Prediction Of Common Single Drug Therapy For Hypertension Based On Machine Learning

Hypertension is the most common chronic disease and the most important risk factor for cardiovascular and cerebrovascular diseases.For patients with insufficient blood pressure control or unable to achieve it through lifestyle interventions,drug therapy is the most important treatment for blood pressure control.However,due to the complexity of blood pressure control factors in hypertensive patients,it is difficult for doctors to give appropriate personalized prescriptions.To date,few research papers have evaluated the efficacy of antihypertensive drugs from the perspective of machine learning.This paper attempts to use machine learning to predict the efficacy of antihypertensive drugs,thus providing some reference for the use of clinical antihypertensive drugs.The main contents of this article are as follows:First,we collected relevant data from a total of 148,735 hypertensive hospitalized patients from 2008 to 2016 in the Information System of the West China Hospital,including clinical test information and basic information of hypertensive patients.Because the study of this article is a single drug treatment plan,the information of five kinds of commonly used antihypertensive drugs(Irbesartan,Metoprolol,Felodipine,Amlodipine,Levamlodipine)is extracted and analyzed.We selected 17 clinical test features and 2 basic features of hypertensive patients to construct A data set of hypertensive patients is presented.We then drew the characteristic spectrum of five antihypertensive drugs based on the 19 extracted hypertensive patient data sets.Then using logistic regression,support vector machine,random forest and gradient lifting tree four machine learning algorithms,the corresponding five antihypertensive drug prediction models were established,and the performance of the established prediction models was evaluated.The experimental results show that the machine learning method has a high predictive ability for the control effect of these drugs,especially with a relatively high sensitivity.Among them,the sensitivity of the Felodipine drug prediction model based on SVM is 99%,and the sensitivity of other drug models is above 80%.The AUC values of the four machine learning algorithms are all close to 0.8.Finally,we used feature engineering algorithms to select features for the existing key features.On the premise of ensuring the performance of the drug prediction model,the number of features for building the model was reduced from 19 to 7 to make the prediction of the construction The model is more convenient when applied to clinical assistant decision-making.The thesis uses feature selection and feature spectrum visualization to explore the characteristic factors that affect the effect of blood pressure control,and provides some new ideas and references for doctors in the treatment of hypertension.In summary,on this data set,machine learning methods can predict the control effects of five commonly used antihypertensive drugs,and the established prediction model has a high prediction sensitivity,indicating that machine learning algorithms based on clinical big data can be used by assistant A powerful tool for medicine.However,due to the complexity of the formation and development of hypertension and the lack of data sets for patients with hypertension,machine learning models also need to be further improved.