Classification of Potential Blood Donors Using Artificial Neural Networks and Alternating Least Squares

Blood has an important role in human life. Lack of
blood can cause illness and even death. To meet the needs of
human blood, it requires blood donors from other people. The
problem that occurs is that everyone can only donate blood once
every three months. The problem is that not all blood donors
routinely donate blood every three months. Therefore, it is
important for the Red Cross to be able to classify active and
passive donors, especially when there is an urgent need for blood.
A method is needed that can be used to provide recommendations
for active donors who have a high probability of donating blood so
that they can do so immediately. So that the Red Cross can conduct
the selection of potential donors more effectively and so that there
are not many expired blood stocks. In this study, the classification
of potential donors was carried out using an Artificial Neural
Network and Alternating Least Squares. The results from the list
of potential donor recommendations are then tested using the
Mean Reciprocal Rank. The test results show a result of
0.0052186103057361, with an average time of 127. With the
Artificial Neural Network model that uses two dense layers of 1024
nodes, 10 epochs, and a sigmoid as the activation function, the
training accuracy is 98.90% and the testing accuracy is 99.20%.