Effect of Artificial Intelligence on the Development of China’s Wholesale and Retail Trade

Abstract

The rapid development of digital technologies and massive data analytics has enabled artificial intelligence (AI), via “machine learning”, to impact many societal sectors, including the wholesale and retail trade (WRT). However, the specific impact pathway and dynamics are still unclear. Based on the panel data of 30 provinces in China from 2015 to 2021, this paper employed the “VHSD-EM” model, random forest algorithm, and partial effect analysis to build an evaluation index system of AI and WRT, then to study the impact of AI on WRT in the temporal and spatial dimensions. Our main discoveries were as follows: (1) the quality of the WRT aligned well with the relative level of AI in the provinces, although the latter developed at a relatively fast pace; (2) the shortcomings that hindered the quality of WRT development varied in different regions, with a stark mismatch between the degree of informatization and the level of economic development in the eastern coastal region, a lack of innovation in the relatively high economic presence of the northern provinces, and a weak sharing of resources in the western region; (3) AI enhanced WRT development jointly with other key factors, particularly the density of employment, the percentage of WRT employees, and the ratio of the year-end financial institution deposits to the regional GDP, which raises the importance of the transaction volume of the technology market; (4) spatial differences exist in the impact pathways of AI on the high-quality development of WRT, and, for most provinces and regions except Shanghai and Guangdong, there is still significant room for expansion in the utilization of AI in WRT.

1. Introduction

China’s super-scale market can serve as a model system for the construction of new development paradigm featuring dual circulation, in which domestic and overseas markets reinforce each other, with the domestic market as the mainstay. However, the wholesale and retail trade or industry (WRT) still faces challenges such as weak core competitiveness and poor connection between supply and demand, which still need to be solved by new technological innovations. With the increasing demands of consumers for quality, brands, and service, China’s WRT is facing enormous market pressure. Therefore, how to meet the needs of different consumers and improve service quality will be an important challenge.

The deep integration of information technology and traditional industries has promoted the transformation of traditional industries into their modern, intelligent counterparts. Due to the rapid development of the internet, browsers, and search engines, and the rapid rise of “machine learning” using massive data, many wholesale and retail enterprises use computers to simulate consumer behaviors such as learning, reasoning, contemplating, and planning. The application of artificial intelligence (AI) technology in various industrial sectors has influenced remarkable changes both inside and outside the industrial sector, and has contributed to the emergence of new forms of WRT. In this context, AI has become a key enabler driving the innovative development of China’s WRT.

2. Literature Review

Opinions vary among scholars regarding the design of an evaluation index system for the high-quality development of the WRT, the definition of which also evolves over time. Therefore, the evaluation indicators for the high-quality development of WRT should be both scientific and innovative [1]. High-quality economic development should follow the development concepts of innovation, coordination, greenness, openness, and sharing, which is also the main framework followed by many scholars in high-quality economic development. Based on this, Chen et al. [2] established a high-quality development evaluation system including the five development concepts and a security guarantee, among which the importance of communication capability is put forward in the security guarantee. In the era of digital economy, the efficient operation of WRT is inseparable from the popularization and application of the computer network, so the definition of high-quality development must involve the level of industrial informatization [3,4,5]. To summarize, this paper sets up quality indicators from six dimensions, i.e., innovation, coordination, greenness, openness, sharing, and degree of informatization, to assess the high-quality development of WRT in a more comprehensive and modern manner. Among these dimensions, informatization refers to the ability of WRT to integrate digital technologies with the information industry, including the abilities of information processing and communication. The level of informatization of WRT can be measured from the perspective of the utilization of the information network and the infrastructure construction.

AI is generally regarded as the capability of a “machine” to perceive, synthesize, and infer information, in contrast to the intelligence shown by humans. There are well-developed research methods for the application of artificial intelligence in both the medical and industrial fields. Allen et al. [6] evaluated the performance of AI models in clinical practice by collecting a volume of AI products for clinical use. Nahavandi et al. [7] believed that intelligent robots, which would play an important role in AI, penetrate into the manufacturing supply chains and production shop floors to an unprecedented level. The similarity of the previous studies is that both AI products in the medical field and industrial robots are relatively easy to quantify. In some other fields, scholars often adopt qualitative research methods. Jallow et al. [8] adopted a qualitative case study research methodology to explore the uses of AI and its benefits for the UK construction industry. In the retail field, Pillai et al. [9] surveyed 1250 consumers to investigate their shopping intentions at AI-powered automated retail stores. Bonetti et al. [10] also combined 74 stakeholder interviews and 14 on-site retail observations to examine how employees’ practices changed when retailers invested in AI. It can be found that studies on AI in WRT are always based on realistic surveys or observations. In WRT, many companies view AI algorithms as core business secrets, which makes it difficult to obtain direct data.

However, when considering AI as an independent factor and exploring its impact on WRT, the samples obtained through surveys and interviews may be slightly insufficient, so we can refer to the overall development evaluation index system of AI. AI should be used to benefit mankind, and the evaluation cannot be separated from its potential applications in the economy and in society [11,12]. Kuang et al. [13] assessed the level of AI technology innovation in China from the perspectives of AI-related infrastructure construction, capital and personnel investment, AI research results, and the economic benefits of AI enterprises. Liu et al. [14] discussed the maturity of AI technology in China; they believe that the level of development of AI is strongly related to its degree of application in the industry. The high development level of AI should meet three conditions: first, the region should have relatively complete AI infrastructure; second, the industry should have the technology and talent for the use of AI; and third, the development of AI should be able to bring about a practical outcome, i.e., capability and/or volume of output.

