Improvements Combination Of Gray Models And Artificial Neural Network-based Prediction Model And Its Application

With the rapid development of information technology, the application scale, scope and depth of database management information system is continuously expanding and deepening.More and more databases have been accumulated in the system, but the present existing database management information systems basically perform the traditional data management functions, traditional data inquiries, statistic operations or fixed calculations about mathematical indicators. The data resources have not been fully utilized, resulting in a great waste of the resources, which can't meet the demand of the decision-making management staff who need data support. As the current competition between the hospitals is becoming increasingly fierce, it is particularly important for the hospital to improve their own competitive advantages in the society. The survival and development of hospitals depends on scientific and effective management. Consequently, the management personnel not only need the traditional statistical results, but also urgently need to do some scientific predictions about the indicators so that they can make scientific and effective arrangements for the next step—preparing materials, assigning members reasonably, in order to achieve the aim of saving resources and improving the comprehensive competitive edge of the hospitals.The hospital management information system has accumulated a large quantity of time-related time sequence data on outpatients, inpatients, drug uses and so on. At present, research results on the prediction of such data have been achieved both at home and abroad, among which the grey model and the artificial neural network model are comparably widely utilized. These two models employ different algorithms to realize the prediction functions of the future data based on historical data, both of which can get better accuracy of data prediction and have respective advantages. The grey model possesses the qualities of less needed sample data, simple principle, convenient calculation, and high precision of short-term prediction, while the artificial neutral network model possesses the qualities of parallel calculation, various functions imitation, good adaptability and learning ability, and strong tolerance of errors, etc. However, as a single prediction model, any of them has its defects. For instance, the grey model has the defects of unfavorable adaptability to the volatility of data, highly dependence on historical data, etc. Meanwhile, the artificial neural network model has the defects of high requirement on sample volume, great influence of personal experience on prediction results, unstability of network learning and memory, etc. It's hard for a single model to solve its own problems.This paper studies the usage of GM (1,1) , analyzes its defects, and then carries on the improvement of GM (1,1). With the improved GM (1,1), the paper predicts the monthly outpatients of a certain hospital, through which its shortcomings are analyzed. In order to enhance the precision of data prediction, assimilate the merits of both models, the paper proposes an improved combined prediction model. First fit the historical data by using improved grey model; then, form the residual sequence data according to the margins produced by the historical data and the fitted data; next, modify the residual sequences by using the artificial neural network model; finally combine the basic data generated by the improved grey model with the revised residual sequence, and obtain outcomes predicted by combined model. The MSE of prediction results produced by improved grey model, artificial neural network model and combined model are respectively 13465, 14235, 10548; the MAPE are respectively 8.09%, 9.60%, 6.27%. According to the error index of the prediction results, the combined prediction model has overcome the defects of one single model to a great extent, and further improved the accuracy of data prediction.