Numerical Investigation on β  = 3/2 in the Generalized Multiquadric Neural Networks for Classification Challenge

Neural networks are an important tool in medical classification tasks that can help healthcare providers make more accurate diagnoses, develop personalized treatment plans, and improve patient outcomes. This study aimed to investigate the effectiveness of the value of beta = 3/2 in a different data structure under the architecture of a neural network, following its success in a previous data structure. To achieve this, a neural network was constructed using the multiquadric radial basis function (MQ-RBF) and two other popular values of beta were included for comparison. The hepatitis C-virus dataset and the breast cancer dataset were used for classification and the accuracy of the classification was evaluated using various metrics and validation techniques. The study found that selecting beta as 3/2 did not yield the expected results, unlike the other two choices which provided dependable confidence. Moreover, the study identified a problem with the interpolation matrix being non-invertible, which created another challenge for analysis. These findings demonstrate that the selection of beta values requires careful consideration and should not be assumed to perform equally well across different data structures. Further research is needed to explore alternative approaches to improve classification performance in challenging datasets.