The Prediction of the Ratio of Current Health Expenditures to GDP with Machine Learning Regression Methods

Health expenditures, which are one of the main areas of health economics, are among the topics that have been discussed recently. The increase in health expenditures puts pressure on the public budget of many countries. In addition, health expenditures increase the health level of society and accelerate development with their contributions to human capital. In this study, Turkey’s current health expenditure (% of gross domestic product (GDP)) was estimated using multiple linear regression, artificial neural networks, random forest (RF) and support vector regression (SVR) methods. In the study, the ratio of current health expenditure to GDP is the dependent variable, life expectancy at birth, the population aged 65 and over (% of the total population), the number of physicians per 1,000 people, the infant mortality rate and the number of hospital beds per 1,000 people were considered as independent variables. 75% of the data set was reserved for training and 25% for test data. The performance of the regression models obtained was compared by calculating the coefficient of determination, mean squared error, root mean square error, mean absolute error and mean absolute percent error, which are goodness-of-fit criteria. Results show that the radial-based SVR and RF models had better prediction results for the entire data set.