Extending the Progression Model of Alzheimer's Disease Combining KNN, LDA and SMOTE

Alzheimer's disease (AD) is a challenging condition that significantly impacts quality of life and leads to cognitive impairment. As the most prevalent form of dementia, AD causes severe damage to nerve cells. Individuals affected by AD struggle with cognitive functions such as memory, thinking, and behaviour, which in turn disrupts their daily activities. Current methods for creating medical concept structures from various data sources— such as ICU and outpatient clinical records—often fall short in capturing the full complexity of disease progression. The primary issues include limitations in attention weight learning and restricted handling of data sources. To address these challenges, this study introduces a novel approach that combines three key techniques: Linear Discriminant Analysis (LDA) for reducing dimensionality, the (SMOTE) Synthetic Minority Over-sampling Technique for enhancing data through augmentation and addressing class imbalance, and K-Nearest Neighbours (KNN) for classification purposes. The proposed method was empirically tested and achieved a high accuracy rate of 98.6%, demonstrating its effectiveness in representing AD progression more comprehensively.