Early Alzheimer's Detection Using Weighted Correlation-Based Feature Selection (Weighted Cfs)

Abstract

Feature selection plays a pivotal role in improving the accuracy and interpretability of machine learning models, particularly in healthcare applications like Alzheimer’s disease detection. In this study, we present a Weighted Correlation-based Feature Selection (Weighted CFS) approach for early Alzheimer’s detection. Weighted CFS extends the traditional CFS method by incorporating domain-specific weights to prioritize clinically significant features while balancing feature relevance and redundancy. This methodology effectively integrates expert knowledge with statistical correlations, making it ideal for handling high-dimensional and complex datasets. By assigning higher importance to features such as cognitive test scores (e.g., MMSE) and demographic factors, the approach ensures that selected features are both domain-relevant and statistically informative. Experimental results demonstrate that Weighted CFS improves model performance, reduces feature redundancy, and enhances interpretability, offering a robust framework for identifying critical predictors in early Alzheimer’s detection.