To summarize, the existing studies have the following deficiencies: Firstly, data on AI technology for WRT use are difficult to obtain, and the evaluation index of AI in WRT is not completely unified. Secondly, the existing research on China’s WRT is either based on the whole country or a single region. However, the imbalanced development of China’s economy may limit the applicability of the conclusions. There is little comprehensive theoretical and empirically proven knowledge regarding how to profoundly develop and utilize AI’s capabilities [15,16]. Meanwhile, there is no unanimous conclusion on whether or how the influence of AI on WRT will change under different conditions in China. This paper evaluates the development level of AI from three aspects: infrastructure construction, AI application capabilities, and AI application achievements; and then analyzes the development of AI, from the technical components to their final impacts, comprehensively. Based on the existing literature, this paper uses the “VHSD-EM” model to study the temporal evolution, spatial differences, degree of impact, and path of the impact of AI on the WRT and to establish evaluation indicators based on the definition of AI and high-quality development of the WRT. We intend to show that AI has a positive impact on promoting the high-quality development of WRT. This study also combines a machine learning model with a high-quality development problem to reduce the impact of the correlation between variables and to improve the generalization ability of the model. The difference between the levels of high-quality development of AI and WRT is analyzed from the perspectives of temporal and spatial dimensions. Finally, we use a random forest model and a partial effect diagram to analyze the impact mechanism of the AI development level on the high-quality development of the modern circulation industry, which enriches the research perspective of this field.

3. Theoretical Analysis of the Influence of AI on the High-Quality Development of WRT

The specific applications of AI differ in different industries. New business forms and models spawned by AI will promote the transformation and upgrading of the industrial structure [17,18,19,20]. There are three hypotheses in this paper.

According to the theory of endogenous growth, technological progress is a determining factor ensuring sustained economic growth. Nowadays, the combination of technology and the economy has promoted the emergence of the digital economy, which is based on big data and AI. Perifanis et al. [21] considered that the development of new business models and competitive advantages through the integration of AI into business and IT strategies held considerable promise. Stallkamp et al. [22] believed that digitalization has enabled firms with platform business models to emerge in many sectors of the economy, which creates more value by connecting different groups of users. Linkov et al. [23] also showed that the growth of a digital economy could increase productivity and benefit the local and global economies. Wang et al. [24] discussed the impact of internet development on green economic growth in China. We can learn from this research that AI will promote industry integration, which means that WRT will connect more closely with other economic fields and develop along with the social economy, driven by technological progress.

H1. 

AI expands the positive externality of WRT.

Cheng’s study [25] of Chinese enterprises and employees showed that there are marked differences in the degree of application of robots in different industries. Additionally, enterprises with greater employment scales and higher capital labor ratios have a higher degree of application of robots. In fact, the impact of AI on the development of WRT mainly follows three paths. First, AI can improve technological innovation and informatization in WRT. The technical methods and knowledge used by AI can permeate all aspects of the industry [26]. On this basis, more technological innovations can be carried out to improve the utilization of information technology in WRT. Second, AI has broken the trade boundaries, alleviated the imbalance of regional development, and enabled the retail industry to operate more efficiently. Enterprises can quickly obtain economies of scale and economies of scope, which has expanded the market boundary to further regions with lower costs. It has also improved the infrastructures of transportation, logistics, and computer and other facilities in the region. As the foundation of AI, data resources, digital technology, and software and hardware infrastructure have become new key production factors circulating around the world, making much of the information no longer unique to a country. The alleviation of information asymmetry in different locations has driven the emergence of more cross-border factories, circulation organizations, and transactions. The efficient data learning and processing functions of AI can quickly access massive databases of information, enabling mutual matching and transportation route planning for both parties involved in the transactions and realizing the integration of the supply chain. Third, with the accumulation of experience in “smart” monitoring, it is becoming a preventive measure [27]. Wilson et al. [28] also found that AI can be used for the “smart” organization and management of circular production works as well as automated and highly accurate sorting of production and consumption waste. Chanhan et al. [29] believed that AI combined with business model innovation are deemed to provide solutions to circular economy transformation by optimizing product service system. As a result, AI can enhance the environmental protection capabilities of the WRT. To summarize, the influence mechanism is shown in Figure 1.

H2. 

AI can promote WRT from six aspects.

Zou [30] conducted research by dividing cities of different sizes and found that the Internet has the characteristics of increasing the “marginal effect” on technological innovation. In more developed regions, this can lower the threshold for enterprises to obtain market information and provide a good foundation of network information for industrial upgrading [31]. Uneven regional development in China has caused some problems, and this can be observed in the economy, finance, technology, and industrialization [32,33]. The difference in factor endowments will lead to differences in the development processes of all industries.

H3. 

AI has varying impact paths on WRT under different conditions.

