Evaluating Dimensionality Reduction Techniques for Liver Disease Classification Using Unlabeled Data

Liver disease poses a significant health burden worldwide, necessitating early and accurate diagnosis to improve patient outcomes. However, real-world clinical datasets often contain high-dimensional and unlabeled features, complicating traditional classification approaches. This study investigates the effectiveness of dimensionality reduction techniques in enhancing liver disease classification performance using unsupervised data. We explore and compare principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), and uniform manifold approximation and projection (UMAP) on a medical dataset containing physiological and biochemical attributes. Each technique is evaluated based on its ability to preserve the intrinsic data structure while improving downstream clustering and classification accuracy. The findings reveal that nonlinear methods, such as t-SNE and UMAP, offer superior separability of liver disease indicators in reduced-dimensional space compared to linear approaches, like PCA. This work highlights the potential of combining unsupervised learning and dimensionality reduction for efficient feature extraction in medical diagnostics, especially where labeled data are limited or unavailable.