Converse or reverse? Machine-learning modeling for disease progression: A study based on Alzheimer’s disease continuum cohort