Evaluating Regional Potentials of Agricultural E-Commerce Development Using a Novel MEREC Heronian-CoCoSo Approach

Abstract

In terms of both economy and sustainability, rural areas can greatly benefit from adopting E-commerce. The Chinese government is currently devoting significant efforts to developing agricultural E-commerce. However, one of the most significant problems is the lack of effective tools for evaluating regional potentials in this regard, possibly leading to inappropriate policymaking, investment allocation, and regional planning. To address this issue, this study proposes a novel and effective method for evaluating regional potentials for agricultural E-commerce development, integrating the method based on the removal effects of criteria (MEREC), Heronian mean operator, and combined compromise solution (CoCoSo) method. The method’s effectiveness is then tested and confirmed in the Chinese city of Yibin through an evaluation of its ten regions. The results suggest that such a method is robust, objective, and able to consider indicator interactions effectively. By applying this method, regional agricultural E-commerce development potentials can be thoroughly evaluated and ranked. This study contributes to the literature by providing new analytical techniques for agricultural studies, as well as by supporting political and investment decision-making for governments and E-commerce practitioners in the agriculture sector.

1. Introduction

1.1. Background

With the development of digital technology and the popularity of mobile phone usage, E-commerce has become one of the most important selling channels in the retail industry. Transaction costs for agricultural products in particular can be significantly reduced by E-commerce [1,2], and intermediate companies like distributors and wholesalers can be removed in the process of online selling. For example, according to a report from Yangcheng Evening News, the Chinese E-commerce giant Pinduoduo launched an event to help farmers sell their products by simplifying logistics, shortening the agricultural supply chain, and providing farmers with E-commerce training. In this way, farmers were directly able to conduct online business with the end customers and earn a higher income, as they were able to make significant savings in terms of distribution and wholesale costs (https://new.qq.com/rain/a/20201216A0ABLX00, accessed on 30 July 2024). Therefore, well-developed agricultural E-commerce can enable farmers and rural regions to acquire higher incomes and reduce poverty, leading directly to enhanced regional economy and social welfare [3,4].

Due to its importance, the Chinese government is currently investing great efforts into developing agricultural E-commerce [5,6,7]. For example, in March 2024, the Chinese government implemented the promotion of high-quality rural E-commerce development; various support will be offered to facilitate its growth, such as providing E-commerce training to farmers, developing logistics systems for agricultural products, and upgrading E-commerce-related infrastructures in rural areas.

However, for these efforts to be effective, first and foremost, the government must evaluate regional potentials for developing agricultural E-commerce [5]. By doing this, they can select the regions with the highest potentials. Otherwise, if certain regional resources are not suitable for agricultural E-commerce, these efforts will be wasted. Because agricultural E-commerce operations involve various activities ranging from agricultural product supply, delivery, sales, technology usage, workforce recruitment and training, etc., multiple aspects should be considered in order to effectively evaluate and select potential regions. This makes evaluating regional potentials for developing agricultural E-commerce a multi-criteria decision-making (MCDM) task. Previous studies in the literature have explored different evaluation methods (e.g., [4,5,6]). However, explorations in this regard are still in their infancy. For example, the objective indicators in the evaluation system have not been fully considered in the literature, and the evaluation approaches do not capture the interactions between indicators, which can make the evaluation results less robust or even biased.

Therefore, in order to properly and thoroughly evaluate the regional potentials of agricultural E-commerce development and render governmental efforts more effective, more objective and robust evaluation frameworks considering indicator interactions should be proposed. To address this research gap, in this study, we aim to answer the following question: how can regional agricultural E-commerce development potentials be properly evaluated? To solve this question, we developed an objective index system and proposed a novel MEREC Heronian–CoCoSo framework to evaluate and select regions. Using empirical regional data from the Chinese city of Yibin, whose principal industry is agriculture, we thoroughly tested our model. The results show that our model has very good capability in terms of evaluating agricultural E-commerce development potential.

We believe that our study offers the following contributions. Academically speaking, our study proposes a novel MCDM framework for evaluating regional agricultural E-commerce development potential. Compared with the existing literature, our framework can outperform others by fully and objectively capturing the importance of indicator interactions. Additionally, we developed an index system for regional agricultural E-commerce development potentials, paving the way for future studies in relevant fields to objectively evaluate policy or financial stimulus effectiveness.

Practically speaking, policy makers and E-commerce companies can benefit from our study. Our study can support governments in designing better agricultural policies for different regions and properly allocating budgets for public expenditure in order to stimulate agricultural E-commerce development. Additionally, E-commerce companies, especially those who tend to enter the agricultural sector, can apply our study to build their own analytical tools and make better strategic plans.

1.2. A Brief Introduction to Yibin

In this study, we proposed a new MCDM method and tested its capability in the case of Yibin city. Yibin is located in the southern part of Sichuan province, China. It consists of ten different regions (i.e., Cuiping, Nanxi, Xuzhou, Jiang’an, Changning, Gongxian, Junlian, Xingwen, and Pingshan). Figure 1 presents geographical information on the city and its ten regions derived from map.baidu.com (accessed on 26 July 2024) and ditu.amap.com (accessed on 26 July 2024).

