Abstract
The aim of this paper was to consider the reliability of the basic metrics of evaluation of classification models: accuracy, sensitivity, specificity and precision. The WEKA software tool was applied to the "Hepatitis C virus (HCV) for Egyptian patients dataset". The algorithms Bayesnet, Naivebayesh, Multilayer Perceptron, J48 and 10-fold cross validation were used in the study. The main results obtained are that, with all four algorithms in question, they achieved approximately the same accuracy of correctly classified specimens. BaiesNet - 22.96%, NaiveBaies - 26.14%, MultilaierPerceptron - 26.57% and J48 - 25.27%. Binary classification metrics - sensitivity, specificity and precision show very different values, depending on the intended class. Metric specificity, for all four algorithms, shows that a value that is in most of the range of possible values [0,1]. Metric sensitivity and precision, for all four algorithms, showed values that are in the lower part of the range of possible values [0,1]. The results of this study showed that WEKA software could not yet be considered as a relevant tool for the diagnosis of Hepatitis C Virus, on whose data set it was applied.
A case study of the application of WEKA software to solve the problem of liver inflamation