4. Evaluation of the WRT and AI Development Level

4.1. Evaluation Index System for High-Quality Development of WRT and AI Development Level

Zhang et al. [34] built a VHSD (vertical and horizontal scatter degree method) model to discuss the effect of digital economy on the carbon emission performance, and showed that at the significance level of 10%, digital economy improves the carbon emission performance of resource-based and non-resource-based cities, but the dividend effect of digital economy in resource-based cities is more obvious. Entropy method (EM) models are always used to evaluate the weights of the indicators. Ferreira et al. [35] used the maximum entropy method with projected climate data to 2100 to project the spatial distribution of the above-ground biomass density across the Brazilian Atlantic Forest domain. This paper, in contrast, uses the “VHSD-EM” model, which makes use of the best of two worlds, to evaluate the development of AI and WRT. The index weight calculated by the “VHSD-EM” model is shown in Equation (1):

$${\delta }_{mk}=\frac{W_m+w_{mk}}{2}$$

(1)

In Equation (1), δ_mk is the weight of the mth index of the evaluation object calculated by the “VHSD-EM” model in period k; W_m is the weight of the VHSD model as the mth index; and w_mk is the weight of the mth index of the EM model in period k. We multiplied δ_mk by the observed values of the evaluation object and summed them up to obtain the corresponding index.

This paper refers to existing research results [36,37] and combines them with actual situations to establish an evaluation index system for the high-quality development of WRT from six perspectives: innovation, coordination, greenness, openness, sharing, and informatization. The specific indicators are shown in

Table 1. Evaluation index system for the high-quality development of WRT (the table was created by the authors).

Primary Indices	Secondary Indices	Tertiary Indices	Description	Index Attribute
Innovation	Innovation investment	Number of scientific studies and development institutions		direct ratio
		R&D expenses of wholesale and retail enterprises in the WRT		direct ratio
	Innovative output	Number of patent applications in the WRT	Number of patent applications that mention the keywords “wholesale” or “retail”	direct ratio
		Number of new product developments		direct ratio
Coordination	Coordinated development of industries	Per-capita e-commerce sales	E-commerce sales/population	direct ratio
		Per-capita retail sales amount of consumer goods	Total retail sales of consumer goods/population	direct ratio
	Coordinated urban and rural development	Ratio of rural and urban retail sales of consumer goods	Rural retail sales of consumer goods/urban retail sales of consumer goods	direct ratio
Greenness	Carbon emission	Carbon emissions of unit WRT added value	Carbon emissions/WRT added value	inverse ratio
	Energy consumption	Energy consumption of unit WRT added value	Energy consumption/WRT added value	inverse ratio
Openness	Foreign investment	Foreign direct investment in the WRT		direct ratio
	Open environment	Total import and export volume of the WRT		direct ratio
Sharing	Circulation facilities	Highway mileage		direct ratio
		Railway mileage		direct ratio
		Waterway mileage		direct ratio
		Per-capita fixed assets investment in the WRT		direct ratio
	Circulation benefit	Profit margin of wholesale and retail enterprises		direct ratio
	Employment opportunities	Number of employees in the WRT		direct ratio
Informatization	Information processing ability	Number of broadband access users per capita in the region	Number of broadband access users/population	direct ratio
	Communication ability	Number of mobile phone users per capita in the region	Number of mobile phone users/population	direct ratio

.

With reference to the existing research results [38], this paper establishes an evaluation index system of the AI development level from three aspects: AI infrastructure construction, AI application abilities and AI application achievements. In evaluating the achievements of AI, the existing literature often collected industry patent data from the computer industry, but the computer industry cannot fully represent the field of AI. Therefore, this paper directly selects the related patents of AI as one of the indicators with which to measure the scientific and technological achievements of AI. In addition, this paper uses industrial robot installation density to measure the AI infrastructure construction, an index which is often applied to manufacturing. However, the application of the industrial robot promotes man–machine collaboration and has become an important part of industrial intelligence, as industrial robot installation density is also very important for the development of AI. The specific indicators are shown in

Table 2. Evaluation index system of AI development (the table was created by the authors).

Primary Indexes	Secondary Indexes	Tertiary Indexes	Description	Index Attribute
AI infrastructure construction	Software popularization and application	Revenue of basic software, support software, and embedded application software products		direct ratio
	Information infrastructure construction	Fixed assets investment in the information technology and software industries		direct ratio
	Funding input	R&D investment in the software and information technology services industries		direct ratio
AI application capabilities	Talent preparation	Number of software developers		direct ratio
	Equipment preparation	Industrial robot installation density	Number of industrial robots used per 10,000 workers in the manufacturing industry	direct ratio
	Information receiving and processing capability	Length of long-distance optical cables		direct ratio
		Number of Internet broadband access households		direct ratio
		Number of computers used at the end of the period	One period indicates one year	direct ratio
AI application achievements	Transformation of basic research and development in artificial intelligence	Number of AI patent applications	Number of patent applications that mention the keywords “AI” or “artificial intelligence”	direct ratio
	Output capability of AI technology products	Revenue of software and information technology service industry		direct ratio

. We can take Beijing as an example, and the detailed data in Beijing are listed in Appendix A.

4.2. Data Source

This paper used panel data from 30 provinces in China from 2015 to 2021. Although China carried out a two-year lockdown during the COVID-19 pandemic, 2020 and 2021 are the latest years to have complete data, and trimming the samples may compromise the timeliness of the results. Additionally, China has mitigated the adverse impact of the nationwide lockdown through “internal circulation”, so the data from 2020 and 2021 are still referential. Considering the availability of the data, the regions covered in this article include 30 other provinces and regions in China apart from Tibet, Hong Kong, and Macao. The data were mainly sourced from the Chinese statistical yearbooks (2015–2021) and provincial statistical yearbooks (2015–2021) by the China National Bureau of Statistics, the China National Intellectual Property Administration (CNIPA) website, the International Carbon Action Partnership Organization (ICAP) website, and the International Federation of Robots (IFR) website. Due to some missing data, the dataset was processed using the interpolation method.

