Prediction by Sigmoid Multilayer Perceptron Artificial Neural Network Function and Model Selection for the Risk Factors Most Affected Human Immunodeficiency

The Multilayer Perceptron (MLP), a widely recognized type of Artificial Neural Network (ANN), was applied in this study to forecast the risk factors associated with human immunodeficiency conditions. A sample of 500 patients with various diseases was collected from hospitals and laboratories in the Kurdistan region. Each patient’s immune level was tested, and the dataset included one dependent variable, immune testing level (classified as either "good immunity" or "poor immunity"), and six independent variables representing potential risk factors (X1 to X6). Statistical analyses, including parameter estimation and variable importance ranking, revealed that X1: Genetic history had the most significant influence on immunity, followed by X5: Cancer treatments such as radiation therapy, X4: AIDS, X2: Diabetes, X3: Human Immunodeficiency Virus (HIV), and lastly, X6: Certain medications. Model selection criteria such as the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC), along with the Likelihood Ratio Test and Chi-square test (p-value = 0.011 < 0.05), a that these risk factors significantly affect immune deficiency outcomes. The results validate the effectiveness of the MLP model in identifying the most influential predictors of immunodeficiency.