Marc Claesen

A robust ensemble approach to learn from positive and unlabeled data using SVM base models

We present a novel approach to learn binary classifiers when only positive and unlabeled instances are available (PU learning). This problem is routinely cast as a supervised task with label noise in the negative set. We use an ensemble of SVM models trained on bootstrap resamples of the training data for increased robustness against label noise. The approach can be considered in a bagging framework which provides an intuitive explanation for its mechanics in a semi-supervised setting. We compared our method to state-of-the-art approaches in simulations using multiple public benchmark data sets. The included benchmark comprises three settings with increasing label noise: (i) fully supervised, (ii) PU learning and (iii) PU learning with false positives. Our approach shows a marginal improvement over existing methods in the second setting and a significant improvement in the third.

Mortality in Individuals Treated With Glucose-Lowering Agents: A Large, Controlled Cohort Study

CONTEXT Several observational studies and meta-analyses have reported increased mortality of patients taking sulfonylurea and insulin. The impact of patient profiles and concomitant therapies often remains unclear. OBJECTIVE The objective of the study was to quantify survival of patients after starting glucose-lowering agents (GLAs) and compare it with control subjects, matched for risk profiles and concomitant therapies. DESIGN This was a retrospective, controlled, cohort study. SETTING The study is based on health expenditure records of the largest Belgian health mutual insurer, covering more than 4.4 million people. PATIENTS A total of 115 896 patients starting metformin, sulfonylurea, or insulin (alone or in combination) between January 2003 and December 2007 participated in the study. Control subjects without GLA therapy were matched for age, gender, history of cardiovascular events, and therapy with antihypertensives, statins and blood platelet aggregation inhibitors. INTERVENTION(S) There were no interventions. MAIN OUTCOME MEASURE Five-year survival after the start of GLA was measured. RESULTS Profiles of patients using different GLAs varied, with patients on sulfonylurea being oldest and patients on insulin having more frequently a history of cardiovascular disease. Excess mortality differed across GLA therapies compared with matched controls without GLAs, even after adjusting for observable characteristics. Only metformin monotherapy was not associated with an increased 5-year mortality compared with matched controls, whereas individuals on a combination of sulfonylurea and insulin had the highest mortality risks. Age and concomitant use of statins strongly affect survival. CONCLUSIONS Differences exist in 5-year survival of patients on GLA, at least partly driven by the risk profile of the individuals themselves. Metformin use was associated with lowest 5-year mortality risk and statins dramatically lowered 5-year mortality throughout all cohorts.