4.3. Analysis of the Evolution in Time Dimension

We calculated the comprehensive scores of each region using VHSD and EM models, namely, the wholesale and retail high-quality development index (WRI) and AI development index (AII), which served as a basis for us to conduct a Spearman correlation test to determine the rationality of the model. The test results are shown in

Table 3. Results of Spearman’s correlation test for the “VHSD-EM” model (the table was created by the authors).

Spearman Correlation Coefficient	2015	2016	2017	2018	2019	2020	2021
WRI	0.846 ***	0.986 ***	0.908 ***	0.886 **	0.830 ***	0.904 **	0.841 **
AII	0.825 ***	0.837 **	0.980 ***	0.841 ***	0.858 ***	0.842 ***	0.879 **

Note: ***, ** represent the significance levels of 1% and 5%.

.

According to Spearman’s test results, the evaluation results are significant and have a strong positive correlation at the 5% significance level, so the results are considered to possess good consistency, and the “VHSD-EM” model constructed in this paper has good stability. Through the “VHSD-EM” model, the changes in the WRI and AII in each province from 2015 to 2021 are shown in Figure 2.

Firstly, the development of WRT and the development of AI exhibited upward trends. We use Pearson correlation analysis and find that at the significance level of 0.01, the Pearson correlation coefficient of AII and WRI is 0.665. As a result, there is a positive correlation between them, based on Figure 2, it can be concluded that there is strong consistency between AII and WRI. However, there is still a difference between the two curves. The WRI curve is more gradual, and the increase is slightly smaller than the AII curve. This is because it might take time for new technology to develop and be applied. The impact of technology on industry is characterized by hysteresis, and it will be clearer many years from now. Second, the growth of the AII in all provinces accelerated after 2019, and the starting point of growth occurred around 2018. In fact, China released a new generation of AI development in 2018, when AI shifted from being primarily focused on research and learning to being widely applied in the fields of production and circulation. Specifically, this process also provided AI technology with a massive source of data for accumulation. In order to obtain greater benefits, enterprises will increase investments, continuously innovate software and hardware facilities, and promote the growth of computing power. Throughout the process, AI algorithms have also been continuously improved and optimized. Ultimately, AI technology has achieved rapid development. Third, the two curves of the central and western regions are relatively flat, while the curve of the eastern region changes greatly, and its growth rate is greater than that of the central and western regions. This is similar to the economic development trends; regions with higher levels of development are more likely to leverage their existing advantages and achieve higher economic growth rates.

4.4. Analysis of the Evolution in Spatial Dimensions

In order to analyze the development of WRT and AI in various regions more intuitively, we obtained the average index of each province through the “VHSD-EM” model and compared them. At the same time, the scores for six dimensions of high-quality development of the WRT were added. The results are shown in

Table 4. WRI and AII in different regions and provinces (the table was created by the authors).

Province	WRI		Innovation		Coordination		Greenness		Openness		Sharing		Informatization		AII
	Score	Ranking	Score	Ranking	Score	Ranking	Score	Ranking	Score	Ranking	Score	Ranking	Score	Ranking	Score	Ranking
Shanghai	0.436	1	0.157	5	0.666	1	0.847	16	0.528	1	0.522	1	0.186	2	0.212	6
Guangdong	0.312	2	0.651	1	0.248	5	0.870	8	0.215	3	0.400	5	0.232	1	0.502	1
Beijing	0.280	3	0.153	6	0.367	2	0.869	9	0.286	2	0.460	2	0.175	4	0.375	2
Jiangsu	0.250	4	0.526	2	0.261	4	0.905	3	0.153	5	0.418	4	0.109	8	0.343	3
Zhejiang	0.218	5	0.401	3	0.201	6	0.912	2	0.152	6	0.328	8	0.176	3	0.224	4
Tianjin	0.191	6	0.086	14	0.287	3	0.880	6	0.176	4	0.430	3	0.113	7	0.064	18
Shandong	0.172	7	0.320	4	0.198	7	0.918	1	0.105	11	0.334	7	0.073	16	0.223	5
Fujian	0.148	8	0.151	7	0.191	8	0.905	4	0.136	7	0.335	6	0.099	11	0.117	8
Chongqing	0.126	9	0.080	15	0.141	13	0.727	24	0.130	9	0.213	11	0.102	10	0.064	17
Ningxia	0.116	10	0.009	28	0.059	29	0.125	30	0.107	10	0.068	29	0.136	5	0.010	30
Jiangxi	0.109	11	0.077	16	0.141	14	0.856	13	0.076	13	0.207	13	0.103	9	0.061	19
Hubei	0.108	12	0.131	10	0.155	10	0.849	14	0.042	19	0.247	9	0.072	17	0.115	9
Henan	0.103	13	0.140	8	0.147	11	0.866	11	0.039	22	0.217	10	0.055	24	0.114	10
Anhui	0.103	14	0.136	9	0.132	15	0.886	5	0.046	17	0.198	15	0.082	13	0.094	13
Hunan	0.101	15	0.119	11	0.141	12	0.872	7	0.058	14	0.210	12	0.062	21	0.097	11
Liaoning	0.100	16	0.074	17	0.098	23	0.766	22	0.134	8	0.160	18	0.066	19	0.095	12
Guizhou	0.091	17	0.028	23	0.158	9	0.668	25	0.009	29	0.092	27	0.035	30	0.041	26
Hebei	0.086	18	0.100	13	0.099	22	0.868	10	0.053	15	0.149	20	0.082	14	0.090	14
Sichuan	0.086	19	0.116	12	0.116	17	0.859	12	0.037	23	0.161	17	0.058	23	0.157	7
Hainan	0.086	20	0.005	29	0.096	24	0.842	17	0.098	12	0.110	24	0.074	15	0.012	29
Shaanxi	0.084	21	0.069	18	0.122	16	0.828	19	0.041	20	0.200	14	0.067	18	0.085	15
Xinjiang	0.080	22	0.014	27	0.055	30	0.615	27	0.045	18	0.067	30	0.117	6	0.043	24
Guangxi	0.077	23	0.031	20	0.090	25	0.833	18	0.052	16	0.136	23	0.087	12	0.061	20
Heilongjiang	0.072	24	0.030	22	0.110	18	0.570	28	0.039	21	0.174	16	0.053	25	0.058	22
Jilin	0.067	25	0.023	25	0.099	21	0.816	20	0.032	24	0.141	22	0.051	26	0.060	21
Yunnan	0.067	26	0.030	21	0.107	19	0.848	15	0.023	27	0.144	21	0.045	29	0.055	23
Nei Monggol	0.065	27	0.028	24	0.103	20	0.491	29	0.025	26	0.159	19	0.047	28	0.068	16
Shanxi	0.059	28	0.032	19	0.073	26	0.768	21	0.029	25	0.105	25	0.065	20	0.043	25
Gansu	0.054	29	0.018	26	0.072	27	0.753	23	0.011	28	0.101	26	0.051	27	0.036	28
Qinghai	0.052	30	0.003	30	0.069	28	0.625	26	0.002	30	0.075	28	0.058	22	0.036	27

