Urban land use classification based on big data: case of Xining

Urban land use is the result of interactions among social, political, economic, technological and other factors within and without cities．Urban land use classification not only helps to analyze land use pattern, but also has great significance for rational urban zoning and promotion of sustainable development．Urban land use classification in Xining is done based on two types of commonly used big data (Easygo, points of interest or POI) and three common classification methods (Maximum Likelihood, Support Vector Machine, Artificial Neural Networks)．By comparing the accuracy of results under different data and methods, optimal data combination and classification method for extracting urban land use information are determined．The classification results are used to analyze urban land use patterns in Xining．Urban land use information obtained by neural network classification method based on Easygo and POI was found to have the highest accuracy, with overall accuracy at 71.25% and a Kappa coefficient at 0.62. Easygo and POI can reflect more information about characteristics of different land use．Artificial Neural Networks can fully integrate information of multi-source big data．Therefore, it provides a potential way to timely and accurately obtain urban land use information with multi-source big data and Artificial Neural Networks．