Abstract Comparison of Machine Learning Techniques with Classical Statistical Models in Pre.pdfVIP

Comparison of Machine Learning Techniques with Classical Statistical Models in Predicting Health Outcomes Xiaowei Songa, Arnold Mitnitskib,c, Jafna Coxb, Kenneth Rockwooda,b a Geriatric Medicine Research Unit, QEII Heath Sciences Centre, Canada b Department of Medicine, Dalhousie University, Canada c Faculty of Computer Science, Dalhousie University, Canada Abstract Several machine learning techniques (multilayer and sin- gle layer perceptron, logistic regression, least square lin- ear separation and support vector machines) are applied to calculate the risk of death from two biomedical data sets, one from patient care records, and another from a population survey. Each dataset contained multiple sources of information: history of related symptoms and other illnesses, physical examination findings, laboratory tests, medications (patient records dataset), health atti- tudes, and disabilities in activities of daily living (survey dataset). Each technique showed very good mortality pre- diction in the acute patients data sample (AUC up to 0.89) and fair prediction accuracy for six year mortality (AUC from 0.70 to 0.76) in individuals from epidemiological da- tabase surveys. The results suggest that the nature of data is of primary importance rather than the learning tech- nique. However, the consistently superior performance of the artificial neural network (multi-layer perceptron) indi- cates that nonlinear relationships (which cannot be dis- cerned by linear separation techniques) can provide addi- tional improvement in correctly predicting health out- comes. Keywords: Health status, Machine learning, Classification, ROC curves. Introduction The prediction of health outcomes from available data is an important problem in health research and health management. It is usually assessed by calculating scores/indices for risk stratification [1]. Conventionally, such scores are based on statistical models, such as logis