.

The following conclusions can be drawn from Table 4:

Firstly, WRI shows a trend of rising in the east and lowering in the west, which is consistent with the level of economic development. The development level of the tertiary industry in coastal areas is relatively high, with Shanghai having the highest WRI, followed by Guangdong, Beijing, Jiangsu, and Zhejiang. There are significant internal differences in rankings in the central region, and there is a trend of polarization. Provinces such as Tianjin and Shandong rank second only to some economically developed areas along the southeast coast, but some central provinces rank relatively low, such as Shanxi, the WRI of which is only 0.059. Northwest provinces have low WRI values, and there were few differences in these regions. The development level was not high, and provinces such as Gansu and Qinghai ranked last.

Second, the scores of the six dimensions can imply the advantages and shortcomings of WRT in various regions. For Guangdong, Jiangsu, Shandong, Fujian, and most of the other eastern coastal areas, lack of informatization has become the most important factor restricting the high-quality development of the local WRT. The layout of information infrastructure and the popularization of technology application in these provinces do not match the level of economic development, and fail to meet the needs of regional WRT development. The weakness of regions with higher economic levels, such as Shanghai and Beijing, lies in innovation. Although these regions have sufficient digital foundations, their lack of innovation capability will hinder the emergence of new products and services in the traditional retail industry, and will hinder their development of more innovative and modern characteristics. The weakness of regions such as Jiangsu and Hebei, where the economy is in a stage of rapid development, lies in coordination. There is a certain gap in the level of consumption between industries and urban and rural areas within the region. Chongqing, Liaoning, and other industrial developed areas have low greenness scores, and pollution emissions have brought negative effects to the retail industry. Guizhou, Shanxi, Qinghai, and other inland provinces have low degrees of internationalization, and factors such as terrain, regional location, and the degree of economic development may limit the development of regional foreign circulation trade. The weakness of Zhejiang, Tianjin, Jilin, and other places lies in sharing; it is difficult for the development of WRT to benefit everyone. Additionally, there are deficiencies in transport infrastructure construction and employment opportunities.

Thirdly, as can be seen from the AII, the AI development index in the southeastern coastal areas ranks among the top in China, with Guangdong being the top-ranking province, with a score of 0.502. The overall score in Northern China is relatively high, followed by the Central South and northeast regions, with the lowest score being found in the northwest region. At the same time, there is a significant difference in the AI development index between Guangdong, which ranks first, and Beijing, which ranks second, with a difference of 0.127, indicating that the development of AI is unbalanced among regions, and Guangdong is significantly stronger than other regions in China.

Fourth, the ranking of WRI and AII is almost consistent across provinces. However, there are also differences, as can be seen by the fact that the ranking of WRI in Shanghai, Tianjin, Guizhou, and other regions is significantly higher than that of AII, which can be attributed to the developed industries such as tourism driving the market demand for retail goods. This has manifested the ability to promote the prosperity of local industries. In some provinces, like Sichuan, the ranking of AII is significantly higher than that of WRI. These regions have a solid economic foundation and the ability to develop cutting-edge technologies such as AI, but their development ability is still limited. These technologies are mostly used in high-end manufacturing or research and development fields, and the degree of technological implementation is far from high. In most of these regions, the economic structure has been dominated by heavy industry that suffers from a lack of innovation. The environmental protection requirements and the fierce competition in the domestic and international markets have led to uncompetitive retail products. At the same time, the brain drain in these regions is severe, so the WRT markets are relatively small. As a result, their applicability in the basic circulation field is weak.

