Abstract Views :150 |
PDF Views:3
Authors
Affiliations
1 Department of Electronics Technology, Guru Nanak Dev University, Amritsar, IN
Source
Research Cell: An International Journal of Engineering Sciences, Vol 17, No 1 (2016), Pagination: 168-174
Abstract
Deciding the criterion for the performance evaluation of a classifier plays a vital role in the selection procedure of a classifier for a certain problem. These criteria empower the researchers to do the selection of a classifiers for effective classifications of unseen data from a range of classifying algorithms. A great number of different measures are currently available for the classification problems based on binary, flat or undistributed data such as in case of images. However in case of hierarchical classifications, where the number of classes to be identified are more than two, the evaluation of a classifier becomes more and more intricate as the classes to be differentiated, are hierarchically attached. The topic of focus of this paper is to provide a knowledge flow which a researcher can use while dealing with such real time based problems where the accuracy and efficiency of a classifier are the major concerns. The problem of interest while discussing the different aspects of various evaluation measures, was the color prediction of paddy crop plant leaf for its health characterization.
Keywords
Machine Learning, Confusion Matrix, ROC, AUC, Cost Curve, Accuracy.
Full Text
