization framework than classifier with global redundancy minimization framework.
Author Keywords: Boosting, Clustering algorithms, Margin theory, Machine learning, feature selection, redundancy minimization.
P. Velvizhy1, S. Abayambigai2, and A. Kannan3
1 Department of Computer Science and Engineering, Anna University, Chennai, India, India
2 Department of Computer Science and Engineering, Anna University, Chennai, India, India
3 Department of Information Science and Technology, Anna University, Chennai, India, India
Original language: English
Copyright © 2016 ISSR Journals. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
Boosting is the general method which converts any weak learning algorithm into strong learner in order to improve the accuracy. The limitations in boosting is overfitting on the training data and filtering out the correct data in the subsequent function since boosting concentrates on regions not predicted well by other learners. So, the cluster based boosting (CBB) approach is used to address limitations in boosting. In this paper, initially X-Means algorithm is used to cluster the data and the clusters are selectively boosted based on the additional structure information provided by clusters and previous function accuracy on the member data. To apply Cluster Based Boosting to the high dimensional data, dimensionality reduction technique is performed. In this paper, we apply Global Redundancy Minimization frame work which considers the redundancy of the feature with all other features. The selected features will contribute more mutual information for prediction. This frame work can be used with any other feature selection technique. We provide experimental results on various dataset. These results demonstrate the effectiveness of Global redundancy framework and also effectiveness of Cluster Based Boosting with Global redundancy minimization framework than classifier with global redundancy minimization framework.
Author Keywords: Boosting, Clustering algorithms, Margin theory, Machine learning, feature selection, redundancy minimization.
How to Cite this Article
P. Velvizhy, S. Abayambigai, and A. Kannan, “Enhanced Cluster Based Boosting in High Dimensional Data,” International Journal of Innovation and Scientific Research, vol. 21, no. 2, pp. 387–396, April 2016.
