IJERT-EMS
IJERT-EMS

Improved kNN Algorithm by Optimizing Cross-validation


Improved kNN Algorithm by Optimizing Cross-validation
Authors : Ms. Soniya S. Dadhania, Prof. J. S. Dhobi
Publication Date: 30-05-2012

Authors

Author(s):  Ms. Soniya S. Dadhania, Prof. J. S. Dhobi

Published in:   International Journal of Engineering Research & Technology

License:  This work is licensed under a Creative Commons Attribution 4.0 International License.

Website: www.ijert.org

Volume/Issue:   Vol.1 - Issue 3 ( May- 2012)

e-ISSN:   2278-0181

Abstract

Nowadays web applications based on short text is increasing rapidly. Moreover, the classification algorithms which are applied to short text data are Support Vector Machines algorithm, k-Nearest Neighbors algorithm and Naive Bayes algorithm. kNN algorithm depends on the distance function and the value of k nearest neighbor. Traditional kNN algorithm can select best value of k using cross- validation but there is unnecessary processing of the dataset for all possible values of k. Proposed kNN algorithm is an optimized form of traditional kNN by reduceing the time and space for evaluating the algorithm. Experiments are performed in developer version of weka 3.7.5.Comparison of proposed kNN algorithm is done with traditional kNN algorithm, Support vector machine and Nave Bayes algorithm. The proposed algorithm is more promising than the traditional kNN algorithm as time taken to process and space used for cross-validation in classification are reduced.

Citations

Number of Citations for this article:  Data not Available

Keywords

Key Word(s):    

Downloads

Number of Downloads:     1628
Similar-Paper

Call for Papers - May - 2017

        

 

                 Call for Thesis - 2017 

     Publish your Ph.D/Master's Thesis Online

              Publish Ph.D Master Thesis Online as Book