Please use this identifier to cite or link to this item:
Title: An efficient implementation of re sampling technique for the dynamic combination of multiple classifiers system
Researcher: Sathiyabama, S
Guide(s): Duraiswamy, K
Upload Date: 8-Jul-2014
University: Periyar University
Completed Date: 02/07/2007
Abstract: Recently mining from data streams has become an important and newlinechallenging task for many realworld applications The classification of large data set newlineis an important problem in data mining Noone classification technique is always newlinesuperior to the others in terms of classification accuracy In recent years some multiple newlineclassifier combination techniques were proposed to improve the performance of newlineclassifiers Multiple classifier combination is a technique of combining the decisions of newlinedifferent classifiers which are trained to solve the same problem but make different newlineerrors A proper combination of multiple classifiers should produce more reliable newlinerecognition of results than any of the individual classifiers Hence multiple newlineindependent approaches can be applied to a classification problem each yielding its newlineown prediction The results of these techniques can then be combined In the proposed newlinemethod the Decision tree classifier K Nearest Neighbor and the Neural network newlineclassifiers are constructed using a new technique known as Resampling based on newlinethreshold RST These classifiers are combined based on the newly proposed method newlineknown as Dynamic Combination of Multiple Classifiers system DCMCS to classify newlinea test instance at run time newlineWhen constructing individual classifiers the new technique RST has been newlineused to improve the performance of classifier The databases considered for the newlineanalysis are the Adult database and the Earthquake database For both the databases newlinethe probability of occurrences of every class for the entire training data set has been newlineestimated Based on RST thresholds have been fixed for all the classes When the data newlineset have been selected randomly the probabilities of the classes have been checked newlineagainst the thresholds If it does not satisfy the threshold Resampling is performed newlineThe study made on three different classifiers shows that the data set which satisfies the newlinethreshold for all the classes produces better performance than the one which doesnt newlinesatisfy the threshold The classifiers constructed using the ran
Pagination: XIX,143p.
Appears in Departments:Department of Computer Science

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File13.45 kBAdobe PDFView/Open
02_certificate.pdf15.58 kBAdobe PDFView/Open
03_acknowledgement.pdf18.65 kBAdobe PDFView/Open
04_abstract.pdf35.07 kBAdobe PDFView/Open
05_contents.pdf39.22 kBAdobe PDFView/Open
06_list of tables,figures,expansion.pdf58.36 kBAdobe PDFView/Open
07_chapter 1.pdf470.62 kBAdobe PDFView/Open
08_chapter 2.pdf210.12 kBAdobe PDFView/Open
09_chapter 3.pdf690.75 kBAdobe PDFView/Open
10_chapter 4.pdf528.97 kBAdobe PDFView/Open
11_chapter 5.pdf623.73 kBAdobe PDFView/Open
12_chapter 6.pdf377.46 kBAdobe PDFView/Open
13_chapter 7.pdf115.91 kBAdobe PDFView/Open
14_appendix.pdf63.49 kBAdobe PDFView/Open
15_references.pdf226.95 kBAdobe PDFView/Open
16_list of publications.pdf55.51 kBAdobe PDFView/Open

Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.