Due to its suitable water resources, weather, and ecological environment, it is a good place for agriculture. The city is famous for its agricultural products such as liquor, fruit, and tea. For example, Yibin’s Da Ta Lychee, Jiang’an Big White Plum, Yibin Lemon, etc., are well known in China. However, according to government statistics (http://tjj.yibin.gov.cn/, accessed on 2 July 2024), the annual disposable income per capita in some regions is lower than 35,000 RMB (lower than 4900 USD), which remains a relatively low level. To enhance the region’s income and welfare, in recent years, the local government realized the importance of and started to develop agricultural E-commerce. These efforts soon saw positive returns. For example, due to E-commerce, the local plums become popular online, with sales increased through E-channels (https://baijiahao.baidu.com/s?id=1772902101167733813&wfr=spider&for=pc, accessed on 31 July 2024).

However, although multiple attempts have been made to develop agricultural E-commerce, some of them may not obtain effective results. According to a local news report, E-commerce in Yibin still faces obstacles, such as a shortage of talent, insufficient logistics networks in multiple areas, and occasionally instable product quality in certain regions (https://www.163.com/dy/article/HUOSGNSI0514AU89.html, accessed on 31 July 2024). Therefore, to make local efforts more effective, the first step is to select the regions with the highest potentials for developing agricultural E-commerce. This study aims to fulfill this need.

1.3. Literature Review

1.3.1. Agricultural E-Commerce

In recent years, agricultural E-commerce has been investigated from multiple angles. For example, some studies examined it from a supply chain perspective in order to identify appropriate decisions for different stakeholders in agricultural E-commerce. For example, the authors in [7] explored the governments’ subsidy plans for agricultural products sold on E-commerce platforms. Using game theory, they developed a supply chain model, solving optimal subsidy decisions and finding that subsidies for farmers can work better than those for customers or than in no subsidy scenarios. The authors in [8] analyzed the barriers to effective agricultural E-commerce management from a supply-chain perspective using big data and MCDM techniques, proposing a new management mode to enhance agricultural E-commerce performance. Other authors are also dedicated to studying the technological developments of agricultural E-commerce. In [9], the authors conducted a systematic literature survey of the recent technological advancements in agriculture E-commerce. In their study, the challenges faced by the industry were summarized, followed by possible solutions via adopting the latest technologies. The authors in [10] focused on farmers’ intentions to adopt live-streaming in agricultural product sales. Drawing on the technology acceptance model, they found that both government and platform support and social learning can lead to increased perceived usefulness and ease of use, which would eventually enhance farmers’ intentions to adopt live-streaming technology. In addition, published studies have explored the economic and environmental value of developing agricultural E-commerce. In [11], the authors investigated the value of agricultural E-commerce in reducing pollution. Adopting a quasi-experimental approach, they found agricultural E-commerce can decrease levels of non-point source pollution and enhance economic growth and technological development. The authors in [12] studied the factors influencing purchase intentions regarding agricultural green products through live-streaming E-commerce. They identified five quality measures for live-streaming sales and found that all of them are significantly related to customer trust and purchase intentions in terms of green agricultural products.

Among the above topics, one of the most important aspects is evaluation, i.e., of agricultural E-commerce suppliers [13], distribution modes [14], or vehicle routes [15]. Recently, there have been a few studies establishing evaluation systems for measuring regional potentials for developing agricultural E-commerce [5,6]. Such explorations are, therefore, still in their infancy, and studies usually adopt qualitative indicators in their measurement system. Therefore, a more robust and objective evaluation system with richer criterion information needs to be developed in order to provide more accurate measurements. This paper aims to address this gap and contribute to the existing literature by proposing a new evaluation system combining the MEREC, Heronian operator, and CoCoSo models.

1.3.2. CoCoSo Model and Its Extension

The CoCoSo model was initially developed by [16]. By combining three compromised values, the model can provide very stable results for ranking alternatives in MCDM problems. In recent years, CoCoSo’s potential has been confirmed across multiple fields, such as supplier management [17], manufacturing [18], construction [19], etc. Additionally, multiple extensions have been developed to enhance the ability of CoCoSo to cope with different decision-making scenarios. For example, the authors in [20] combined the entropy method with CoCoSo and proposed an objective weighting framework to evaluate road transport sustainability. In [17], the authors extended the real number CoCoSo model to a fuzzy number model, establishing a supplier selection protocol. The authors in [21] extended the CoCoSo model to the hesitant Fermatean fuzzy context and proposed an evaluation framework for blockchain platforms. In [22], the authors combined the SWARA method with the CoCoSo model and built a location selection model for logistics centers.

