Research on Spatial Pattern and Its Industrial Distribution of Commercial Space in Mianyang Based on POI Data

The rational layout of urban commercial space is conducive to optimizing the allocation of commercial resources in the urban interior space. Based on the commercial POI (Point of Interest) data in the central district of Mianyang, the characteristics of urban commercial spatial pattern under different scales are analyzed by using Kernel Density Estimation, Getis-Ord * G , Ripley’s Function and Location Entropy method, and the spatial agglomeration characteristics of various industries in urban commerce are studied. The results show that: 1) The spatial distribution characteristics of commercial outlets in downtown Mianyang are remarkable, and show a multi-center distribution pattern. The hot area distribution of commercial outlets based on road grid unit is generally consistent with the identified commercial density center distribution. 2) The commercial grade scale structure has been formed in the central urban area as a whole, and the distribution of commercial network hot spots based on road grid unit is generally consistent with the identified commercial density center distribution. 3) From the perspective of commercial industry, the differentiation of urban commercial space “center-periphery” is obvious, and different industries show different spatial agglomeration modes. 4) The mul-ti-scale spatial agglomeration of each industry is different, the spatial scale of location choice of comprehensive retail, household appliances and other industries is larger, and the scale of location choice of textile, clothing, culture and sports is small. 5) There are significant differences in specialized functional areas from the perspective of industry. Mature areas show mul-ti-functional elements, multi-advantage industry agglomeration characteristics, and a small number of developing areas also show multi-advantage


Introduction
The spatial layout of the commercial space has been a hot spot in modern urban research; the western scholars have paid close attention to the commercial space theory such as the Central Place Theory [1] [2], the Reilly's Law of Retail Gravitation [3], the theory of spatial utilization structure [4] [5] etc., which have be- gradually enriched. Increasingly, scholars study the structure of urban commercial space layout by investigating consumer shopping trips and travel and residents' consumption behavior [6] [7] [8] [9], however, the traditional research is limited to data and technology, and lacks the analysis of the overall pattern of urban commerce and the characteristics of local spatial differentiation based on large samples. With the rise of large data, the new technology methods and means provide the possibility for the research of urban commercial space structure. For example, Zhou Suhong et al. [10] used GPS data to explore the attractive attenuation effect of commercial space. Wang De et al. [11] made a comparison of different business districts in Shanghai by using cell phone signaling data. Hu Qingwu et al. [12] used Weibo check-in data to study the urban business districts.
In recent years, based on POI big data, the empirical study of space has emerged to identify urban hot spots based on KDE (Kernel Density Estimation), Moran's I, Getis-Ord * i G etc. For example, Chen Weishan et al. [13] identified the characteristics of hot spots in Guangzhou retail industry based on POI data, and the relationship between the spatial agglomeration characteristics of different retail formats and its market and positioning, business model and location strategy is discussed. Xu Dong et al. [14] analyzed the distribution pattern of the urban leisure tourism space based on the POI data, and drew the conclusion that the leisure tourism space has obvious dependence on the scale. Zhao Hongbo et al. [15] took cultural facilities as the research object, quantitatively analyzed the evolution of the spatial pattern of Zhengzhou cultural facilities and its location

Study Area and Data Sources
The  characteristics, such as "shopping related places" and "specialty stores". After spatial matching, deduplication, and deletion of low-recognized commercial outlets, a total of 30,979 valid POI data were obtained ( Table 1).
The spatial distribution of commercial outlets in the study area is shown in Figure 1. Ensure that the POI data is consistent with the geographic coordinate system of the administrative division data of Mianyang, and converting them into projections. 1 The National Economic Industry Classification Standard (GB/T 4754-2017) states that the retail industry refers to the sales activities of department stores, supermarkets, specialized retail stores, brand stores, and stalls that are mainly targeted at end consumers (such as residents) Sales activities by Internet, post, telephone, vending machines, etc. According to statistical standards, it includes 9 types of retail industries, including comprehensive retail, retail of food, beverages and tobacco products, retail of textiles, clothing and daily necessities, cultural, sports products and equipment. Due to data limitations, this article mainly focuses on retail establishments with fixed stores, and does not consider stalls, no stores, and other retail industries. Journal of Data Analysis and Information Processing

Kernel Density Estimation
The kernel density estimation method is used to calculate the density of the point features around each output raster cell. The points closer to the center have greater weight.
where h is the threshold radius, that is broadband, n represents the number of sample points, d − d i is the distance from the estimated point to the sample point, k is equal to the spatial weight function.

Getis-Ord i G *
The statistic is used to analyze the agglomeration degree of attribute values at the local spatial level by comparing the local summation of adjacent elements with the summation of all elements in a given distance range. The formula is expressed as follows: In the formula: x j is the element attribute value of the j spatial unit; n is the total number of elements; w ij represents the spatial weight matrix. If the distance between the i and the j spatial units is within a given critical distance d, they are considered to be neighbors and the elements in the spatial weight matrix are 1, otherwise, the element is 0.
where j δ is the density index of commercial outlets in the j-th block unit, m j is the number of commercial outlets in the block, p j is the area of the j-th block.

