ANALISIS PERBANDINGAN ALGORITMA C4.5 DAN REGRESSION TREE DALAM PENDETEKSIAN PENYAKIT TUBERCULOSIS PARU
Daftar Isi:
- The classification technique has been widely applied in many areas of the application to perform a prediction. One of the commonly used classification techniques is decision trees. This technique is widely used because it can display the results of the classification that is easily understood by humans in the form of a tree-like. C4.5 and CART are the algorithms of decision trees that are unique in the formation of trees. Where C4.5 will generate a multi-branch tree and CART will generate a tree with a binary branch. The two algorithms will be compared in the detection of Lung Tuberculosis, which according to the WHO is one of the ten most deadly diseases in the world by 2015, where Lung Tuberculosis is responsible for causing more human deaths than HIV and malaria diseases. There are 102 training data used in decision tree formation and 30 data testing to be tested for classification using established trees. The results of the test show that C4.5 has accuracy and processing time faster than CART that is 86,66% with time required 31ms compared to 73,33% and 78ms when not done pruning. When done pruning the result is 93.33% with time required 32ms for C4.5, while CART 80.00% and 172ms. In the test using crossvalidate technique, used data training as much as 132 data and the amount of fold as much as four fold. The results show the accuracy of C4.5 of 93.93% with a time of 15ms, better than the result of CART accuracy of 90.15% with time 109ms.