Among all the studies in the literature, extending the CoCoSo model to include a weighting approach has attracted much attention, as appropriate criteria weighting can significantly enhance the accuracy of alternative selection. To obtain a rigorous weighting process, the MEREC method developed by [23] has been integrated into the CoCoSo model to calculate objective weights for criteria established in recent studies. For example, the authors in [24] proposed a MEREC–CoCoSo method to evaluate the financial performance of hospitals, and those in [25] integrated the methods for robot selection. Some authors extended the hybrid MEREC–CoCoSo model to fuzzy contexts, applying it to offshore wind turbine selection [26], urban transport planning [27], and waste material recycling [28]. As can be observed, MEREC has great potential in MCDM and can significantly enhance the capability of CoCoSo in problem evaluation due to its objectivity and stability in weighting processes.

In the meantime, the interactions between different criteria in MCDM have drawn increasing academic attention. This has led to recent publications integrating the Heronian mean operator with the CoCoSo model to enable a full consideration of criterion interactions. For example, the authors in [29] proposed a framework integrating a fuzzy power Heronian operator with the CoCoSo model in order to rank autonomous vehicles; the authors in [30] applied a combination of the Heronian operator and CoCoSo for stock portfolio selection. However, as can be observed in the existing literature, the Heronian–CoCoSo method is still in its infancy and has not yet taken full advantage of the MEREC method in MCDM applications. To respond to this gap and fully realize its potential, the model should be extended. Therefore, this study innovatively integrates the MEREC with Heronian–CoCoSo to propose a novel approach, enabling a robust and objective evaluation that captures criterion interactions.

1.3.3. Research Gap Summary

Based on the above literature, the research gaps can be summarized as follows. First, although there are several studies focusing on evaluating the regional potentials of agricultural E-commerce development, they adopted qualitative indicators without fully considering the quantitative properties of the measurement system. However, according to the previous literature related to agricultural study [31,32], quantitative indicators and index systems produce comprehensive and rational evaluations of agricultural systems, thus supporting better decision-making for agricultural development. Therefore, failing to construct a quantitative index system can probably undermine the inclusion and robustness of the evaluation processes. In addition, the existing evaluation models are weak in terms of considering objective indicator weighting and indicator interactions. However, the previous literature demonstrates that the objective weighting process can mitigate possible decision bias or expert subjectivity [23,25], while indicator interactions can capture richer information in order to support decision-making [30]. Therefore, in agricultural E-commerce development research, failing to consider objective weighting and indicator interactions can produce possible evaluation bias.

Therefore, to fill the above gaps and better evaluate the potentials of agricultural E-commerce development, the objective of this paper is as follows. Firstly, we aim to develop an objective index system whose indicators are all quantitative data. By doing this, we intend to eliminate the ambiguity and subjectivity of the previous literature, mitigating any possible decision bias and thus providing more solid and informative decision support for agricultural E-commerce practitioners and policy makers. Additionally, we attempt to develop a more robust and stable evaluation model by integrating MEREC weighting. Using the Heronian operator and the CoCoSo model, we will better realize the potential of CoCoSo in MCDM and offer new analytical tools for future research in relevant fields.

2. Methods

In this section, we introduce the methods, including the development of the index system and establishment of the MEREC Heronian–CoCoSo model. The methodological flow is visualized in Figure 2. There are three main components in our methodological framework, namely, index system development, MEREC calculation, and Heronian–CoCoSo modeling. This means that our framework, just like other MCDM frameworks [16,26,30], generally entails three main procedures, including decision matrix formation, criteria weight assignment, and alternative ranking and selection. First, the index system developed enables us to specify the candidates considered in our framework, as well as their criterion values. The matrix made up by both candidates (usually in the first column) and criteria (usually in the first row) is the matrix for decision-making. After that, to ensure a fair comparison among candidates, appropriate criterion weights for candidates should be determined and assigned, as is addressed by MEREC. Finally, a comparison covering multiple criteria is conducted to calculate the overall performance of each candidate, the rankings of which can be derived to support alternative selection.

2.1. Index System Development

To properly evaluate the potential of developing agricultural E-commerce in different regions, the first step is constructing an index system. In this study, we applied a three-step approach. First, we searched the relevant literature for indicator selections. After that, by considering the properties of the regional agricultural E-commerce, we built the index system by analyzing the local annual statistics yearbook. After that, based on the existing literature and data availability, and while carefully justifying indictor appropriateness, we eventually selected 11 indicators to form the final index system for measuring regional potentials for developing agricultural E-commerce; this is presented in

Table 1. Indicator abbreviations of index system.

Indicator Abbreviations	Definition	Effect
OF	Gross Output Value of Farming (10,000 RMB)	Beneficial
CA	Cultivated Area (10,000 Hectare)	Beneficial
GA	Government Expenditure for Agriculture, Forestry, and Water Conservancy (10,000 RMB)	Beneficial
GT	Government Expenditure for Transport (10,000 RMB)	Beneficial
HM	Highway Mileage (KM)	Beneficial
NP	Number of Postal Offices (Unit)	Beneficial
NM	Number of Mobile Telephone Subscribers (10,000)	Beneficial
NB	Number of Internet Broadband Subscribers (10,000)	Beneficial
GS	Government Expenditure for Science and Technology (10,000 RMB)	Beneficial
GE	Government Expenditure for Education (10,000 RMB)	Beneficial
NE	Number of Employed Persons (10,000)	Beneficial

. All the indicators are beneficial, meaning that higher values can enhance regional agricultural E-commerce development potential.

