An Improved Ensemble Classifier Results with Consideration of Classifier Importance

International Journal of Computer Trends and Technology (IJCTT)          
© 2016 by IJCTT Journal
Volume-36 Number-2
Year of Publication : 2016
Authors : C. Gayathri, R.Umarani


C. Gayathri, R.Umarani "An Improved Ensemble Classifier Results with Consideration of Classifier Importance". International Journal of Computer Trends and Technology (IJCTT) V36(2):96-100, June 2016. ISSN:2231-2803. Published by Seventh Sense Research Group.

Abstract -
Financial fraud detection is the most concerned and complicated task in the real world environment which would requires complete and structured financial data set for accurate detection. These issues are resolved in our previous work namely Optimal Ensemble Architecture Selection using FFA (OEAC-FFA). This work would select the most optimal ensemble classifier architecture for the accurate detection financial fraud behaviour by changing the ensemble architecture of different combination of base classifiers and subset features in every iteration. This methodology is proved to be better in the accurate classification result. OEACFFA makes use of weighted average methodology for obtaining the final classification result from the ensemble classifier. However the weighted average method cannot produce an accurate ensemble result where the classifiers that are not provided with equal importance. In real world environment, classifier should be differentiated in importance based on their classification result. This problem is resolved in this research work by introducing the novel approach called the Dempster Shafer based Ensemble Classifier method (DSECM). The dempster shafer theory can give better ensemble classifier result than the weighted average method in presence of different importance values of classifiers too. This approach would retrieve the ensemble result as the base classifier result which is having more confidence in terms of their weight. The experimental tests were conducted in the matlab simulation environment and the evaluation is done by comparing it with the OEAC_FFA which is based on weighted average method. This is proved that the DSECM approach is better than the other mechanisms.

[1] Ordonez Jon Geiler, Li Hong, Guo Yue-jian , “An Adaptive Sampling Ensemble Classifier for Learning from Imbalanced Data Sets “, Proceeding of an international conference , 2010.
[2] Shuo Wang and Huanhuan Chen and Xin Yao, “Negative Correlation Learning for Classification Ensembles”, 2007. [3] Albert Bifet , Geoff Holmes and Bernhard Pfahringer , “New Ensemble Methods For Evolving Data Streams”, 2009.
[4] Ashfaqur Rahman and Brijesh Verma , “ Novel Layered Clustering-Based Approach for Generating Ensemble of Classifiers”, IEEE transactions on neural networks, vol. 22, no. 5, may 2011 781.
[5] Brijesh Verma and Ashfaqur Rahman,, “ Cluster-Oriented Ensemble Classifier: Impact of Multicluster Characterization on Ensemble Classifier Learning”, IEEE transactions on knowledge and data engineering, vol. 24, NO. 4, APRIL 2012.
[6] Nicolás García-Pedrajas, “Constructing Ensembles of Classifiers by Means of Weighted Instance Selection”, IEEE transactions on neural networks, vol. 20, no. 2, February 2009.
[7] Michael D. Muhlbaier, Apostolos Topalis, and Robi Polikar,” Learn .NC: Combining Ensemble of Classifiers With Dynamically Weighted Consult-and-Vote for Efficient Incremental Learning of New Classes”, IEEE Transactions on neural networks, vol. 20, no. 1, January 2009.
[8] Daniel Herna´ndez-Lobato, Gonzalo Martı´nez-Mun˜ oz, and Alberto Suarez, “ Statistical Instance-Based Pruning in Ensembles of Independent Classifiers”, IEEE Transactions on pattern analysis and machine intelligence, vol. 31, no. 2, February 2009.
[9] Gonzalo Martınez-Munoz, Daniel Hernandez-Lobato, and Alberto Suarez,” An Analysis of Ensemble Pruning Techniques Based on Ordered Aggregation”, IEEE transactions on pattern analysis and machine intelligence, vol. 31, no. 2, February 2009

Financial Fraud detection, Fusion, Dempster shafer, confidence score.