Research On Key Methods Of Data Intelligent Analysis Of Big-health

In the new era,with the continuous enhancement of people's awareness of health management and the yearning for healthy life,as well as the increasing proportion of the aging population,chronic diseases and sub-health population,China is facing severe health challenges.For this reason,general secretary Xi Jinping put forward the strategy of healthy China in the nineteen major reports of October 18,2017,which provided important guarantee for the practice of big health in China.With the promotion of Health China strategy,the focus of health management has gradually shifted to the direction of national health management and disease prevention,but the contradiction between the overall lack of health service supply and the growing demand is still prominent,and the uneven distribution of health care resources still exists and will be maintained for a long time.It is an inevitable choice for health service industry to carry out large-scale health service management and promote health equity by means of science and technology.The means of big science and technology will certainly promote the development of big health.In the 21 st century,the process of comprehensive informatization of health care has been promoted with the development of the Internet.Especially in the second decade of this century,the rise of new technologies such as big data,artificial intelligence and Internet of things has provided a new way for big health services,played an important role in the process of optimizing big health management services,and promoted the transformation of big health service towards refinement,digitalization,intellectualization and scientization,and made it possible to guarantee people's health in an all-round and full cycle way.On June 18,2019,the blue book of artificial intelligence health management in Guangdong Province was released,and the first artificial intelligence health management platform in China was officially launched,which means that the era of intelligent health has come.Intelligent health is not facing patients,but a large number of healthy people.It analyzes individual health status through big data of artificial intelligence,provides accurate health management for individual health,improves individual health management awareness,and creates great value for economic society.This paper analyzes the research status and development trend of big health intelligent service at home and abroad based on the background of big health,closely around the two research objects of big health service platform and health data,launches the basic research of big health intelligent analysis,and promotes the deep integration of emerging technologies such as artificial intelligence and big health.The specific research contents are as follows:1.To overcome the serious problems of high retrieval response delay and poor accuracy in the whole research of health data,a large-scale health information resource integration architecture based on hybrid cloud computing is proposed.Based on the general mode of health information service platform,a large-scale health information resource integration architecture is constructed by using the cloud architecture idea.At the same time,through the sample specification and dimension specification,the large health information data is regulated,and the resource space is optimized.The field matching method is used to clean the data resources,and the matching degree is used to judge whether it is a redundant matching field.According to the weight value of data resources obtained by hybrid cloud computing,the data of big health information is arranged to realize the classification and integration of big data information.The simulation results show that the proposed method has good performance,low data redundancy,low retrieval response delay,high classification and integration accuracy,and is feasible.2.Aiming at the problem of dynamic load imbalance of health data flow in large-scale health service system,the processing ability of traditional methods is limited to the window range of a certain operator's node,which is obviously insufficient when the data volume is gradually increasing,and the situation of data flow congestion is easy to occur,and the research of overall load distribution and Migration Decision-making in dynamic load balancing is ignored.Therefore,a new dynamic load balancing method for health data flow in parallel computing environment is proposed.We can get node corresponding data block with using the Hash of tuple key,then record the target node of response with the data block and output data tuple.At the same time,the entropy of parallel computing is extended to heterogeneous clusters and solved.As a measure of the load balance degree of healthy data flow,it can judge whether load migration is needed,determine the migration task mode and migration amount,and then make migration decision.The simulation results show that the method is feasible and has good calculation performance and dynamic load balancing.3.According to the problems of apriori,parameter setting and randomization in the existing baseline solution methods of health sign data(such as ECG,EEG,pulse,etc.),a weighted matrix regression algorithm is proposed.According to the cut-off frequency of the corresponding sign signal under the different regularization parameters,combined with the range of the baseline frequency of the corresponding sign data,the regularization parameter value is determined,and the effect of effectively and quickly calculating the drift baseline contained in the corresponding sign data is achieved.Meanwhile,the effective original periodic signal is retained.Finally,through the experiment of PPG signal and simulation signal,the results show that the signal drift in the time domain and the drift baseline term in the frequency domain can be separated effectively.4.In view of the analysis of the interaction of multiple health variables(blood pressure,heart rate,respiration,etc.)in physiological control system during postural change in disease and health state,a P-SLD model combining physiological control model with switching linear dynamics is proposed.In this model,the improved semi supervised machine learning EM algorithm is applied to solve the posterior probability estimate of the latent handover variables and the maximum likelihood set of the model parameters.In the improvement of EM algorithm,the penalized least squares factor is introduced into the maximum likelihood function,and the nonnegative constraint is introduced as a priori information,combined with semi supervised machine learning method.Finally,the improved problem is transformed to the minimum solution problem.Finally,the unsteady health data of mean arterial pressure and heart rate in ICU patients and the blood pressure and heart rate of healthy people with different posture(supine,non supine)were choose and tested.It is verified that the model can be used to reveal the changes associated with severe systemic response syndrome(SIRS)by analyzing the transfer function and power spectrum of p-sld model.It is also verified that the modified model can automatically capture the effect of postural changes on the pressure reflex gain,providing a hypothesis for the automatic regulation of physiological control under health and disease conditions.5.For one-dimensional health signs,time series data are time-varying,nonlinear,non-stationary and so on.As a result,the existing single and mixed forecasting models are difficult to describe their complex change rules.A model for predicting time series data of health signs by using deep reinforcement learning method is proposed.At the same time,Saras learning method is introduced into the algorithm,which effectively improves the stability,effectiveness and accuracy of the model Experimental verification of historical blood pressure data shows that the model improves the accuracy,stability and convergence speed of the prediction model,and can fully express the trend of one-dimensional health sign time series data.