Here, we have also provided justifications and supporting studies for the selection of each factor. First, the previous literature suggested that agricultural E-commerce adoption and development can be linked to the development level of the agricultural industry [33]. It can be reasonably inferred that the more developed the regional agricultural industry, the more cost and quality benefits for E-commerce sellers of agricultural products. Therefore, Gross Output Value of Farming, OF, and Cultivated Area (CA) were selected in the index system to reflect the current conditions of the regional agriculture sector. Furthermore, as Government Expenditure for Agriculture, Forestry, and Water Conservancy can be regarded as an important factor for enhancing the future development of agriculture, we also included GA in our index system.

The literature revealed that well-developed regional logistics can firmly support the development of agricultural E-commerce [33]. As agricultural E-commerce logistics include transportation and postal delivery, we selected Highway Mileage (HM) to reflect their current condition. Moreover, previous studies found that a well-designed and developed logistics network enhances delivery efficiency [34]; thus, we thus selected Number of Postal Offices (NP) for the index system. In addition, logistics performance can be leveled up by advanced transportation [35]. Government Expenditure for Transport can help in achieving this, thus justifying the selection of GT. Moreover, agricultural E-commerce company transactions are directly enabled by Internet and mobile communication tools [36]. Therefore, we selected the Number of Mobile Telephone Subscribers (NM) and Number of Internet Broadband Subscribers (NB) to reflect the business environments for agricultural E-commerce sellers.

Furthermore, the previous literature suggested that a lack of E-commerce talent can be a critical barrier to agricultural E-commerce development. Strengthening regional education can be a possible solution in this regard, which can also lead to increased agricultural E-commerce acceptance and adoption [37]. Therefore, we selected Government Expenditure for Education (GE) in our index system. Meanwhile, publications have revealed that technological advancement and the emergence of digital economy can effectively support the development of agriculture and agricultural E-commerce [38], thus justifying the selection of Government Expenditure for Science and Technology (GS) in the index system.

Finally, the previous literature has indicated that rural labor is an important factor in agricultural development [39]. Therefore, the large number of employed people in primary industry will contribute to the developmental potentials of agricultural E-commerce. Furthermore, as logistics and production are additional indispensable activities [40], the number of employed people in the secondary and tertiary industries is also connected to agricultural E-commerce development. Therefore, the Number of Employed Persons (NE) in all industries should be a useful indicator.

Based on the index system developed above, we collected the data for decision-making; these are publicly available and were taken from the Yibin Annual Year Book 2023, published by the Yibin Municipal Bureau of Statistics and accessible at http://tjj.yibin.gov.cn/sjfb/tjnj/202404/t20240409_1974463.html (accessed on 2 July 2024). Data related to all 10 different regions were collected, and each region’s potential for E-commerce development was measured according to the 11 criteria.

2.2. MEREC Procedures

We follow the previous literature (e.g., [23]) in conducting a MEREC analysis and deriving criterion weights. The steps of MEREC can be classified into six categories. The first step is the formation of a decision matrix, $x_{n\times m}$. There are $n$ alternatives evaluated using $m$ criteria, with the criterion value under each alternative as $x_{ij}$. Based on the data collected, the overall decision matrix for $x_{ij}$ is $x_{n\times m}$. After that, the second step is to derive the normalized matrix $n_{n\times m}^x$. According to [23], the normalization operations are conducted based on Equation (1), and $n_{ij}^x$ is the element of $n_{n\times m}^x$ with $i=1,2,3,\dots n$ and $j=1,2,3,\dots m$. Based on the normalized value, the third step is to calculate the overall performance of each alternative $S_i$ using the logarithmic relationship in Equation (2). The fourth step is to calculate the performance of each alternative by removing each criterion, which is $S_{ij}^{\prime }$ in Equation (3). After that, the fifth step is to add up the absolute deviation of each criterion, $e_j$. Finally, the criterion weights $w_j$ can be derived using Equation (5).

$$n_{ij}^x=\left\{\begin{array}{ccc}{\displaystyle \frac{\underset{j}{min}{(x}_{ij})}{x_{ij}}}& if& j\in \mathrm{b}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{l} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\\ {\displaystyle \frac{x_{ij}}{\underset{j}{\mathit{max}\left(x_{ij}\right)}}}& if& j\in \mathrm{c}\mathrm{o}\mathrm{s}\mathrm{t} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\end{array}\right.$$

(1)

$$S_i=ln\left(1+\left({\displaystyle \frac{1}{m}}\sum _j\left|ln\left(n_{ij}^x\right)\right|\right)\right)$$

(2)

$$S_{ij}^{\prime }=ln\left(1+\left({\displaystyle \frac{1}{m}}\sum _{k,k\ne j}\left|ln\left(n_{ik}^x\right)\right|\right)\right)$$

(3)

$$e_j=\sum _i\left|S_{ij}^{\prime }-S_i\right|$$

(4)

$$w_j={\displaystyle \frac{E_j}{\sum _k{E}_k}}$$

(5)

