The current paper introduces an unsupervised data mining methodology applied to medical data to uncover insights into the cognitive states associated with Alzheimer’s disease. The methodology emphasizes the integration of algorithmic feature selection with existing knowledge about the disease’s manifestation. The data clustering analysis is conducted with two primary objectives: first, to uncover the underlying structure of the data, and second, to examine how well the identified cognitive groups align with existing diagnoses. The results demonstrate that the most effective clustering approach identifies three distinct cognitive groups. However, while each cluster represents a dominant cognitive group, it does not strongly differentiate between the subjects’ diagnoses.