Research And Realization Of Key Technology Of Sleep Respiratory Disease Analysis And Decision System

With the increasing development of the Internet,the amount of users and data in all walks of life are showing explosive growth.In the face of more and more data,the traditional database and analysis system can not meet the business needs of the industry Big data framework has become a hot spot in recent years,which not only provides the storage and query mode for both structure and non-structural,but also supports intelligent decision-making large data analysis and intelligent mining technology.In the medical industry,there are not only a large number of traditional structured management data of patients,doctors and the others,but also a large number unstructured data such as diagnostic records,monitoring data,medical images.Therefore,it is essential to find the relative effective big data technologies to achieve multi-dimension statistical analysis to provide the multi-view and direct display of patient information to doctors and to support diagnosis decision.It is also important doctor to use data mining technology to discover potential rules from huge historical cases to achieve intelligent medical care such as the disease cause retrieval,disease modeling and automatic diagnosis.Therefore,under the large data structure,the study of medical data analysis,intelligent computing related technologies provide experience to construct intelligent medical platform experience,which has a good application significance.In order to realize the medical big data analysis platform,this paper uses sleep respiratory disease as the case to achieve the statistical analysis module of OLAP by using the data real-time processing framework Druid as the data warehouse.Spark is used as the analysis engine of data mining module,and the data is clustered and analyzed by using weighted FP-growth algorithm.The main work and innovation are described as follows:1.Design and realization of data modeling and OLAP temporal and spatial analysis for sleep respiratory analysis.Based on the idea of OLAP,the "fact constellation" multidimensional data model suitable for the system transaction is constructed according to the information tables at different dimensions of patients,doctors and diagnostic records to achieve the OLAP temporal and spatial analysis.OLAP drill-up,drill-down query operators based on time dimension and the statistical analysis based on different spatial dimension based on region and latitude and longitude.to study the physiological and disease indicators for cluster analysis.2.The K-means-based patient profile and diagnostic recommendation algorithm are studied and implemented.According to patients with respiratory sleep disorders diagnosis and treatment data,we study the clustering analysis algorithm to physical and disease factors to achieve the disease profile.On this basis,to extract the similar treatment and machine settings during non-invasion therapy with CPAP In this paper,the performance comparison experiment is carried out on several typical clustering algorithms,and the k-means algorithm with balanced performance of accuracy and efficiency is selected.The distance between classes is used as the basis and the K value is determined as 5.3.An important factor mining algorithm based on weighted FP-growth is proposed.According to the improved algorithm,the association analysis of the patient factors is realized,the correlation between the different factors is excavated to assist doctors making the diagnosis decision.Comparing to the traditional Aprior association rules learning method,FP-growth improve the frequent set of mining efficiency based on the tree structure of the itemset storage.Moreover,this paper makes use of doctors' experience as aprior,which provides the corresponding weight according to the importance of different factors in practice.This enhances the ability to describe the association rules.4.Design and implementation of the overall structure of the sleep respiratory analysis system.In addition to the statistical analysis layer and data mining layer,we also realize the system user management module,data import module,data storage module and front-end module.The data management module is mainly based on the Druid data warehouse for real-time data and offline data storage;Front-end module is mainly used in the data processing module is mainly used for the import and export of data;Echart visualization tools and map interfaces is for data display,in the form of line charts,histograms,thermograms etc.