Hospital Revenue Forecast Using Multivariate and Univariate Long Short-Term Memories

Currently, most hospitals in Vietnam have Health Insurance services. Therefore, the hospital’s revenue includes the actual revenue from patients and the revenue paid by Health Insurance companies. Hospital revenue forecasting is essential in management, so that hospital leaders can make policies, plan advances, and make appropriate decisions. The hospital revenue forecast is a complex problem. The considered hospital revenue in this study is Health Insurance Company payment. This study mainly analyzes approaches to select neural network models in deep learning for forecast on two datasets of hospital revenue on the Health Information System of Vietnam Posts and Telecommunications Group (VNPT-HIS), recorded daily from 2018 to March 2021 in a provincial hospital of Ca Mau Province. Dataset 1 includes all revenue values recorded daily with 1182 records and an average value of 218 million VND. Dataset 2 does not include revenue values recorded on weekends or particular days with 982 records and an average value of 245 million VND. We adapt the models with both datasets by using different test sets for comparing the prediction performance. The empirical results show that the proposed method achieves positive results in both datasets. The models could produce acceptable prediction results. Therefore, the system could support the hospital manager in financial management activities.