2.3. Heronian–CoCoSo Model

After deriving the MEREC weights, we constructed the Heronian–CoCoSo model. First, based on the decision matrix $x_{n\times m}$, we normalized the value by using the widely adopted approach (e.g., [41]) in Equation (6). It should be noted that such a normalization process is slightly different from that originally posed in the CoCoSo model [16]. The justification for modifying the normalization process is to eliminate the possible zero value in the normalized decision matrix. This is because the existence of a zero value will lead to the geometric Heronian mean being zero as well, which probably means that the subsequent calculation of the CoCoSo would contain less information. Therefore, to obtain more information in the final CoCoSo values, we chose the normalization approach in Equation (6).

$$r_{ij}=\left\{\begin{array}{ccc}{\displaystyle \frac{x_{ij}}{\underset{j}{\mathit{max}\left(x_{ij}\right)}}}& if& j\in \mathrm{b}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{i}\mathrm{a}\mathrm{l} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\\ {\displaystyle \frac{\underset{j}{\mathit{min}\left(x_{ij}\right)}}{x_{ij}}}& if& j\in \mathrm{c}\mathrm{o}\mathrm{s}\mathrm{t} \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\end{array}\right.$$

(6)

In this study, we extend the original CoCoSo model and calculate the arithmetic and geometric Heronian means as two important components for CoCoSo calculation. The arithmetic Heronian mean for each alternative is notated as $E_i$, while the geometric Heronian mean for each alternative is notated as $P_i$. By doing this, the interaction between different criteria can be fully considered [30,42].

$$E_i={\left({\displaystyle \frac{2}{m(m+1)}}\sum _{l=1}^m\sum _{j=l}^m{\left(m{w}_i{r}_{il}\right)}^p{\left(m{w}_j{r}_{ij}\right)}^q\right)}^{{\displaystyle \frac{1}{p+q}}}$$

(7)

$$P_i={\displaystyle \frac{1}{p+q}}{\left(\prod _{l=1}^m\prod _{j=l}^m\left(p(r_{il}{)}^{m{w}_i}+q(r_{ij}{)}^{m{w}_j}\right)\right)}^{{\displaystyle \frac{2}{m(m+1)}}}$$

(8)

The $p$ and $q$ are the parameters of arithmetic and geometric Heronian mean. Consistent with the previous literature, we assume that $p=q=1$ [42]. After that, according to [16], $E_i$ and $P_i$ were integrated into three values, $K_{ia}$, $K_{ib}$, and $K_{ic}$, as follows.

$$K_{ia}={\displaystyle \frac{P_i+E_i}{\sum _{i=1}^m\left(P_i+E_i\right)}}$$

(9)

$$K_{ib}={\displaystyle \frac{E_i}{\underset{i}{min}{E}_i}}+{\displaystyle \frac{P_i}{\underset{i}{min}{P}_i}}$$

(10)

$$\begin{array}{c}{K}_{ic}={\displaystyle \frac{\lambda E_i+\left(1-\lambda \right)P_i}{\left(\lambda \underset{i}{max}{E}_i+\left(1-\lambda \right)\underset{i}{max}{P}_i\right)}}\end{array}$$

(11)

In Equation (11), the value of $\lambda$ determines the reliability of the CoCoSo. In this study, this was set to 0.5 according to the previous literature [16,43]. Finally, the final CoCoSo value integrated $K_{ia}$, $K_{ib}$, and $K_{ic}$, as shown in Equation (12).

$$K_i=(K_{ia}{K}_{ib}{K}_{ic}{)}^{{\displaystyle \frac{1}{3}}}+{\displaystyle \frac{1}{3}}(K_{ia}+K_{ib}+K_{ic})$$

(12)

3. Results

3.1. Model Results for Potential Evaluation

In this section, the data processing will be presented with the final ranks of agricultural E-commerce development potentials for each region. The data collected based on the index system to evaluate the regional potentials of E-commerce agricultural development are presented in

Table 2. Decision matrix of agricultural E-commerce development potential evaluation.

Region	OF	CA	GA	GT	HM	NP
Cuiping	427,224	5.3	94,256	51,927	2888	669
Nanxi	441,712	2.52	125,920	10,832	1807	265
Xuzhou	506,592	7.66	92,456	13,173	4635	208
Jiang’an	344,169	3.43	63,373	13,824	2423	135
Changning	284,055	2.38	54,595	6119	2110	109
Gaoxian	308,643	3.94	78,136	6257	2707	115
Gongxian	293,352	2.74	67,509	15,813	2362	64
Junlian	269,765	2.59	53,746	7597	2354	197
Xingwen	303,859	3.27	78,313	13,234	2428	76
Pingshan	297,152	1.88	74,994	5508	2375	57
Regions	NM	NB	GS	GE	NE
Cuiping	133.54	42.62	4825	142,937	55.33
Nanxi	33.21	10.95	4139	66,592	18.37
Xuzhou	100.33	30.37	1277	147,533	54.88
Jiang’an	44.06	13.87	5547	84,383	23.18
Changning	37.43	12.57	1155	60,313	17.92
Gaoxian	41.75	13.42	284	69,558	20.43
Gongxian	37.59	11.79	1053	68,612	17.85
Junlian	34.87	11.13	2691	74,836	17.56
Xingwen	40.08	12.87	1571	89,086	20.74
Pingshan	27.53	9	2225	54,698	12.49