5. Analysis of the Influence of AI on the High-Quality Development of WRT

5.1. The Direct Impact of AI on the High-Quality Development of WRT

The random forest algorithm is one of the most commonly used algorithms in machine learning. Its purpose is to judge the influence degree of each index on the target variable [39]. Choosing WRI as the target value, the influence of AI on the high-quality development of WRT was studied using the random forest algorithm. The development level of AI is not the only variable that affects the development of WRT, and the influence of other related factors cannot be abandoned when studying this problem. The interaction of AI with other factors will also bring new impetus to the industry. Therefore, in reference to existing studies [40,41], this paper selects 10 additional variables as characteristic variables. These variables mainly cover the following five aspects: scientific and technological development level, socio-economic development, government policy support, foreign trade, and financial development level. The specific explanations are as follows: the development level of science and technology refers to the turnover of the technology market and the number of patent applications. Socio-economic development includes the employment situation and the enterprise development situation, including employment density, the number of enterprises above the designated size, and the proportion of retail employees. The employment density is the ratio of the number of people employed to the land area, and according to the China National Bureau of Statistics, industrial enterprises with an annual output value of over CNY 20 million and commercial enterprises with an annual turnover of over CNY 2 million are deemed as enterprises above the designated size. Government policy support is measured by the degree of government intervention, calculated as the proportion of government budget expenditure to GDP. Foreign trade is measured by the proportion of foreign investment, calculated as foreign direct investment divided by GDP. The level of financial development includes three indicators: the ratio of deposit balance of financial institutions to GDP at the end of the year, the ratio of loan balance of financial institutions to GDP at the end of the year, and the scale of social financing. After determining the target value and characteristic variables, the model was trained and learned using the random forest algorithm. The contribution rate of each characteristic variable to the high-quality development of WRT is shown in

Table 5. Contribution rate of AII and other characteristic variables and interaction items (the table was created by the authors).

Interactive Item Not Added		Add Interactive Item
Variable	Contribution Rate	Variable	Contribution Rate	Interaction Item	Contribution Rate
AII	38.70%	AII	15.70%
The proportion of foreign investment	5.30%	The proportion of foreign investment	1.40%	AII × The proportion of foreign investment	4.40%
Technology market turnover	1.20%	Technology market turnover	3.20%	AII × Technology market turnover	7.20%
Number of domestic patent application granted	1.90%	Number of domestic patent application granted	1.20%	AII × Number of domestic patent application granted	5.50%
Number of enterprises above the designated size	5.30%	Number of enterprises above the designated size	0.50%	AII × Number of enterprises above the designated size	1.20%
The proportion of wholesale and retail employees	7.60%	The proportion of wholesale and retail employees	0.40%	AII × The proportion of wholesale and retail employees	15.20%
The ratio of loan balance of financial institutions to GDP at the end of the year	2.60%	The ratio of loan balance of financial institutions to GDP at the end of the year	0.80%	AII × The ratio of loan balance of financial institutions to GDP at the end of the year	6.20%
Government intervention	3.00%	Government intervention	1.30%	AII × Government intervention	0.40%
Employment density	26.40%	Employment density	3.20%	AII × Employment density	16.20%
Scale of social financing	2.30%	Scale of social financing	0.80%	AII × Scale of social financing	4.50%
The ratio of deposit balance of financial institutions to GDP at the end of the year	5.70%	The ratio of deposit balance of financial institutions to GDP at the end of the year	3.50%	AII × The ratio of deposit balance of financial institutions to GDP at the end of the year	7.00%

.

The results demonstrate that the contribution rate of AI reached 38.70%, indicating that AI could promote the development of WRT. Secondly, the remaining 10 characteristic variables also played a role in promoting the high-quality development of WRT. Among them, employment density was the most important, with the contribution rate reaching 26.40%. Secondly, the contribution rate of the ratio of the deposit balance of financial institutions to the GDP at the end of the year reached 7.60%. The two indicators with the highest contribution rates are both employment-related indicators, indicating that, as one of the traditional labor-intensive industries, WRT is still inseparable from the employment situations of the relevant personnel. In addition, the ratio of the deposit balance of financial institutions to the GDP at the end of the year also had a great impact on the development of WRT. In comparison, other indicators, especially those related to scientific and technological development and patent application, played small roles.

In reality, various economic factors do not exist independently, and there may be interactions and connections between AI and other indicators. Therefore, there will also be a mesomeric effect and a regulatory effect. Succinctly put, the influence of AI on industrial development may be disturbed by other factors, and vice versa. This effect is generally verified by adding interaction terms to the model, but prior to that, the contribution rate obtained by the random forest algorithm refers to the total contribution rate of the characteristic variable itself and the interaction with other variables. The closer the relationship between a certain variable and other characteristic variables, the greater the importance of the interaction term, which will also lead to a decrease in the contribution rate when the variable is completely independent. After adding the interaction item, the contribution rate of AI decreased from 38.70% to 15.70%, and the contribution rate of the interaction item with other factors reached 67.80%, indicating that in addition to the direct impact, AI also had a strong indirect impact on the high-quality development of WRT through its interaction with other factors. This also confirms the previous view that the role of AI in promoting WRT development may have different effects due to other factors.

The contribution rate of other characteristic variables also showed some changes after the addition of interaction terms. For the interaction items between AII and characteristic variables, the contribution rates of employment density and the proportion of wholesale and retail employees were relatively high, reaching 16.20% and 15.20%, respectively, while the direct contribution rates of these two variables decreased from 26.40% and 7.60% to 3.20% and 0.40%, respectively. The proportion of foreign investment and the number of enterprises above the designated size saw a significant decrease in their contribution rates after including interaction items, with contribution rates dropping from 5.30% to 1.40% and 0.50%, respectively. This indicates that improving employment levels has a significant driving effect on promoting the high-quality development of the WRT, both in terms of its own individual impact and its interaction with AI. The influences of government intervention, social financing and other factors cannot be ignored; nevertheless, these variables will have a stronger positive effect if combined with AI. In contrast, the number of enterprises, the number of patent applications, and the utilization of foreign capital have less significant impacts on the development of WRT.

