Vol 17 no.2 2017

Ibraheem Oluwashola RAHEEM, Abdulrauph Olanrewaju BABATUNDE, Taye Oladele ARO, Abdullahi Yola MUSA

University of Ilorin, Department of Computer Science, Ilorin, Nigeria


The liver is a very important organ in the human body as it carries out vital functions such as clearing of toxins from the blood, metabolizing drugs, makes proteins for blood clotting amongst others. The complexity of the liver makes it easily affected by diseases. Data mining is a method used in the classification of diseases including liver diseases. This study applies three classification algorithms; Naïve Bayes, K-nearest neighbour and decision trees, their bagged and boosted versions, then the algorithms were combined together by ensemble methods of stacking and voting on liver diseases dataset using 10-fold cross validation. Results show that bagging, boosting, voting and stacking the algorithms in the classification of liver diseases do not necessarily increase the classification accuracy, but increase their complexity except the boosted version of Naïve Bayes which shows an increase in classification accuracy when compared with Naïve Bayes. The stacking and voting has a reduced root mean squared error as compared to the other algorithms, while it was observed that C4.5 decision tree algorithm gave the best classification accuracy of all the algorithms used.

Full Text: