Alzheimer’s disease is a chronic, prolonged, and irreversible neurodegenerative disease of unknown cause. In recent years growing research interest assumes that by processing data of essential factors effective models can be defined for recognizing and predicting the disease development. The present article aims to propose classification models for the diagnosis of Alzheimer’s disease cognitive states. For this aim medical data of biomarkers and cognitive assessment data are used. The novelty of the paper is to explore both the Amyloid/TAU/ Neurodegeneration framework and the biologically determined process of delay between the brain impairment and visibility of its appearances by incorporating these concepts in the model development procedure. The study explores the ability of three classifiers – Random Forest, Extreme Gradient Boosting, and Logistic Regression. Conclusion results have been done by comparison of the grouping abilities in different data spaces. The practical result of the study is helping to determine medical examinations that give accurate results for the diagnosis and prediction of the progression of the disease in possible earlier stages of the disease development.