Prediction of hospitalization cost for day surgery patients based on machine learning
Objective To explore and compare the effect of the three forecasting models(extreme gradient boosting,back propagation neural network and support vector machine)on the hospitalization expense of day surgery,and to put forward suggestions on how to effectively control allocation of medical resources.Methods A total of 9 064 pieces of data from January 1,2018 to August 31,2021 were collected from the hospital information system.Excel was used to establish a database,and make a descriptive analysis by SPSS 21.0.Python was used to conduct models fitting for the hospitalization expense of day surgery.Select the best model for exactly forecasting the hospitalization expense of day surgery by comparing the evaluation indi-cators.Results The results showed that the median of hospitalization expense is 2 872.11.The coefficient of determination(R2)achieved 0.854 and the mean absolute percentage error(MAPE)was 0.209 when the extreme gradient boosting was used to predict the hospitalization expense.R2 achieved 0.837 and MAPE was 0.240 by using the back propagation neural network.R2achieved 0.730 and MAPE was 0.225 by using the support vector machine.The extreme gradient boosting performed better than the other methods by comparing the evaluation indicators.Conclusion Compared with the back propagation neural network and support vector machine,the extreme gradient boosting has more advantages in predicting the hospitalization expense of day surgery patients,which has higher estimation precision and reliability.The accurate prediction of hospitalization expense can pro-vide decision-making reference for relevant medical operation managers,and control the expense actively under the condition of ensuring medical quality,guiding the medical behavior and improving the efficiency of the use of hospital resources.
Day surgeryHospitalization expenseExtreme gradient boostingBack propagation neural networkSup-port vector machine
万方数据
维普
