Asian Journal of Research in Computer Science

Alzheimer’s disease (AD) is a progressive and incurable neurological disorder that affects both cognitive and motor functions, including fine motor skills such as handwriting. Early diagnosis is crucial to slow the progression and improve patient outcomes, especially as current treatments are largely palliative. In this study, we present a non-invasive and cost-effective approach for early AD detection using handwriting analysis combined with machine learning (ML) techniques. The DARWIN dataset, comprising 174 samples and over 25 kinematic and dynamic handwriting features, was used to train and evaluate classification models. Feature selection was performed using Analysis of Variance (ANOVA) and Recursive Feature Elimination with Cross-Validation (RFECV). Multiple ML classifiers were applied, and their performance was validated using repeated K-Fold and Monte Carlo Cross-Validation strategies. A voting ensemble classifier achieved 100% accuracy with ANOVA-selected features and 88.6% with RFECV-selected features. While these results are promising, the exceptionally high accuracy may indicate potential overfitting due to the limited dataset size, warranting further external validation. This research highlights the potential of handwriting-based screening tools as accessible, low-cost aids for early AD diagnosis in both clinical and remote healthcare environments.