Prediction of Demand for Red Blood Cells Using Artificial Intelligence Methods

Authors : Seda Hatice GÖKLER, Semra BORAN

Pages : 86-93

Doi:10.21541/apjess.1078920

View : 12 | Download : 5

Publication Date : 2022-05-01

Article Type : Research Paper

Abstract :Blood is a vital product with limited resources, available only from volunteers. For this reason, the blood components to be sent from the blood bank to the transfusion centers insert ignore into journalissuearticles values(hospitals); should be accurately predicted. There are many variables that affect the demand prediction. In this study, fifteen different qualitative and quantitative variables were determined. Artificial intelligence insert ignore into journalissuearticles values(AI); methods are used because the prediction has nonlinear, complex and uncertain relationships and thus it is also difficult to mathematically express on relationship in between input and output variables. AI methods have the feature of predicting the information that is not given or that may occur in the future by learning the past data. In the study, AI methods such as Decision Tree insert ignore into journalissuearticles values(DT);, Support Vector Machine insert ignore into journalissuearticles values(SVM);, Artificial Neural Network insert ignore into journalissuearticles values(ANN); and Deep Learning insert ignore into journalissuearticles values(DL); were applied to blood bank providing blood supply to public and private hospitals operating in four provinces. The data obtained from the prediction results of AI methods were compared with performance criteria insert ignore into journalissuearticles values(MAPE, MSE, MAE RMSE and R2); and values of overprediction, underprediction, minimum and maximum deviation. The weekly average over predictions are calculated as 9.69, 5.29, 8.45, and 15.65 and weekly average underpredictions as 17.57, 3.03, 3.94, and 14.69 for DT, SVM, ANN, and DL methods, respectively. SVM method was determined as giving the best prediction values. Therefore, it is envisaged that the blood component demand prediction can be calculated using the SVM method.
Keywords : Demand prediction, decision tree, support vector machine, artificial neural network, deep learning