Frontiers in Business, Economics and Management ISSN: 2766-824X | Vol. 15, No. 2, 2024 350 Research on Supplier Selection of Fresh Food Supermarkets under New Retail Mode Yang Zhao Master of Engineering Management, Henan Polytechnic University, Jiao Zuo, China * Corresponding author: Zhao Yang (Email: zhaoyang0303a@163.com) Abstract: Under the new retail mode, the market competition in the fresh produce industry is fierce. Optimizing the procurement of fresh agricultural products and selecting high-quality suppliers are conducive to reducing costs and enhancing market competitiveness of fresh supermarkets under the new retail mode. This paper constructs a supplier selection index system for fresh produce supermarkets, uses hierarchical analysis and entropy weight method to determine the comprehensive weight of the indexes, and establishes the TOPSIS supplier selection model with gray correlation improvement. And the model is applied to A fresh food supermarket under the new retail mode, and the best supplier is selected for this supermarket from four strawberry suppliers. Keywords: Supplier Selection, Gray TOPSIS Method, Fresh Produce. 1. Introduction In recent years, the new retail industry is becoming more and more mature, and scholars in China have launched relevant research on the connotation, characteristics, motivation and development trend of new retail [1, 2]. Generally speaking, the research system of "new retail" has been initially formed, but not yet mature, and will be in the new normal of change for a long time. However, at present, many fresh food enterprises under the new retail model have not formed a complete evaluation system, and the quality of fresh food problems are endless, which restricts the development of the industry. The choice of suppliers determines the cost and quality of goods and the stability of fresh supermarket operations. Therefore, scientific and reasonable selection of suppliers for fresh supermarkets is of great significance to the development of fresh supermarkets under the new retail model. The main methods of supplier selection are hierarchical analysis method [3, 4], mathematical planning method [5, 6], gray correlation analysis method [7, 8], fuzzy comprehensive evaluation method [9, 10] and so on. In the supplier selection of fresh agricultural products, Zhu Xue Zhen uses hierarchical analysis to construct the supplier selection model of agricultural products, and comprehensively evaluates the strength of agricultural products suppliers [11], however, the hierarchical analysis method has more human factors and is more subjective when determining the weights. Thus, scholars began to use a combination of subjective and objective weights to determine the weights of indicators, and combined with the TOPSIS method to construct the supplier selection model of fresh produce. Song Bao'e comprehensively used the hierarchical analysis method and entropy value method to determine the index weights, and combined with the TOPSIS method to construct the fresh food supplier selection model for supermarkets to select long-term strategic partners [12]. Su Zong Rong constructed a fresh food supplier selection index system for Metro and combined the TOPSIS method to build a model to solve Metro's dilemma under new retail [13]. Feng Meng jie used the rooting theory to generalize an evaluation system containing 17 factors influencing the selection of fresh food e-commerce enterprises, and used the entropy weight-TOPSIS method to select the best suppliers for fresh food e-commerce platforms and realize the procurement of farmers by e-commerce platforms [14]. There are many indicators and complex data for fresh produce supplier selection, while TOPSIS method can normalize the data and eliminate the influence between multiple indicators. However, the single TOPSIS method focuses on selecting the program with the optimal level of synthesis when evaluating the target program, and it will ignore the relationship between the internal indicators and cannot reflect the level of each indicator dimension. Therefore, this study applies gray correlation to improve the TOPSIS method to select the best supplier of fresh food supermarket [15]. 