Tree-Based Classification of AD Risk Using KappaTree

 
   Lead Investigator:    Jeannie-Marie Leoutsakos
   Institution      :    Johns Hopkins School of Medicine
   E-Mail           :    jeannie-marie@jhu.edu
   Proposal ID      :    310



   Proposal Description:
   We propose to use a specific tree-based machine learning algorithm (Kappatree  Petras et al, 2013), to find optimal combinations of baseline predictors of AD conversion among non-demented individuals. Kappatree allows for differential  weighting of sensitivity and specificity we will compare prediction algorithms  when they are equally weighted (which will maximize classification accuracy),   when sensitivity is maximized (for screening purposes) and when specificity is  maximized (for the purpose of identifying individuals to enroll in prevention  trials).