Abstract: Facing high-frequency and massive customer orders in O2O takeaway platform, how to divide O2O takeaway business zone and analyze the spatial distribution law of customer demand is of great value for optimizing the allocation of rider resources and improving the distribution efficiency. Based on a large amount of historical data in O2O takeaway platform, including the information of orders, merchants and customers. Firstly, the distribution area of takeaway orders is divided into network cells, and the statistical data in the grid represent the characteristics of orders in this network cell; A new soft quadratic surface support vector regression machine is introduced to characterize the relationship between the average price of orders (i.e. customer demands) and order location, the average price of discounts, the average price of additional costs, and then determine the optimal grid size. From the perspective of customers, based on the grid model and the density of O2O takeaway orders, the O2O takeaway business zone is divided by using the grid density clustering method. From the computational results, compared with the business zone division adopted by current O2O takeaway platforms, the average distance of takeaway delivery is decreased by 8.97%, the Davies-Bouldin Index is decreased by 73.61%, and the number of business zones is decreased by 56.10%. The newly-divided O2O takeaway business zones make the riders distribute within the designated business zone as much as possible, and reduces the cost of resource allocation in the business zones. The proposed division also verifies that the customer demand density is much more helpful than the information of traditional administrative regions and merchant locations in dividing the O2O business zone. Finally, based on the newly-divided O2O takeaway business zones and correlation coefficients, some spatial distribution laws of O2O takeaway customer demands are discovered, including the distribution law of different types of merchants and orders and the cost preference of customers, so as to provide some guiding managerial insights for the siting and pricing of different types of O2O takeaway merchants. For instance, managers may get more customer demands by locating the stores near campus or commercial buildings. The non-key merchants should try to locate in areas with high demands. If they can only be located in areas with low demands, areas with dense merchants should be avoided. When the key merchants open additional branches of stores, areas with low demands should be selected. Merchants should appropriately increase the box price and reduce the delivery price as much as possible, which can promote consumer consumption. However, residents in some first-tier cities are not sensitive to the price of takeaway. Merchants may not make more profits by adopting the strategy of small profits but quick turnover, they need to pay more attention to the timeliness and quality of takeaway services.

Key words: O2O mode; business zone division; grid density clustering; spatial distribution law; big data mining