Research On Hospital Department Performance Management Methods Based On Pattern Recognition

With the development of economy and society, medical and health services are becomingimportant public utilities, which are related to people’s livelihood. Modern public hospitalsencountered challenges and opportunities in every aspect with the reform of the health caremechanisms and a rapidly changing of external environment. However, due to the differencesin structures, technologies, capital and operating philosophy of these medical institutions,some of the uncompetitive have troubles in development, even unsustainable.When it comes to some predictive decisions, the hospitals usually convene experts to voteon. This method is simple but has many drawbacks. In the case of relation tickets, the hospitalmay make wrong decisions, which will cause some damages and affect the hospital’sdevelopment and survival.So if the hospitals can pay more attention to the daily management, they will be moreconducive to the development of the hospital. If the tendency of important data can bepredicted, the work can be arranged reasonably and the hospitals can make right decisions.Neural network algorithms (NNA) have been very mature in forecasting area. Manyscholars applied NNA to prediction experiments in various fields. However, the result is notsatisfactory when NNA is used alone. Fruit Fly Optimization Algorithm (FOA) is similar tothe genetic algorithm (GA), but it is simpler, more understandable and workable than GA.FOA has a superior character for global optimization features. Therefore, this paper choosesFOA to optimize NNA and predicts various indicators in hospital.Based on the two ideas above, this paper uses the radial basis function neural networkalgorithm (RBF) and Fruit Fly Optimization neural network optimization (FOA-RBF)algorithm to build two mathematical models respectively, and then predicts four importantindexes (outpatient, average length of stay, bed usage and discharges) of one hospital inChangchun. It uses the first24months of data as training samples, and uses the last6monthsof data as test samples. The result can be obtained that the FOA-RBF algorithm has a better performance and a smaller average error and mean relative error.This paper also makes classification and evaluation of every department by the fourimportant indexes using the Extension Theory-Neural Networks algorithms, which aredivided into excellent, good, fair and poor four grades. Department performance can beevaluated by the result of classification and evaluation. The simulation experiment shows thatthe method can evaluate every department’s month performance well and the predicted datacan be classified easily with a high accuracy. Hospitals can use the result to do bonusesassessment of the department’s doctors, nurses and other staff. This approach will encouragemedical staff’s working attitude and enthusiasm withbonus and forfeit.