5.2. The Impact Path of AI on the High-Quality Development of WRT

5.2.1. Univariate Partial Effect

The partial effect reflects the magnitude of the impact of characteristic variables in different states on the high-quality development of WRT. This paper analyzes whether the economic factors of each province reach or deviate from optimal points through the partial effect function, and explores the best path for the high-quality development of WRT in each region with the contribution rate obtained by the random forest algorithm. Figure 3 shows the partial effect of AI on the high-quality development of WRT as a single variable. The figure shows some provinces which are closer to the turning point.

As can be seen in Figure 3, there are three obvious turning points in the partial effect diagram. Turning point A is located at a lower point of AII. Although the slope near this point is relatively high, with Ningxia and Hainan just before point A, the development of AI is relatively low, so it can hardly promote the development of WRT. Turning point B is located at the point where the curve rises the most quickly, and when AII reaches this point, it makes its greatest contribution to the WRT. Hubei, Henan, and other provinces have crossed turning point A and are situated between the two turning points A and B, and these provinces have a certain foundation in AI technology. However, their overall levels are not high enough to maximize the scale effect brought on by AI technology, and they have not promoted large-scale industrial structure optimization nor the emergence of new industries through technological advantages. These provinces need to continue to increase their investments in science and technology development and improve their own levels of AI technology so as to give full thrust to the role of science and technology in promoting the WRT. The region between turning points B and C is the most rapidly growing stage involved in the promotion effect of AI on the development of the retail industry. It is also the target stage for the provinces before turning point B. Sichuan and Shandong are in this stage, with the scale effect of AI receiving more full play and serving as an important driving force for development that constantly promotes the emergence of new forms. The influx of knowledge elements in the WRT maximizes production efficiency and promotes interconnectivity between industries, with mature infrastructure also providing convenience for the application of digital technology. This stage can be considered as a golden period during which AI can maximize its functions. The partial effect curve at turning point C tends to be horizontal, meaning that at this time, although the development level of AI has reached a high level, the scale effect can no longer continue to play its role. Zhejiang, Shanghai, Jiangsu, Beijing, Guangdong, and some other provinces have crossed turning point C, at which the promotion effect of AI on the WRT has encountered bottlenecks; thus, provinces in this region need to change their development strategies and make more substantial and sustainable changes in AI technology to meet the development needs of the new phase.

5.2.2. Bivariate Partial Effect

In order to combine various elements and to explore a more specific and scientific path for the high-quality development of WRT, this paper adds interaction terms in order to draw a bivariate partial effect diagram of AII and other characteristic variables. We selected four variables with the highest contribution rates of the interaction terms with AII, namely, employment density, proportion of wholesale and retail employees, technology market turnover, and the ratio of deposit balance of financial institutions to GDP at the end of the year, to visualize the results. The actual observation data of each province in 2021 are labeled in the graph, and some provinces that were closer to the optimal investment point are indicated. The results are shown in Figure 4.

In Figure 4, starting from the respective dimensions of AII and the characteristic variable, we found the corresponding maximum point of the partial derivative. The joint partial effect at the intersection of the maximum partial derivative of two elements was the largest; this was the optimal investment point. We used this point as the origin, created a planar coordinate system, and divided each bivariate partial effect diagram into four parts, marking the actual observation values of each province in the diagram with dots. Provinces located in the first quadrant have crossed the optimal investment points for their AI technology and characteristic variables. In the second quadrant, only the characteristic variables of provinces have reached the optimal investment points. Neither AI nor characteristic variables of provinces in the third quadrant have reached their optimal investment points. For the provinces in the fourth quadrant, only AI has reached the optimal investment point.

Considering the interaction with other characteristic variables, AII, in most other provinces except Guangdong and Beijing, has not exceeded the optimal investment point; that is, the development of AI in each province in China is still limited. Shanghai, Jiangsu, Beijing, and some other provinces have crossed the optimal investment point where simple element investments have been unable to provide the greatest support for regional WRT development, and increasing factor input is not the most effective development strategy. Thus, these provinces need to change their strategies, adjust the industrial structures, and provide policy support to fuel new economic growth. They should upgrade the factors and shift from increasing the quantity to improving the quality. With the existing technological advantages, these regions could improve the agglomeration ability of high-end elements and the efficiency of high-tech factor resource allocation.

In Figure 4a, it is shown that the employment density in provinces such as Zhejiang, Sichuan, and Tianjin has not reached the optimal input point, and these regions have relatively mature technological endowments and economic foundations. At this time, it is necessary to vigorously promote the development strategy of actively attracting talent and encouraging employment in order to maximize the impact under the interaction with AI.

In Figure 4b, it is shown that the proportion of wholesale and retail employees in Shandong, Henan, Sichuan, and other central and southwest regions have not reached the optimal investment point, and the developments of WRT in these regions do not match their own economic foundations. The development of WRT lags behind the overall level of economic development, and those with relevant abilities tend to flow into other industries. This is in line with the fact that north and southwest China are in a period of rapid economic growth, and AI achievements are mainly concentrated in advanced fields and technologies that are difficult to implement. In order to meet the needs of economic development, these regions tend to invest limited resources into the high-tech field, first developing high-tech industries and then driving the development of basic industries such as WRT, which temporarily creates a shortage of employment in the WRT.

