Feature-weighted survival learning machine for COPD failure prediction

Abstract

Chronic obstructive pulmonary disease (COPD) yields a high rate of failures such as hospital readmission and death in the United States, Canada and worldwide. COPD failure imposes a significant social and economic burden on society, and predicting such failure is crucial to early intervention and decision-making, making this a very important research issue. Current analysis methods address all risk factors in medical records indiscriminately and therefore generally suffer from ineffectiveness in real applications, mainly because many of these factors relate weakly to prediction. Numerous studies have been done on selecting factors for survival analysis, but their inherent shortcomings render these methods inapplicable for failure prediction in the context of unknown and intricate correlation patterns among risk factors. These difficulties have prompted us to design a new Cox-based learning machine that embeds the feature weighting technique into failure prediction. In order to improve predictive accuracy, we propose two weighting criteria to maximize the area under the ROC curve (AUC) and the concordance index (C-index), respectively. At the same time, we perform a Dirichlet-based regularization on weights, making differences between factor relevance clearly visible while maintaining the model s high predictive ability. The experimental results on real-life COPD data collected from patients hospitalized at the Centre Hospitalier Universitaire de Sherbrooke (CHUS) demonstrate the effectiveness of our learning machine and its great promise in clinical applications.