Using Machine Learning to Predict Frailty from Cognitive Assessments.

This paper explores the relation between cognitive and physical aspects of the human body from a machine learning standpoint. We propose to use performance on cognitive assessments to predict frailty of elderly adults with different regression and classification models. We propose a preprocessing scheme with oversampling and imputation to overcome the challenge of an imbalanced data distribution on the existing dataset. We validate the capability of classification models to predict frailty on patients given cognitive input data and provide evidence that machine learning models depend on clinically-defined thresholds.