Research On Spatio-Temporal Clustering Algorithm For Infectious Disease Surveillance

Infectious disease surveillance is the prerequisite and guarantee for the prevention and control of infectious diseases.It is the key to whether the epidemic situation can be controlled in the outbreak period.The clustering algorithm can be used to predict the epidemic situation and cause in the early stage of the outbreak of infectious diseases,and to support the epidemic prevention and control measures.At present,there are time cluster monitoring,spatial clustering monitoring and time and spatio-temporal clustering monitoring method.As the outbreak of infectious diseases outbreaks and uncertainties,which requires clustering algorithm must have timeliness,sensitivity and accuracy characteristics.However,the current spatio-temporal clustering algorithms for infectious disease surveillance,using the regular spatial range to determine the geographical distribution of the epidemic,it is difficult to find the disease distribution in more complex and diverse situation,resulting in lower accuracy of monitoring.Although some scholars have proposed to generate irregular shape of the spatial range to find the exact distribution of the epidemic,but due to the impact of computational efficiency,it can only be used for small area monitoring.In addition,in the real infectious disease monitoring process,because the infectious diseases and common diseases in the early have similar symptoms,making the analysis of the effect of the algorithms are difficulties.The above problems seriously affected the development of infectious disease prevention and control work.In order to solve the problems existing in the current surveillance of infectious diseases,this paper establishes a spatio-temporal clustering framework for infectious disease surveillance through the in-depth study of epidemiology time clustering and spatial clustering algorithm,and proposes a spatial clustering of irregular spatial scale infectious diseases.Then we introduces the time factor to extend the spatial clustering into a spatio-temporal clustering monitoring algorithm.The heuristic method is used to provide the speed of the algorithm to solve the contradiction between the accuracy and computing time faced by the previous space-time clustering algorithm.Finally,the simulation data is used to verify the proposed algorithms.The main work of the paper is summarized as follows:(1)Analyzes the key scientific problems that need to be solved to monitor the outbreak of infectious diseases,and summarize the theoretical and difficulties of current infectious disease surveillance from three aspects: time,space and spatio-temporal.The spatial clustering monitoring method is selected as the main research point,and the representative monitoring methods are compared and analyzed.The technical feasibility of the spatio-temporal clustering monitoring of infectious diseases are demonstrated and analyzed.In the end,the framework of the spatio-temporal clustering for infectious disease surveillance is proposed.(2)Based on the analysis of the advantages and disadvantages of the previous spatial clustering algorithm,a spatial clustering monitoring algorithm is proposed to accurately monitor the outbreak area of irregular shape.First of all,according to the administrative area will be between the blocks of geographical relations into a network structure,in order to reduce the computational complexity,to facilitate the solution of the problem.Then,through the initial positioning of the epidemic range,reduce the size of the problem,to ensure the accuracy of the case,to improve the operation speed of the algorithm.Then,based on the likelihood function,we propose the test statistic to determine whether spatial clustering exists.Finally,according to the theory of graph theory to determine the constraints of monitoring problems,the problem is abstracted as a constrained optimization problem,and through a simple diagram shows the effect of the algorithm.(3)According to the proposed framework of spatio-temporal clustering,the proposed spatial clustering algorithm is extended to the clustering algorithm in spatio-temporal data.The basic idea is to first use the time clustering algorithm to determine the general outbreak time range of the epidemic and transform the spatio-temporal data into spatial data divided by time slice(days).Then,by analyzing the computational complexity of the algorithm,it is found that the spatio-temporal clustering problem belongs to the NP problem,and the genetic algorithm is used to improve the computational efficiency of the algorithm for large scale monitoring.Finally,the simulation data of H1N1 influenza from Beijing is combined with real outpatient data,for real-time monitoring of outbreaks,which were carried out on spatio-temporal and spatial clustering monitoring.Simulation of the data can be obtained because of the state of the patient at all times,it can clearly know the data source,and then the accuracy of the algorithms can be a reliable evaluation.The experimental results show that the algorithm proposed in this paper can be used in spatial clustering and spatiotemporal clustering,and compared with the previous methods.