Advancing Medical Predictive Models with Integrated Approaches

This study proposes a thorough methodology for improving the precision and dependability of liver disease prediction utilising cutting-edge methods. The study uses demographic information, blood test results, and machine learning to distinguish between healthy blood donors and people who have liver disorders including cirrhosis, hepatitis C, and fibrosis. To create a reliable prediction model, the study integrates PCA, PSO, & the XGBoost algorithm. Age, sex, and test results of patients are among the information gathered from the University of California Irvine Machine Learning Repository. Categorical variables are numerically encoded while missing values are substituted using column means to maintain data integrity. By using PCA to keep considerable data variance in a lower-dimensional space, dimensionality reduction is achieved. The suggested feature selection strategy combines PSO and Random Forest, with the latter being used to optimise the feature subset that has the greatest impact on prediction accuracy. The population size, iterations, cognitive and social components (C1 and C2), as well as inertia weight (w), are simulation parameters used by the PSO method. The database is categorized into training & testing sets in way to assess the performance of the design. With feature-selected data, the XGBoost algorithm—known for its powerful prediction abilities—is trained. Grid search is used to fine-tune hyperparameters such as maximum depth, learning rate, and number of estimators. The conductance of the design is analyzed by a variety of assessment criteria, like accuracy, precision, sensitivity, specificity, F1 Score. The design’s predictions are further illustrated by a confusion matrix and ROC curve, & the AUC measures the model’s discriminating capability. In order to help doctors, diagnose liver illness early and administer efficient therapy, this research attempts to generate precise and trustworthy forecasts for the condition. The combination of PCA, PSO, and XGBoost gives a methodical approach, illuminating the potential of cutting-edge approaches to improve patient care and medical diagnostics.