2. Fresh Supermarket Supplier Selection Index System 2.1. Selection of Supplier Selection Indicators for Fresh Food Supermarkets It is known through literature research that the basic guidelines for supplier selection are quality, price, delivery capability, technical level, and service capability. Combined with the reality of the fresh food industry for targeted analysis, the quality indicators are refined into quality rating system, product quality pass rate, food traceability rate, freshness. Refine the price into, unit product price, distribution costs. Delivery capacity is refined into on-time delivery rate, continuous supply capacity, and average delivery period. For meat products, aquatic products, fruits and vegetables and other fresh commodities are easy to rot and deteriorate, not fresh, time-sensitive requirements are very strong characteristics. Fresh product suppliers have to ensure product freshness and reduce the loss rate, so the technical level is refined into three secondary indicators: cold chain transportation capacity, distribution cold storage facilities, and loss rate of transported products. In addition, in order to reduce procurement costs, the fresh food platform must increase the proportion of direct purchases, with high demand and the need for stable supply, so the volume discount rate 351 indicator is incorporated into the price, and the order fulfillment rate is incorporated into the delivery capacity. Finally, considering the characteristics of fresh food new retail and consumer demand response, the service ability is refined into two secondary indicators: customer satisfaction rate and ability to handle customer complaints. The supplier selection index system of fresh produce supermarket is constructed, as shown in Table 1: Table 1. Indicator system for supplier selection in fresh food supermarkets Level 1 indicators Secondary indicators Description of indicators Quality (A1 ) Quality level (A11 ) Scoring by experts based on the quality characteristics of the product, such as appearance, flavor, nutrition and other indicators, ten-point system Product quality pass rate (A12 ) Test the content of bacteria, viruses, pesticides, additives, etc. in commodities with reference to the Food Sanitation Law, the Agricultural Products Quality and Safety Law, the Pesticide Regulations, etc., and test the number of qualified food categories/total number of food categories Food traceability (A13 ) Number of traceable fresh produce types/total fresh produce types Freshness (A14 ) Supermarkets receive deliveries of freshness, scored by the receiving clerk based on experience, ten-point system Price (A2 ) Price per unit of product (A21 ) Quotation per unit of fresh produce Volume discount rate (A22 ) Discount rates available from suppliers for bulk orders Distribution costs (A23 ) Additional charge per unit of fresh produce when suppliers make deliveries Delivery capacity (A3 ) On-time delivery rate (A31 ) Examining the ability of each supplier to deliver products from a time perspective, number of orders arriving on time/total number of orders over time Continuity of supply (A32 ) Scoring by experts based on suppliers on a ten-point scale Average delivery time (A33 ) Time taken from acceptance of order to delivery by supplier Order fulfillment rate (A34 ) Ability to deliver according to actual demand, number of products actually supplied by the supplier/original order quantity Level of technology (A4 ) Cold chain transportation capacity (A41 ) Delivery speed of suppliers, cold chain transportation speed of fresh products to target supermarkets Distribution cold storage facilities (A42 ) Whether the cold storage area and other types of equipment are complete Wear and tear rate of transportation products (A43 ) Number of fresh produce spoiled and bumped during transportation/total number of fresh produce Service capacity (A5 ) Customer satisfaction rate (A51 ) Feedback on customer evaluations of fresh products after suppliers have supplied the products Ability to handle customer complaints (A52 ) Ability of fresh food suppliers to deal with problems arising after supply, number of after-sales satisfaction/number of customer complaint problems 2.2. Determination of Weights of Supplier Selection Indicators for Fresh Supermarkets In existing studies related to supplier selection indicators, the hierarchical analysis method and entropy weight method are often used to determine indicator weights. Scholars often use the combination method to determine the indicator weights, which can make up for the subjectivity of the hierarchical analysis method, but also avoid the entropy weight method to produce indicator weight values that are different from the actual, further improving the accuracy of the conclusions. 2.2.1. Entropy Weighting Method to Determine Objective Weights Entropy weight method is a mathematical method used to determine the degree of discrete of a certain indicator, can be based on the degree of difference between the indicators, the use of information entropy, calculated the weight of each indicator. There are m suppliers, n evaluation indicators, each supplier's evaluation indicator values form a matrix 𝑋 the values of the evaluation indicators of each supplier form a matrix. π‘₯𝑖𝑗 is the first 𝑖 of the supplier. 𝑗 The value of the indicator of the supplier is the first value of the supplier. (1) Standardization of raw data Data are standardized using the extreme value method, and since there are benefit-type indicators and cost-type indicators, evaluation indicators with different attributes need to be handled differently: Benefit-based indicators: 𝑦𝑖𝑗 = π‘₯π‘–π‘—βˆ’π‘šπ‘–π‘›{π‘₯𝑖𝑗} π‘šπ‘Žπ‘₯{π‘₯𝑖𝑗}βˆ’π‘šπ‘–π‘›{π‘₯𝑖𝑗} (𝑖 = 1,2, … , π‘š; 𝑗 = 1,2, … , 𝑛) (2-1) Cost-based indicators: 𝑦𝑖𝑗 = π‘šπ‘Žπ‘₯{π‘₯𝑖𝑗}βˆ’π‘₯𝑖𝑗 π‘šπ‘Žπ‘₯{π‘₯𝑖𝑗}βˆ’π‘šπ‘–π‘›{π‘₯𝑖𝑗} (𝑖 = 1,2, … , π‘š; 𝑗 = 1,2, … , 𝑛)(2-2) (2) Normalize the data: 𝑝𝑖𝑗 = 𝑦𝑖𝑗 βˆ‘ 𝑦𝑖𝑗 π‘š 𝑖=1 (2-3) (3) Calculate the information entropy: 𝐻𝑗 = βˆ’ 1 ln π‘š βˆ‘ 𝑝𝑖𝑗 𝑛 𝑖=1 ln 𝑝𝑖𝑗 (𝑗 = 1,2, … , 𝑛) (2-4) The coefficient of variationβ„Žπ‘— For: β„Žπ‘— = 1 βˆ’ 𝐻𝑗 (2-5) (4) Calculate entropy weights: πœ”π‘— = β„Žπ‘— βˆ‘ β„Žπ‘— 𝑛 𝑗=1 (2-6) 2.2.2. Hierarchical Analysis to Determine Subjective Weights Hierarchical analysis is a decision-making method that breaks down the factors that are always relevant to decision- making into levels such as objectives, guidelines, and programs, on the basis of which qualitative and quantitative 352 analyses are conducted. In this paper, we use 𝑒𝑗 denote the subjective weight of the jth evaluation index obtained by the hierarchical analysis method. Its calculation steps are as follows: (1) Establishment of a hierarchical model. Generally when the decision-making problem is transferred to a hierarchical structural model for