Prediction and Classification of Patients Length of Stay in a Medical Hospital in Birnin Kebbi, Kebbi State, Northwestern Nigeria

Abstract

The objective of this paper was to predict and classify the attributes that influences patients Length of Stay (LOS) in a Medical Hospital in Birnin Kebbi, Kebbi State, Northwestern Nigeria using tree-based machine learning algorithms after data collection. When training the modes, random forest achieved an R-squared value of 0.573541 using a continuous response and classification rate of about 87% using categorical response variable. In testing the performance of the top identified modes, random forest mode had an accuracy of 72%. Linear regression model was also used in predicting patient’s length of stay. Tree based models performs better than the linear regression model. The result shows that random forest outperforms decision tree and boosted tree in predicting and classifying patient LOS.

Keywords

• Length of Stay, Prediction, Classification, Decision Tree, Boosted Tree, Random Forest