Implemented a comprehensive approach for feature engineering in chronic liver disease detection by combining PCA, replication dependence using mixed-effect models, Fourier and wavelet transforms, and standard scaling normalization for replication-aware feature sets. In 10-fold cross-validation, an MLP model achieved a 95% accuracy, outperforming other methods. In a train-test split, both Random Forest (RF) and MLP achieved perfect 100% accuracies. This methodology significantly enhanced the model performance and demonstrated robust results in comparison to current state-of-the-art techniques

Applied a combination of linear and non-linear dimensionality reduction techniques (LDA, FA, t-SNE, and UMAP) to form a comprehensive feature set. Conducted train-test split and 10-fold cross-validation evaluation for Logistic Regression, KNN, Random Forest, and MLP, which improved the accuracy (98.31% for RF) of the models, outperforming recent state-of-the-art studies.