A proposed hybrid PCA–t-SNE approach for early prediction of Alzheimer’s disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by memory loss and cognitive impairment due to neuronal degeneration. Early-stage detection is essential for enabling timely intervention and slowing disease progression. This study proposes a hybrid dimensionality reduction approach combining Principal Component Analysis (PCA) and t-distributed Stochastic Neighbor Embedding (t-SNE) to enhance early diagnostic accuracy. By analyzing two temporal baselines, the model captures inter-period variations in cognitive health indicators The hybrid PCA-t-SNE workflow, coupled with various machine learning classifiers, depicts significant enhancements relative to conventional approaches. Interestingly, the hybrid classifier combined with a Neural Network presented the best diagnostic accuracy (93.81%), F1-score (93.29%), and ROC AUC (97.81%) in comparison to PCA-only and t-SNE-only models. The Voting Classifier was also strong, with 93.52% accuracy and 96.37% Receiver Operating Characteristic - Area Under Curve (ROC AUC). These empirical findings support the effectiveness of the suggested hybrid framework, which, unlike existing applications of PCA or t-SNE alone, integrates both in a unified pipeline tested across multiple classifiers for early AD detection—providing a novel, interpretable, and computationally efficient alternative to deep or single-method architectures.