A Demand Forcasting Model for the Blood Platelet Supply Chain with Artificial Neural Network Approach and Arima Models

Background and Objectives One of the major issues in global healthcare systems is the issue of improving supply chain performance and uncertainties in demand. The aim of this study is to forecast blood platelet demand with artificial neural network and Arima Models in the blood transfusion supply chain in Sistan and Baluchistan province.
Materials and Methods In this applied study, the data on demand for 8 types of blood platelets were collected from the Zahedan Blood Center between 2011 and 2015. Then, using artificial neural network models and ARIMA models, daily demand forecasts were made. Then, according to MSE performance evaluation criteria, the results of the above-mentioned methods were compared. The data were analyzed by MetlabR2016b and Eviews 6 softwares.
Results The results of this study indicate the high accuracy of neural network models followed by Arima compared to that calculated in the current profile of IBTO. The average accuracy according to MSE of the two models for platelet types are: O (0.0132±0.0048), O- (0.0115 ± 0.0041), A (0.0205 ± 0.0043), A- (0.0108 ± 0.0033), B (0.0221 ± 0.0086), B- (0.0045 ± 0.0009), AB (0.0136 ± 0.0031), AB- (0.0034 ± 0.0005) which represent the mean and standard deviation of the error, respectively.
Conclusions The results of this study indicate the high accuracy of artificial neural network models followd by Arima in predicting blood platelet demand. Therefore, using artificial neural network models for prediction of demand is recommended instead of common statistical prediction methods in blood centers.