Multilayer Perceptron: Architecture Optimizationfor Classifying Anemia Patients

This research extends the work done in an earlier study in 2022, wherein the authors used multiple data sets for data analysis in the task of classifying anemia patients. This study uses the same four data sets in the data nalysis as used and defined in this study. This research investigates the computational performance of various configurations of a multi -layer perceptron architecture for classifying anemia patients. The data analysis has been done with four data sets. The performance of various setups of a multi-layer perceptron neural network is rigorously assessed by analyzing key metrics such as accuracy, precision, recall, and computation time. These evaluations have been conducted using a 10-fold cross-validation framework. The configurations of the MLP architecture are defined by the parameters of Hidden layers, Momentum, Learning rate and Number of neurons in the hidden layer. The computations find that among the configurations of MLP neural networks, used in the data analysis for classifying anemia patients, the best network configuration represented the data set obtained by applying Synthetic Minority Oversampling Technique (SMOTE) on the original input data set in which the study finds the parameters with the highest accuracy of 97.526, precision = 0.965,time = 2.13(seconds) and recall = 0.961 with two hidden layers and 15 neurons in the first and 5 neurons in the second hidden layers in the MLP configuration. The computations have been done using the WEKA tool.