Liver Disorder Detection Using Fractional Brown-Bear Optimization-Based Feature Fusion and Ensemble Learning

In the last few decades, liver diseases have quickly augmented in prevalence and severity, making them one of leading global killers. Any abnormality in liver is defined as liver disease. Diagnosis and evaluation of liver diseases constitute the process of identifying liver disorders as generally; the cases are the worst where there are already signs and symptoms. A significance of the research focuses on Machine Learning (ML) and Deep Learning (DL), which are progressively more included into society and applicable in nearly all situations, with even more potential in future scenarios. This research develops a method for prediction of liver diseases using optimization-enhanced feature fusion and ensemble learning. Initially, preprocessing is done using data standardization to transform the data into a form in order to maintain consistency and enable comparability. After standardization, a bootstrapping-based oversampling method is applied to address class imbalances. Then, feature fusion is done using Minkowski distance-based feature sorting and Fractional Brown-Bear Optimization Algorithm (FrBBOA), where FrBBOA is proposed by integrating Fractional Calculus in Brown-Bear Optimization Algorithm (BOA). Finally, liver disease prediction is performed using proposed FrBBOA-based 1D Convolutional Neural Network–Recurrent Neural Network–Multilayer Perceptron (FrBBOA-based 1D CNN-RNN-MLP). This model integrates 1D CNN, RNN and MLP, and is trained using proposed FrBBOA. Results for the proposed FrBBOA-based 1D CNN-RNN-MLP demonstrated better performance than conventional approaches, as confirmed by its 97.3% accuracy, 98.1% sensitivity and 96.2% specificity.