In Figure 4c, it is shown that the technology market turnover in Jilin, Liaoning, Heilongjiang, Qinghai, Ningxia, Xinjiang, and other northeast and northwest provinces has not reached the optimal investment point. This situation is caused by local development policies. Industries in the northwest are dominated by traditional agriculture and the demand for digital technology applications is low. The modern industry and the service industry have small development spaces, resulting in limited development of high-tech industries. At the same time, due to the limitations of elements such as capital, science, and technology, much network infrastructure and data-intelligent infrastructure construction is incomplete, so the short boards restrict the development of AI and are averse to the application of digital technology in production. There is also a shortage of investments in technology market transactions in northeast regions such as Jilin and Heilongjiang, which may be related to the small population base and the brain drain in the region, resulting in insufficient endogenous growth momentum and a lack of vitality in the development of circulation enterprises and making it difficult to develop large-scale technology transactions.

In Figure 4d, it is shown that the ratio of the deposit balance to the GDP of financial institutions at the end of the year in Xinjiang and Yunnan did not reach the optimal investment point, demonstrating the phenomenon of insufficient social financing and low ability to attract financial support. These provinces are located in remote inland areas, and the unfavorable terrain conditions increase the restrictions on trade with neighboring countries, hinder the dissemination of knowledge and the application of technology, and lack some of the financing channels needed to attract and cultivate talents and to build AI infrastructures. In order to improve the development of AI and WRT, it is necessary to understand the restrictive factors of different regions, solve the key problems, and accelerate the application of AI, virtual reality, and other digital technologies in the process of WRT.

6. Discussion

6.1. Summary of the Study

With the advancement of technology, AI is increasingly penetrating into the social economy, and there are also more connections between WRT and AI [42]. There are still shortages of evaluation index systems for AI used in WRT, although scholars have established index systems for the overall evaluation of AI [38]. Based on this research, this paper built an evaluation index system for WRT, including innovation, coordination, greenness, openness, sharing, informatization, and an evaluation index system for AI. This included AI infrastructure construction, AI application capabilities, and AI application achievements. Semenov et al. [43] posited that AI, a modern analysis tool for consumer demand in retail trade, would bring maximum profit to WRT. This paper confirms that AI expands the positive externality of WRT, but there is a time lag between them because it takes time to realize the knowledge spillover, industrial transformation, and environmental improvements. Zou et al. [30] believed that, in cities with high degrees of marketization and Internet development, the role of AI in promoting industrial upgrading can be strengthened. We confirm that there are regional differences in the impact of AI on WRT. In addition, in this paper, we selected other factors that interact with AI and found that the synergistic effects of different factors and artificial intelligence vary. When some factors are underdeveloped, AI has difficulty promoting the high-quality development of WRT.

6.2. Conclusions

This paper draws the following main conclusions by studying the impact of AI on the high-quality development of WRT:

First, there are spatial differences in the development of AI and WRT among different regions, but they are consistent with the social economy development level of the region. However, the impact of AI on the industry has a certain time lag.

Second, the shortcomings of the high-quality development of the WRT vary in different regions. In general, the degree of informatization in eastern coastal areas does not match the level of economic development. The weakness of provinces of higher economic levels in Northern China lies in innovation and sharing, while the weakness of regions with rapid economic development, i.e., the central and southeastern regions, lies in coordination. Environmental pollution problems in some industrial developed provinces in the northeast and southwest regions have brought about negative effects on the development of WRT, while the internationalization levels of inland remote areas in the southwest and the northwest are relatively low.

Third, AI and other factors produce synergy, promoting the high-quality development of WRT. Before adding interaction items, the direct contribution rate of AI to the development of WRT reached 38.70%. After adding interaction terms, the effects of element interactions were also taken into account. The direct contribution rate of AI elements was reduced to 15.70%, but the contribution rate of interaction terms was actually higher. When it interacts with four factors—employment density, proportion of employees in the WRT, technology market turnover, and the ratio of the deposit balance of financial institutions to the GDP at the end of the year—it will have the greatest indirect impact on the WRT.

Fourth, there are also spatial differences in the influence path of AI on the high-quality development of WRT. AI technologies in economically developed areas such as Shanghai and Guangdong are widely used, but in other provinces, the AI elements have not reached the optimal input point, which indicates that AI still does not have widespread applications in modern industry, and there still exists significant room for expansion in its utilization scenarios in the central and western regions. In addition, the insufficient investment in regional science and technology will restrict the application of AI and induce the region to create a suitable development environment for the digital economy.

6.3. Limitations and Future Research

The research of the effect of AI on the development of China’s WRT is still facing limitations. Many retail companies view AI algorithms as core business secrets, which makes it difficult to obtain direct data. The theoretical system is not perfect, so there is still a lack of a comprehensive indicator system for AI technology in WRT. In addition, it could be detrimental to allow the creation of AI applications to progress without any oversight [44,45]; however, this paper mainly discusses the positive impact of AI on the WRT. In future research, efforts can be made in the following three areas: First, we will attempt to obtain some internal indicators of enterprise management and establish a more comprehensive evaluation index system. Second, we will select specific regions as research objects to make the data more available. Third, we will attempt to use other methods to investigate the negative impacts of AI in order to make our conclusion more complete.