Ripley's K Function
Ripley's K function can be used to analyze the distribution patterns of spatial point features on different spatial scales. The calculation formula is as follows: In the formula: A is the area under study; n is the number of commercial outlets in each industry; d is the distance threshold; w ij (d) is the distance between commercial outlets i and j in a certain industry within the range of d. In 1977, Besag proposed to replace K(d) with L(d) and to square K(d) with a linear transformation to keep the variance stable. The formula is:

of Data Analysis and Information Processing
The relationship between L(d) and d can test the spatial distribution pattern of various industries within the range of distance d. L(d) = 0 means that the industry is randomly distributed, L(d) > 0 means that the industry is agglomerated, L(d) < 0 means that the industry is decentralized.

Location Quotient
Location Quotient is often used to analyze the degree of specialization of regional leading industries, which is helpful to measure the spatial distribution of a certain factor. This study uses location entropy indicators to analyze the degree of regional specialization of the retail industry. The higher the value, the higher the degree of specialization of the industry type in the region. The formula is as follows: where Q is Location Quotient, e K-A is the ratio of the number of outlets of industry type A in area K to the total number of all outlets of industry type A in the entire area, e K is the total number of outlets in area K and the number of outlets in the entire area ratio.

Overall Distribution Characteristics of Commercial Outlets
Using the Average Nearest Neighbor Distance module in ArcGIS to analyze commercial outlets, it was found that the NNI of the commercial outlets was  ing a relatively small-scale cluster "Island". In general, a more reasonable commercial-grade scale structure system has been formed in the downtown area of Mianyang as a whole.   trict are mainly used as the center of the area to serve the surrounding residential groups, which has radiation effect in a certain range. In the second category, with 95% confidence, hotspots with a Z-score greater than 1.96 (P value < 0.05) for testing are mainly distributed in Fucheng District. Although these areas have small-scale commercial centers, In terms of distribution, most of its blocks form relatively significant hotspots that are not surrounded by other neighboring high-density commercial network blocks. The third category is the area where the Z-score used for the test is between −1.65 -1.65. The spatial autocorrelation is weak, and the distribution of commercial outlets in the neighborhood is random.

Spatial Aggregation Degree
According to the calculation of Average Nearest Neighbor (Table 2)

Spatial Distribution Characteristics of "Center-Periphery"
To better study the "Central-Peripheral" differences in the number and density

Characteristics of Multi-Scale Spatial Agglomeration
The calculation results of Ripley's K function show that within a 10 km observation range, the distribution of commercial outlets of automobiles, motorcycle fuels, and spare parts shows a "double peak" feature, while the distribution of commercial outlets in other industries shows a "single peak" feature ( Figure 7). influence from spatial location. In order to bring greater service coverage, they are dispersed in the core and fringe areas of the city. Some industries that have high requirements for location selection, such as textiles and clothing, culture and sports, are mostly distributed in urban business districts, and they tend to be located in core urban areas. Automobile, motorcycle fuel and spare parts, hardware, furniture and interior decoration materials, household appliances and electronic products industries due to the requirements of site size, tend to layout in the outer areas of the city. The service coverage of the leisure and entertainment industry is less than that of the comprehensive industry mainly based on basic social services. The "double peak" feature of the former Ripley's K function also further shows the difference between the two types of commercial outlets in the industry. Relatively speaking, the motorcycle fuel and spare parts industry has a smaller range of spatial scales and is more likely to form dense distribution.

Characteristics of Specialized Functional Area
The study selects the administrative division of Mianyang for the measurement of location entropy (  speaking, there is a great difference in the degree of industry specialization among sub-districts and towns. Mature sub-districts and towns show the characteristics of multi-advantage industry agglomeration, [16] but the degree of specialization is not high enough, such as Chaoyang and Gongqu sub-district.
There are also a few fast-developing sub-districts and towns that exhibit the

Discussions
In general, the research on the distribution of urban commercial outlets and in- standing of the specialized functional agglomeration area of the industry. However, due to the lack of POI data attribute information, such as lack of creation time, outlet size, etc., it is difficult to discuss the formation mechanism of the commercial space pattern. In the future, we should further explore the causality behind the spatial pattern of urban commercial industry in combination with the "small data" of the traditional survey. In addition, although the paper analyzes and obtains the spatial scale range of the maximum agglomeration in each commercial industry, whether it is really the scope of producing the best agglomeration benefit, or whether it can guide the layout of all kinds of commercial industries based on this scope, still needs to be deeply explored.

Conclusions
The 1) Urban commercial spatial distribution is characterized by significant agglomeration and has shown a polycentric distribution pattern. On the whole, it presents a "Y-type" agglomeration with the commercial area from Nanhe road to Linyuan road as the core, the commercial distribution spatial pattern with the area around Wanda shopping center and railway station, the peripheral area radiated by Tanghong international shopping center as the sub-center, and several secondary centers gradually formed in the urban fringe. The cluster area of road grid of commercial network is consistent with the distribution of identified commercial hot spots, but the influence of different commercial centers on adjacent grid is different. In addition, the hierarchical scale structure of the commercial center system is reasonable, with more primary and secondary centers and fewer secondary centers, indicating that the polycentric pattern of commercial centers has been formed.
2) The "center-periphery" of urban commercial space is clearly differentiated from the industry perspective, and the multi-scale spatial agglomeration perfor-