Sina Technology News Beijing time on April 17 morning news, machine learning algorithm predicts the accuracy of heart disease is 7.6% higher than the American College of Cardiology / American Heart Association (ACC / AHA) indicators.
It is estimated that 20 million people die of cardiovascular disease each year. Fortunately, a group of researchers from the University of Nottingham in the UK have developed a machine learning algorithm that can predict the likelihood of a heart attack or stroke like a doctor.
The ACC/AHA has developed a set of guidelines to assess the risk of cardiovascular risk in patients based on eight factors: age, cholesterol levels and blood pressure. On average, the system's valuation accuracy rate can be as high as 72.8%.
Although this is quite accurate, Stephen Weng and his team want to do better. They built four computer learning algorithms and then entered data from 378,256 patients in the UK. The system first uses approximately 295,000 records to generate an internal prediction model. The algorithm is then tested and refined using the remaining records. The calculation results are much higher than the ACC/AHA standard, and the accuracy ranges from 74.5% to 76.4%. After testing, the neural network algorithm has the highest accuracy rate, 7.6 percentage points higher than the existing standard, and the false positive rate has also increased by 1.6%.
The system saved an additional 355 lives in the 83,000 patient records that participated in the test. Interestingly, the artificial intelligence system identified risk factors and predictive values ​​that were not included in existing guidelines, such as severe mental illness and oral doses of corticosteroids. “There are a lot of interactions in biological systems,†Weng told the media. "This is the reality of the human body. Computer science can help us explore these connections."
Shanghai J.Shine Co.,Ltd , https://www.jshinechem.com