Forecasting Reasons for Students’ Health Failure in Tertiary Institutions

Abstract

One of the prime requirement of healthcare administration is to bring excellence service to patients, by making comprehensive conclusions. Such conclusions can only be made with the existence of satisfactory knowledge de rivative from healthcare data that cannot be acquired by simple observation. The usage of data mining is very supportive in health care administration for forecast and conclusion making, since it is a convergence of numerous disci plines, which comprises database management systems (DBMS), Statistics, Ar tificial Intelligence, and Machine Learning. In this study, Decision tree tech nique was used to forecast the periodic causes of students’ health failure. Four dissimilar decision tree models were articulated, which comprise the J4.8, Ran dom Forest, Random Tree and Decision Stump. This was attained by perform ing a 10-fold cross validation on a dataset containing of seven nominal varia bles and ninety instances. It was detected that the J48, which is the Weka’s im plementation of the C4.5 decision tree model performed better with 76.6% ac curacy, 0.829 precision and 0.708 recall than the remaining three mode