The Data Of Non - Traumatic Emergency Chest Pain Were Predicted By Data Mining

Objective:patients with chest pain in cardiovascular disease specialist hospital emergency department exists obvious characteristics, such as multiple quantities, numerous pathogeny and critical level disparity. It’s a challenging thing for outpatient and emergency young physician to quickly and precisely identify disease of patients with chest pain. Through big data, data mining and so on the novel method, this study attempts to build a classification and predictive model which can be based on patient disease information to determine the type of disease, to assist physician obtain rapid, efficient diagnostic results and reduce missed diagnosis and misdiagnosis.Methods:By a variety of classification in data mining algorithm to build the prediction model, firstly, using the CART decision tree calculates the probability of each disease in patients with chest pain. Then, making use of six classification algorithm, including the nearest neighbor method, C5.0 decision tree and support vector machine(SUV), artificial neural network, naive Bayes and random forest, evaluates each disease to determine whether the patients have the disease.Results:In the diagnosis of the coronary heart disease model, artificial neural network has the best performance, the accuracy rate is 93.75%, AUC value is 0.849; SVM has the best performance of subdividing coronary heart disease into myocardial infarction and angina model, accurate rate is 82.35%, AUC value is 0.826; In aortic dissection and other diseases diagnosed models, artificial neural network has the best performance, accuracy rates is 96.30% and 89.36% respectively, AUC values is 0.924 and 0.867 respectively; predicting whether serious adverse events of patients discharged from hospital after acute myocardial infarction will occur or not, random forests in 30 days, half a year and a year has the most reliable prediction, accuracy rate is 94.36%,92.63% and 92.72% respectively, each AUC values is above 0.9.Conclusion:each models of optimum algorithm classified prediction accuracy is above 82%, and patients discharged from hospital after acute myocardial infarction serious adverse events prediction is more than 92%. It is proved that classification and prediction model built by means of data mining can accurately classify patients with heart chest pain in vascular disease specialist hospital emergency departments, providing some assistance for the medical staff and reducing missed diagnosis and misdiagnosis phenomenon.