THE EFFECT OF USING DUMMY VARIABLE ON CLASSIFICATION OF WOMB DISEASE WITH C4.5 METHOD

Abstract

The use of dummy variables is recommended because the symptoms of the womb disease compounds that have the possible values that appear more than two (non-binary), there is a possibility that not all types of occurrence related to the disease symptoms as other content that needs to be done solving the symptoms so that the value to binary and symptoms become more specific. By applying the dummy variable, is expected to improve the accuracy of the probabilistic approach Naïve Bayes classifier, because the assumption of independency between the symptoms of the disease are met. Besides Naïve Bayes classifier, Decission Tree is also commonly used in classification, one of Decission Tree method is C4.5. This study discusses the effect of the use of dummy variables in the womb disease classification using C4.5. From the results of this study concluded that the use of dummy variables to produce an average value accuracy, precission, recall, and F-measure which remained stable at 87.2% in testing k-fold cross validation with  value of k (5, 10, 15, 20, and 25). However, the use of dummy variables reduces the average value of accuracy, precission, recall, and F-measure sequentially from 89.6%, 89.74%, 89.7%, and 89.6% to 87.2%, 87.2%, 87.2% and 87.2%. Besides, the use of dummy variables to specify the attributes of disease symptoms used in the classification of disease womb.

Keywords: Dummy Variable, C4.5 method, Womb Disease.