Neural Networks Approach In Diagnosing Classes Of Anaemia

Hundreds of haematology forms are directed to Haematology unit every day from various departments from physicians that need the right diagnosis in patient’s blood. The
processing may take several days depending on the workload and available resources. A combination of various factors has to be considered before a haematologist can diagnose
classes of anaemia and is normally performed in several stages. The process can actually be performed using neural network approach, as it is capable in pattern recognition.
Knowing the relevant factors that influence anaemia classification, a model of neural network can be produced if it is trained with sufficient data sets. Hence, this thesis presents the neural network model for anaemia classification and identifies parameter that affects its performance using backpropagation. The model is then implemented and the performance of the neural network is assessed. The model was able to diagnose classes of anaemia with 7 1.5 6% generalization. Finally, the model was compared with Radial Basis Function and Regression model to show that Multilayer Perceptron outperforms the
other two models.