Hospital Bed Support System Based on Machine Learning

Recent years, the phenomenon of hospital overcrowding has become more and more severe and occurs in all levels with 2-3 people per bed, which has become an urgent issue for the health system as well as the whole society. One of the solutions to reduce this situation is to arrange beds in a reasonable way. Based on the large amount of data on examination and treatment in hospitals, we propose a solution to use Machine Learning to solve the above problem. A district hospital in Binh Dinh has provided 15066 medical records of 20 different hospitalizations, including information on patients such as year of birth, ethnicity, hometown, diagnosis, date admission and discharge date. With these records, we have built into a dataset and conducted experiments that predict the number of days the patient will be inpatient. The best experimental result is the use of XGBoost Regression, the R2 coefficient is about 0.84.