A Process For Online Dynamic Learning With Cost Sensitivity In Data Mining

  Mr. Prashant Mahakal, ,  Prof. Pritesh Jain

ABSTRACT:- In general, performance of the classifier is measure using accuracy i.e. on the basis of number of incorrectly predicted instances in testing phase. Cost of what is misclassified is not considered for the measuring performance in general approaches; cost sensitive classification considers cost of the misclassified label. In online learning, prediction model is updated is predicted label and actual label are not same in each round, but in real applications every time getting the actual class is not possible so there come concept of online dynamic learning. Current online dynamic learning systems not consider cost of the misclassification. The systems propose online dynamic learning system which considers the cost of the misclassification. Malicious uniform resource locator (URL) detection is one of the applications where getting actual label of the instance is not possible and class distribution of malicious and normal URL is unbalanced. To evaluate proposed system implemented the Malicious URL detection system using real world dataset which outperforms that existing Malicious URL detection system.

Keywords- Cost-sensitive classification, online anomaly detection, online learning.

Unique Identification Number - IJEDR1604123
Page Number(s) - 828-833
Pubished in - Volume 4 | Issue 4 | December 2016
DOI (Digital Object Identifier) -   
Publisher - IJEDR (ISSN - 2321-9939)

  Mr. Prashant Mahakal, ,  Prof. Pritesh Jain,   "A Process For Online Dynamic Learning With Cost Sensitivity In Data Mining", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.4, Issue 4, pp.828-833, December 2016, Available at :http://www.ijedr.org/papers/IJEDR1604123.pdf