Single decision tree classifiers' accuracy on medical data

• Main Authors: Hasan, Md. Rajib, Abu Bakar, Nur Azzah, Siraj, Fadzilah, Sainin, Mohd Shamrie, Sumon, Md. Shariful Hasan
• Format: Conference or Workshop Item
• Language: English
• Published: School of Computing, Universiti Utara Malaysia 2015
• Online Access: http://psasir.upm.edu.my/id/eprint/59080/1/PID188.pdf; http://psasir.upm.edu.my/id/eprint/59080/
• Institution: Universiti Putra Malaysia

Decision tree is one of the classification techniques for classifying sequential decision problems such as those in medical domain. This paper discusses an evaluation study on different single decision tree classifiers. There are various single decision tree classifiers which have been extensively applied in medical decision making; each of these classifies the data with different accuracy rate. Since accuracy is crucial in medical decision making, it is important to identify a classifier with the best accuracy. The study examines the performance of fourteen single decision tree classifiers on three medical data sets, i.e. Wisconsin’s breast cancer data sets, Pima Indian diabetes data sets and hepatitis data sets. All classifiers were trained and tested using WEKA and cross validation. The results revealed that classifiers such as FT, LMT, NB tree, Random Forest and Random Tree are the five best single classifiers as they constantly provide better accuracy in their classifications.