The Detection And Mining Of Urban Hotspots Based On Social Geographic Big-data

With the arrival of the big data era,the analysis and mining of geographic big data provide a solid support in respects of macro decision making and public opinion monitoring in social development,urban planning.Geographic big data contributed by users are open resources for the public and organizations on social networks and they have large amount of location and time information.This thesis applies spatial statistics theories to detect the pattern,the cluster and the outlier of sign-in data.the author also studied the characteristics of the distribution,space-time evolution and users behaviors within a hotspot area.Besides,by applying the conclusion of the thesis to the study of the correlation between economic indexes and user activities in Shenyang,this thesis has provided references for its economic decisions and urban planning.The body of the thesis are as follows:(1)This thesis expounded the meaning and characteristics of geographic data of social media and summarized results of relevant scholars.Besides,social media geographic big data can be applied to detect and mine urban hotspots.The author took Weibo as an example,and provided a solution to API calling and parallel processing in sign-in data acquisition.The author pre-processed time,location and semanteme and put forward a method to process repeated data and made a basic descriptive statistics about sign-in data.(2)Based on the theory of spatial statistics,this thesis concluded the central tendency,dispersion,direction,distribution and relation of spatial data.Besides,this thesis put forward a time sequence and semanteme based analysis mode to study the space-time evolution and users behaviors in urban hotspots.This thesis upgraded the sampling of spots,which is carried out by weighting sign-in data attributes.By doing so,user activity and unit can be calculated and presumed.(3)The time and number of sign-in information of Weibo were introduced to the sample and the author put forward an improved sampling by using multi-time and weighting methods and used this method to analyze the effect of modifiable areal unit on the spatial correlation of sign-in data.The author applied spatial autocorrelation of increment and concluded that the result of spatial correlation analysis is relative,hence the uncertainties brought about by scale effect and zoning effect must be considered.(4)The spatial detection to the sign-in data identified the correlation between the sign-in data and their related values.By measuring the correlation of spaces,studying the spatial pattern of sign-in data,the author got clusters and outliers.The author used the analysis model to explore the characteristics of the central space,coverage,direction,space-time evolution and users behaviors of urban hotspots.By combining natural language processing method and semantic information of sign-in data,The author studied users behaviors in the research zone.(5)The author analyzed the closest index—economic index by calculating Pearson Correlation Coefficient,Grey Relational Analysis in order to find out influence factors of economic index variation.The analysis results show that the hot spots can be used within the division of the active users of national economic development trend forecast,provide the reference for the government planning.