Research On Scan Statistic Methods For Detecting Geographical Spatio-temporal Anomalies

Spatio-temporal data is the addition of spatial attributes and time information to the data with thematic attributes.The spatio-temporal anomaly refers to the spatio-temporal entity whose subject attribute value deviates significantly from its time or(and)the spatial reference domain.The purpose of spatio-temporal anomaly detection is to use some method to find spatio-temporal clusters in spatio-temporal data.The study of spatio-temporal anomalies can help to understand the distribution of spatio-temporal data and the movement characteristics of spatio-temporal hotspots,and can find potentially useful information and laws.The common method of space-time anomaly detection is also extended by the method of spatial anomaly detection.The space-time scanning statistical method is one of the most commonly used methods for space-time anomaly detection.This paper mainly studies the statistical method of geographic space-time anomaly scanning.The main contents include:(1)Research on algorithm: Two improved methods of gravitational search algorithm are implemented.CPGSA algorithm combines chaos optimization and memory group idea of particle swarm optimization algorithm to optimize the velocity term of gravitational search algorithm,and replaces resultant force calculation and speed update with chaotic sequence.The random sequence improves the problem that the algorithm is not easy to converge later in the iteration;the LGSA algorithm combines the idea of local search,and uses the mental search process as the neighborhood action of the local search,which improves the shortcomings of the local optimization ability in the late iteration of gravitational search algorithm.The test results of the 23 standard test functions on the LGSA algorithm show that the performance of LGSA is greatly improved compared with the GSA algorithm.(2)Design of spatio-temporal scanning window: 3 different spatio-temporal scanning windows are designed,which are: straight cylindrical spatio-temporal scanning window,oblique cylindrical spatio-temporal scanning window,variable-radius cylindrical spatio-temporal scanning window,and for each The scan window uses the artificially generated test data sets I,II,and III for performance test evaluation.The results show that the three shapes of spatio-temporal scan windows have very good detection ability for their corresponding shape spatio-temporal anomalies,and the design of three kinds of windows improved detection of irregular shaped anomalies.(3)Application of the algorithm: The data adopts the national hand-foot-and-mouth disease dataset of 2016,the hand-foot-and-mouth disease dataset of Guangdong Province in2009 and the PM2.5 dataset of Shandong Province in 2016 and 2017,spatio-temporal anomaly detection of datas using three different spatio-temporal scanning windows and kullldorf method.The detection result indicates that:for the national hand-foot-and-mouth disease dataset of 2016,the result of the radius-variable cylindrical scanning window is better;for the hand-foot-and-mouth disease dataset of Guangdong Province in 2009 and the PM2.5dataset of Shandong Province in 2016 and 2017,the results of the tilted cylindrical scan window are better.