Abbreviation explanations: OF: Gross Output Value of Farming (10,000 RMB); CA: Cultivated Area (10,000 Hectare); GA: Government Expenditure for Agriculture, Forestry, and Water Conservancy (10,000 RMB); GT: Government Expenditure for Transport (10,000 RMB); HM: Highway Mileage (KM); NP: Number of Postal Offices (Unit); NM: Number of Mobile Telephone Subscribers (10,000); NB: Number of Internet Broadband Subscribers (10,000); GS: Government Expenditure for Science and Technology (10,000 RMB); GE: Government Expenditure for Education (10,000 RMB); NE: Number of Employed Persons (10,000).

in matrix form. Based on Table 2, the weights of each criterion can be calculated using the MEREC model. The performance scores, considering removal effects for each criterion, were calculated using Equations (1) to (3) and are presented in

Table 3. MEREC performance scores.

Region	${\mathit{S}}_{\mathit{i}}$	${\mathit{S}}_{\mathit{i}1}^{\prime }$	${\mathit{S}}_{\mathit{i}2}^{\prime }$	${\mathit{S}}_{\mathit{i}3}^{\prime }$	${\mathit{S}}_{\mathit{i}4}^{\prime }$	${\mathit{S}}_{\mathit{i}5}^{\prime }$
Cuiping	0.884856	0.867453	0.845187	0.863551	0.796907	0.867104
Nanxi	0.519655	0.492633	0.503688	0.472532	0.482405	0.519655
Xuzhou	0.745004	0.717431	0.682463	0.721314	0.706645	0.703501
Jiang’an	0.547182	0.534287	0.515044	0.538477	0.49757	0.531632
Changning	0.290939	0.287425	0.274782	0.289873	0.283764	0.280348
Gaoxian	0.312195	0.303197	0.261715	0.286985	0.303676	0.284937
Gongxian	0.349532	0.344145	0.325093	0.334811	0.279545	0.332216
Junlian	0.405354	0.405354	0.385744	0.405354	0.385672	0.389195
Xingwen	0.432215	0.425167	0.399008	0.409752	0.379105	0.414631
Pingshan	0.223994	0.216943	0.223994	0.199489	0.223994	0.203933
Regions	${\mathit{S}}_{\mathit{i}\mathbf{6}}^{\prime }$	${\mathit{S}}_{\mathit{i}\mathbf{7}}^{\prime }$	${\mathit{S}}_{\mathit{i}\mathbf{8}}^{\prime }$	${\mathit{S}}_{\mathit{i}\mathbf{9}}^{\prime }$	${\mathit{S}}_{\mathit{i}\mathbf{10}}^{\prime }$	${\mathit{S}}_{\mathit{i}\mathbf{11}}$
Cuiping	0.787889	0.823771	0.824729	0.772479	0.848145	0.827384
Nanxi	0.432918	0.509462	0.508995	0.36317	0.50896	0.498576
Xuzhou	0.687516	0.687575	0.691086	0.677925	0.701238	0.678989
Jiang’an	0.50077	0.522135	0.52417	0.377196	0.524114	0.514116
Changning	0.24588	0.269841	0.267972	0.190743	0.284275	0.2661
Gaoxian	0.264373	0.284099	0.285255	0.312195	0.296076	0.278909
Gongxian	0.34208	0.329367	0.332073	0.261805	0.334899	0.32638
Junlian	0.32721	0.390925	0.392395	0.258825	0.386171	0.384487
Xingwen	0.415094	0.409803	0.410884	0.32582	0.403011	0.401834
Pingshan	0.223994	0.223994	0.223994	0.061964	0.223994	0.223994

. After that, using Equations (4) and (5), the results of final criterion weights were recorded and are presented in

Table 4. Criterion weights of index system.

Indicator Abbreviations	Definition	Weights
OF	Gross Output Value of Farming (10,000 RMB)	0.031031
CA	Cultivated Area (10,000 Hectare)	0.078105
GA	Government Expenditure for Agriculture, Forestry, and Water Conservancy (10,000 RMB)	0.050118
GT	Government Expenditure for Transport (10,000 RMB)	0.098662
HM	Highway Mileage (KM)	0.048787
NP	Number of Postal Offices (Unit)	0.128278
NM	Number of Mobile Telephone Subscribers (10,000)	0.069011
NB	Number of Internet Broadband Subscribers (10,000)	0.066202
GS	Government Expenditure for Science and Technology (10,000 RMB)	0.29436
GE	Government Expenditure for Education (10,000 RMB)	0.053106
NE	Number of Employed Persons (10,000)	0.082339

.

After obtaining the weights for each criterion, the arithmetic and geometric mean values for each alternative were calculated.

Table 5. Heronian mean results of each region.

