PUBLISHED RESEARCHES IN THE YEAR 2010 VOLUME 4 No.1 Title: An application of Decision Tree Algorithms with Diagnosis of the Diseases of the Respiratory System ∗ Researchers: Ditapol , † Mantuam Lily Ingsrisawang† ∗ Corresponding author, e-mail:
[email protected] † Department of Statistics, Faculty of Science, Kasetsart University Abstract: The Knowledge Discovery in Database (KDD) has been extensively used through “data mining”, a statistics and computer sciences, to organize the crucial useful data into knowledge base form for further research purpose. The data classification is a techniques applied by the KDD to various fields and medical research. The primary purpose of this study was aimed to applications and compare the performance of the 3 decision-tree algorithms, including ID3, C4.5, and CART, which have currently become famous in sorting data. The results would be expected to support the screening process and to be used as guidelines for primary diagnosis. The medical record of 7,327 out-patients at the Pranakorn Sri Ayutthaya Hospital during 2004-2006 was examined. The results have demonstrated that algorithm C4.5 which percentage split method was used to divide the data into 70:30 was 99.41% accurate in respect of no selection of variables. The Kappa was 0.9881. Sensitivity was 99.31% while specificity 99.50%. Positive predictive value (PPV) was 99.40% while negative predictive value (NPV) was 99.41, ROC area was 99.70%, regarded as the most effective classifier. On the other hand, it found that algorithm ID3 which percentage split method was used to divide the data into 70:30 was 95.85% accurate in case of selection for variables.