A STUDY ON EFFICIENT MAKING METHOD OF DETAILED TIME-SERIES URBAN DATASET BY SPATIAL INTEGRATION OF YELLOW-PAGE DATA AND DIGITAL MAPS FOR URBAN ANALYSIS

摘要

There were many studies of urban analysis with various ways in various disciplines. But information used often has low spatial resolution especially when they use regional statistics, though the statistical data can over large areas with homogeneous quality. For detailed analysis, some studies rely on field survey that has very fine spatial resolution, but they fail to cover an entire urban area or large regions including suburb and rural area. In this study, there are three aims. First is to develop a method of making time-series tenant dataset by integrating of multi-year yellow page. Second is to develop a method of generating a new digital map with time-series tenant data that can cover whole urban area or national land by integrating detailed digital maps with time-series tenant data. Third is to develop a new urban analytical method using the dataset of this study. In this study, we developed two kinds of spatial integration method. First is to develop the dataset of time-series changes about all tenants by integrate multi-year yellow page data. Using this database, we can acquire many patterns of all tenant changes. In addition, we can also acquire floor information from building name. Second is to integrate detailed digital house map of Japan called Zmap with dataset of time-series tenant changes. Both methods integrate data based on the similarities of name and position. Name information is identified by one of the natural language processing called N-gram. N-gram can calculate similarities of texts. All of this system is automated by C++, so anyone can acquire results easily in short time with high accuracy (accurate test showed accuracies of object identification are about 93 percent). This dataset allows us to cover whole extent of Japan with "accuracy" and "flexibility".