Abstract
In high-dimensional data analysis, unsupervised feature selection plays a crucial role in enhancing model interpretability and reducing computational cost. While Principal Component Analysis (PCA) and Multi-Criteria Decision-Making (MCDM) methods such as MOORA have individually been employed for dimensionality reduction and feature evaluation, their combined use remains underexplored in the context of unsupervised feature selection. This study proposes a structured hybrid approach that integrates PCA for extracting dominant components and MOORA for ranking original features based on their alignment with those components. Unlike traditional methods that rely on a single criterion or lack interpretability, our fusion method incorporates multiple decision metrics in a unified framework. The proposed approach is evaluated on both bioinformatics datasets and diverse real-world applications, demonstrating consistent improvements in classification accuracy and feature reduction compared to standalone PCA, MOORA, and other baseline techniques. These results suggest that the synergy between PCA and MCDM can provide a more robust and generalizable strategy for unsupervised feature selection across domains.