A model to predict the sequential behavior of healthy blood donors using data mining

Urgent need for blood and lack of alternative to replace it, necessitate the presence of a system for predicting continuous behavior of healthy blood donors in blood transfusion organization. Predicting donors’ behavior aims to determine blood unit number and blood groups for providing future contingency blood bank; and also restraining the dangerous bullwhip effect on blood supply chain. This cross-sectional study was conducted by using census. The study population consisted of blood donors’ data, which was collected from blood transfusion organization. Clementine software version 12.0 was used to analyze the data. Four different data mining algorithms including a decision tree algorithm that it used as data mining technique (C5.0), C R Tree, CHAID and QUEST were employed for analyzing the data and knowledge. Results from data mining were respectively: 57.49%, 55.9%, 55.56% and 55.34%. High accuracy of C5.0 indicates the better performance for this algorithm. According to measurement accuracy of training and test data, the reliability score for algorithm of C5.0 was 0.981. Clustering method was used to verify the accuracy of the best model. C5.0 algorithm, distributs the continuous participants blood among clusters based on variables such as age, gender and marital status by using clustering. Repeated participantss were assigned to a cluster and the first time participantss were assigned between two clusters based on gender or marital status variables. Using this model helps to predict contingency blood bank faster with more accuracy.