Research On Characteristics Of Population Distribution Based On Positioning Data

The development of smart phones, global positioning system(GPS) and mobile Internet jointly promote a new type of service, Location-based Service. Location-based Service, which is different from the traditional Internet services, based on the smart phone positioning, provide information service that is highly relevant to location to the users. Mobile positioning technology has played a basic role in the booming of smart phone usage and the diversification of mobile Internet applications. Mobile positioning technology not only easy user’s life, but also support the mobile application service, most importantly this service is producing large amount of position information of users. Massive positioning information, can be considered as a detection of the state of city. Due to the popularity of smart phones, the massive location data is a sampling on spatial distribution of population. Supporting by the Hadoop platform, we made an integration on the log of Baidu mobile positioning services system and generated a positioning density image data. Based on this data, comparing with the DMSP/OLS night-time light data and location share data from Weibo, we characterized the distribution of population based on the analysis, we reached these conclusions: 1. the number of user’s positioning changes periodically; 2. the space distribution of population follows Power-law; 3. The positioning data characterized the spatial distribution of population and its variation. Based on the characteristics of population distribution, the positioning density image can be considered as a sampling on population distribution, we photographed the population density of China. The analysis of spatial-temporal changes of positioning density data, shows that large amount of information lays in positioning data, based on which two research are applied: 1. A majority of human lives in urban area, based on the random forest algorithm, we carried on a classification of the city area. 2. The spatial and temporal variations of positioning density has a strong correlation between urban land use types, training the gradient boosting decision tree model for urban land type classification and verified its validity. By the exploration of the distribution model of population and digging of the information lays in positioning density data, we fulfilled the idea of human as sensors, verify potential of the user generated positioning data in the geography research.