Region	Arithmetic Heronian	Geometric Heronian
Cuiping	0.921552715	0.896622918
Nanxi	0.52947379	0.46629304
Xuzhou	0.563951582	0.616631006
Jiang’an	0.613381222	0.522328533
Changning	0.269506943	0.331768843
Gaoxian	0.266657503	0.350323961
Gongxian	0.286874916	0.356716564
Junlian	0.388734203	0.385537252
Xingwen	0.340097992	0.390240915
Pingshan	0.312334313	0.312491341

shows the results. Based on the arithmetic and geometric Heronian mean values, the modified CoCoSo model was applied to calculate the final score to rank regional potentials for agricultural E-commerce. The results are listed in

Table 6. CoCoSo results for regional ranks.

Region	${\mathit{E}}_{\mathit{i}}$	${\mathit{P}}_{\mathit{i}}$	${\mathit{K}}_{\mathit{i}\mathit{a}}$	${\mathit{K}}_{\mathit{i}\mathit{b}}$	${\mathit{K}}_{\mathit{i}\mathit{c}}$	${\mathit{K}}_{\mathit{i}}$	Rank
Cuiping	0.921552715	0.896622918	0.199328152	6.32521429	1	3.588489731	1
Nanxi	0.52947379	0.46629304	0.10916677	3.477774025	0.54767362	1.970635949	4
Xuzhou	0.563951582	0.616631006	0.12942828	4.088165012	0.649322633	2.32269441	2
Jiang’an	0.613381222	0.522328533	0.124508833	3.971756255	0.62464249	2.249622466	3
Changning	0.269506943	0.331768843	0.065918379	2.072375487	0.330702808	1.17915272	10
Gaoxian	0.266657503	0.350323961	0.067640206	2.121067739	0.339340959	1.207827876	9
Gongxian	0.286874916	0.356716564	0.070557485	2.217342542	0.353976518	1.261793818	7
Junlian	0.388734203	0.385537252	0.084884043	2.691556755	0.425850749	1.527364681	5
Xingwen	0.340097992	0.390240915	0.08006768	2.52421683	0.401687765	1.434993689	6
Pingshan	0.312334313	0.312491341	0.068500172	2.171293926	0.343655279	1.232267005	8

.

3.2. Sensitivity Analysis

In our model, the $\lambda$ for CoCoSo calculation was 0.5 in the initial experiment. According to the previous literature [43], this is an important parameter determining the reliability of the results, so we conducted sensitivity analysis. We altered the value of $\lambda$ from 0.1 to 0.9 and examined the rankings of alternative potentials for developing agricultural E-commerce. Figure 3 indicates the alternative rankings according to the change in $\lambda$. It can be observed that our results remain very stable under changes in $\lambda$, indicating the robustness of our method.

4. Discussion

4.1. Comparison with Past Research

The above results reveal that different regions have different potentials for agricultural E-commerce development. Such a finding is consistent with the previous literature, as well as with agricultural development in China. In different areas, agricultural E-commerce can have different regional properties and thus development potentials [6]. Therefore, local governments need to implement diverse forms of policy support and financial stimulus [3].

Furthermore, our method indicates that some factors are more important as they have higher weights in decision-making. For example, local governments may consider dedicating efforts to enhancing Government Expenditure for Science and Technology in order to improve the effectiveness of agricultural E-commerce development. This is consistent with suggestions in previous studies that governmental promoting of scientific and technological innovations can contribute significantly to agricultural E-commerce development [44,45].

Finally, differing from the previous literature [5,6], our paper proposed a quantitative and objective evaluation. Our findings regarding the regional rankings of agricultural E-commerce development potentials confirm the capability of our system. Therefore, it can be stated that our method can be more informative and supportive regarding agricultural E-commerce development, as it can mitigate possible subjectivity and judgmental bias. In the following subsection, the insights from our results will be discussed.

4.2. Insights into Research Results

The above results indicate two main insights. On the one hand, the results reveal the rankings of regional potentials for agricultural E-commerce. On the other hand, the model also indicates the relative importance of their criteria, the implications of which we will now discuss.

First, it can be observed that the Cuiping region has the largest potential for developing agricultural E-commerce, followed by the Xuzhou and Jiang’an regions, while the Changning region has the smallest potential. Such results can support local governments in adopting better regional planning strategies. Specifically, in order to enhance the economic and societal development of Yibin city in terms of the agricultural E-commerce industry, the local government can prioritize the Cuiping, Xuzhou, and Jiang’an regions, rather than giving similar policy or financial support to all regions. To achieve this, the government can provide agricultural subsidies [46], farmer and E-commerce employee training [47], and infrastructure (e.g., E-commerce industrial zones) and transportation investments [48] in these three regions in order to maximize the effectiveness of agricultural E-commerce development. Additionally, governments can encourage people in the above three regions to work in agricultural E-commerce; creating more jobs in this sector would also support its development [47].

Moreover, our results can help local companies wishing to extend their business to agricultural E-commerce. For example, as the Cuiping, Xuzhou, and Jiang’an regions have the largest potentials, companies can build their product-processing factories, E-commerce centers, or live-streaming studios in these regions [49]. Considering the current popularity of live-streaming sales for agricultural products, companies can even train up farmers in these three regions to be social media influencers as they are familiar with local products. By doing so, E-commerce companies can benefit from the cost and quality advantages of regional agricultural products.

