Heart failure (HF), the terminal stage of all kinds of cardiovascular disease, has a high level of morbidity and mortality. But the heavy burden of curing and managing HF can be largely reduced by early detection of it. Motivated by this problem, we study methods to determine its risk factors based on extreme learning machine (ELM). Several state-of-the-art data mining algorithms are employed to estimate the performance of various classification of the HF patients. Our data sets are extracted from reality hospital patients' data, which consist of patients' basic demographic, disease and assay information. The results show that ELM will have a better performance larger than 93% if the selected attributes have more strong correlation with the label. © 2016 IEEE.
(8.0+极速模式)