Research On Recommendation And Curative Effect Prediction Of Hypertension Drugs Based On Spark

Hypertension is the most common cardiovascular disease and it has great harm. It is a major cause of coronary heart disease and myocardial infarction. Hypertension not only has a high morbidity and mortality, but only a serious drain on the country’s medical resources. It is the first cause of death among the top ten risk factors. Hypertension is characterized by life-long medication, effective control of blood pressure is an important factor in the prevention of cardiovascular complications. Currently, the treatment of high blood pressure is mainly controlled by medication, but patients with hypertension awareness, treatment and control rate has been very low, how to improve the treatment and control of hypertension patients had important significance.Development of new technologies of mobile Internet has changed the traditional medical model of hypertension, innovative fusion of traditional blood pressure monitoring equipment and technologies of big data and mobile Internet, through a way that connection between wireless or Bluetooth mobile terminal APP and professional blood pressure monitoring device to help users complete the "measured data- interpret data- data sharing- close to the health" of a complete solution to human health needs of the closed-loop. This approach not only gives users convenience and speed, and has accumulated huge amounts of data. Mining the value of the vast amounts of data has the great significance.To solve the above problem,firstly, this thesis researches the field of hypertension, data mining, and big data field, then proposed a research on recommendation and curative effect prediction of hypertension drugs based on Spark. Specifically, the main work is as follows:(1)To counter the problem of lower hypertension treatment, a recommendation model of hypertension drugs combined of Bayesian and case-based reasoning was researched. Firstly,a lot of similar cases was obtained by case-based reasoning,then a list of drugs was obtained from Bayesian. Case-based reasoning is a common model for recommendation of drugs in patients with a simple, efficient features. However, it lacks ability to deal with a large number of cases, while the probability of Bayesian has rich expression ability. Through the integration of case-based reasoning and the advantages of Bayesian, auxiliary physician decision-making and improve the accuracy of the drug in patients with hypertension.(2) To counter the problem of lower hypertension control, A hierarchical time series clustering algorithm was proposed, by introducing the hierarchical model,effectively solved the problem of low accuracy and high time complexity of the time series clustering model. First of all, hypertension electronic medical records were controlled by hierarchical model threshold, thereafter, the time series curve of blood pressure levels was established for each patient, using a two-stage clustering analysis method, according to the structure of the two curves similarity,multiple fragments with high match degree were divided and calculated the difference of the value of all the matched fragments as the basis for clustering. Finally, drug curative effect was predicted through the time-series curves clustering of patients.(3)To counter the problem of time performance of single machine for mass data processing, a parallel design and implementation based on Spark for the model of hypertension drug recommendation and drug curative effect prediction model were put forward. The parallel Bayesian algorithm and parallel K-Means clustering algorithm were mainly designed and implemented, and through the performance comparison with the single machine and Hadoop, the efficiency of data processing and time performance was greatly improved.The experimental results show that the recommendation model and curative effect prediction model of hypertension drugs have good accuracy and practicability. Through the parallel implementation based on Spark, the execution time and performance of the model algorithm were greatly improved. At the same time, the platform of the hypertension drugs prediction was developed,the recommendation model and the curative effect prediction model were integrated to the system, greatly simplify the operation complexity and help physician make decisions conveniently.