Besides the rankings of agricultural E-commerce potentials, the relative weights of the criteria can also provide useful information for local governments and companies. As generated by our model, the most important factors contributing to potential agricultural E-commerce development are GS (i.e., Government Expenditure for Science and Technology), followed by NP (i.e., the support Number of Postal Offices), and GT (i.e., Government Expenditure for Transport). Combined, these can local government policy making. Specifically, if local governments intend to make the agricultural E-commerce the region’s main industry, they may need to consider prioritizing public budgets for areas relevant to science and technology, especially those enhancing the efficiency of E-commerce, such as Internet-of-Things [50] and AI-enabled E-commerce technology (e.g., recommendation systems) [51]. Moreover, governments should provide budget support to increase post office numbers, especially in rural towns, in order to form a more connected logistics network for agricultural product delivery. Finally, governments can issue policies to subsidize aspects of transport such as drivers or logistics companies, as well as to upgrade transport infrastructures for effective regional agricultural E-commerce support.

5. Conclusions

E-commerce has become one of the most important approaches to selling products. Through E-commerce channels, transaction costs for agricultural products can be largely reduced. Economic growth in rural areas can be enhanced by the development of agricultural E-commerce, as well as by environmental sustainability. In this study, we proposed a novel framework integrating the MEREC, Heronian mean operator, and CoCoSo models to evaluate regional potentials for developing agricultural E-commerce. By applying our model to the case of Yibin city, we tested the model by evaluating the agricultural E-commerce potentials of the city’s ten regions. The results showed that the Cuiping, Xuzhou and Jiang’an regions rank as the top three, while the Changning region has the lowest potential.

We believe that our study makes a significant contribution to the existing literature and practices. In terms of academic contributions, this paper is the first to integrate the MEREC weighting method with Heronian–CoCoSo. Compared with the previous literature, by doing this, the criterion weighting can be derived objectively and the subjective bias kept at a minimum. Meanwhile, the interactions between criteria can be effectively captured using the Heronian mean, leading to a more appropriate comparison between alternatives. This makes the evaluation results more informative. Additionally, this paper developed an index system for evaluating regional agricultural E-commerce potentials, enriching the agricultural E-commerce literature and offering new ideas for the agricultural sector, developmental economics, and rural planning studies.

In addition, this study offers practical contributions. For companies, especially those aiming to enter the agricultural E-commerce business, our methods can provide an easily implementable analytical tool for making better decisions about location choices. Moreover, our results can better inform local governments in allocating their public investments and establishing appropriate policies for stimulating the local economy and enhancing social welfare.

However, we acknowledge that our study has limitations, thus paving the way for future research to refine our theory and model. In terms of research content, first, the index system developed in this paper only takes a specific city—and its unique features—into consideration, meaning that generalization is limited. In addition, as there are only ten regions in Yibin city; the model may not fully capture the economic and geographical diversity necessary for agricultural E-commerce development in different regions in China. Therefore, future research could extend our theoretic constructions, and may use our index system in other cities or areas, by adding more indicators to fit different regional features. Furthermore, the newly proposed model is only tested in the evaluation of regional potentials for agricultural E-commerce development, without verifying its capability in other industries. Future research can thus apply our model to other industrial contexts for evaluating various types of development potentials to test its effectiveness. Moreover, the index system developed in this paper mainly considers economic and technological perspectives (e.g., OF, HM, or GS), with relatively little attention paid to social or cultural factors. This is largely because of data availability issues. However, as social–cultural factors may also influence agricultural development, future studies could benefit from adding these into our index system, once relevant data are obtained, for a more comprehensive evaluation. Finally, although this study demonstrates the importance of the policy for developing agricultural E-commerce, it does not provide a detailed analysis for the effect of policy measures due to the MCDM purpose of this study. Therefore, future studies of the relevant fields can conduct in-depth analyses for the specific policy measures and their effects by using statistical models such as regression and structural equation modelling.

In terms of methodological limitations, the model only utilized statistics from a single year, without considering regional dynamics. This means that the current model could perhaps be expanded to reflect changes in regional agricultural E-commerce development potentials. Future research can extend the model to integrate data from multiple years, enabling a dynamic comparison of alternatives. In addition, although the current model fully considers and captures criteria interactions in the evaluation process, the mechanism behind their effects can be hard to deconstruct in detail due to the nonlinear nature of the evaluation. Therefore, in the future, researchers should conduct further explorations of the effects of criterion interactions on regional agricultural E-commerce development potentials in order to facilitate better interpretations. To achieve this, other methods like system dynamics or interpretive structural modeling can be applied. Finally, as this model is the first to integrate the MEREC, Heronian mean operator, and CoCoSo in agricultural E-commerce studies, its performance can be further examined by comparing with other methods in the future. For example, the MCDM with entropy method or Bonferroni mean operator can be comparatively analyzed against our model.