analysis, the goals and considerations of the decision are divided into the goal level, the guideline level and the program level. (2) Construct judgment matrix. The factors at the same level are compared between the two, and the judgment matrix is established by using the nine-scaled method to assess the rank according to the degree of importance. (3) Calculate the relative weights. Calculate the relative weight of the factors on the level relative to the factors on the upper level, the commonly used methods are square root method, sum and product method and eigenvalue method, etc., and this paper adopts the sum and product method to calculate the weight. (4) Consistency test. Calculation of consistency indicators𝐢𝐼: 𝐢𝐼 = πœ†π‘šπ‘Žπ‘₯βˆ’π‘› π‘›βˆ’1 (2-7) πœ†π‘šπ‘Žπ‘₯ is the largest eigenvalue of the judgment matrix, calculate the test coefficient𝐢𝑅: 𝐢𝑅 = 𝐢𝐼 𝑅𝐼 (2-8) RI is a random consistency indicator, which can be obtained from the random consistency indicator table, when𝐢𝑅 ≀ 0.1 when, the judgment matrix is considered to have satisfactory consistency. 2.2.3. Geometric Averaging to Determine Composite Weights The weights obtained from entropy weighting and hierarchical analysis were combined using geometric averaging to obtain a composite weight of πœƒπ‘—: πœƒπ‘— = βˆšπ‘’π‘— Γ— πœ”π‘—(𝑗 = 1,2, … , 𝑛) (2-9) 3. Gray-based TOPSIS Modeling The TOPSIS method, also known as the Approximate Ideal Solution Ranking Method, is a multi-objective decision- making method. By calculating the relative closeness of alternative suppliers to the positive and negative ideal solutions, the suppliers are ranked and the best alternative is selected comprehensively. However, the TOPSIS method cannot reflect the level of each index dimension when evaluating suppliers, and it may appear that two suppliers have equal distance to the positive and negative ideal solutions. Gray correlation analysis is mainly used to rank the solutions through the shape or similarity with the ideal solution, to react to the trend of change within the sample and the difference between the sample and the target sample, and to emphasize the correlation between the internal elements of the project, and it is used to improve TOPSIS, which can reflect the status of the change of the indicators within the suppliers, and also the deviation between the suppliers and the target solution. The calculation steps are as follows: (1) Raw data standardization the raw data are first normalized and then standardized: 𝑍𝑖𝑗 = π‘₯𝑖𝑗 βˆšβˆ‘ π‘₯𝑖𝑗 2π‘š 𝑖=1 (3-1) (2) Construct the weighted standardized decision matrix using the indicator weights calculated in Equation 2-9πœƒπ‘— the weighted decision matrix is calculated by using the weights of the indicators calculated in Eq. 2-9. 𝐡𝑖𝑗 (b) Construct the weighted standardized decision matrix: 𝐡𝑖𝑗 = πœƒπ‘—π‘§π‘–π‘— (3-2) (3) Determine the positive and negative ideal solutions: 𝐡+ = (π‘šπ‘Žπ‘₯𝑖𝐡𝑖𝑗|𝑗 ∈ 𝐽1), (π‘šπ‘–π‘›π‘–π΅π‘–π‘—|𝑗 ∈ 𝐽2), |𝑖 = 1,2, … , π‘š (3-3) π΅βˆ’ = (π‘šπ‘–π‘›π‘–π΅π‘–π‘—|𝑗 ∈ 𝐽1), (π‘šπ‘Žπ‘₯𝑖𝐡𝑖𝑗|𝑗 ∈ 𝐽2), |𝑖 = 1,2, … , π‘š (3- 4) Among them: 𝐽1 and 𝐽2 are sets of benefit and cost indicators, respectively. (4) Calculate the distance from the supplier to the positive and negative ideal solutions: 𝑑𝑖 + = βˆšβˆ‘ (𝐡𝑖𝑗 βˆ’ 𝐡𝑗 +) 2𝑛 𝑗=1 (3-5) 𝑑𝑖 βˆ’ = βˆšβˆ‘ (𝐡𝑖𝑗 βˆ’ 𝐡𝑗 βˆ’)2𝑛𝑗=1 (3-6) (5) Determine the gray correlation coefficient: 𝑠𝑖𝑗 + = min 𝑖 min 𝑗 |𝐡𝑗 + βˆ’π΅π‘–π‘—|+𝜌 max 𝑖 max 𝑗 |𝐡𝑗 + βˆ’π΅π‘–π‘—| |𝐡𝑗 +βˆ’π΅π‘–π‘—|+𝜌 max 𝑖 max 𝑗 |𝐡𝑗 +βˆ’π΅π‘–π‘—| (3-7) 𝑠𝑖𝑗 βˆ’ = min 𝑖 min 𝑗 |𝐡𝑗 βˆ’βˆ’π΅π‘–π‘—|+𝜌 max 𝑖 max 𝑗 |𝐡𝑗 βˆ’βˆ’π΅π‘–π‘—| |𝐡𝑗 βˆ’βˆ’π΅π‘–π‘—|+𝜌 max 𝑖 max 𝑗 |𝐡𝑗 βˆ’βˆ’π΅π‘–π‘—| (3-8) (6) Determination of relevance: 𝑠𝑖 + = 1 𝑛 βˆ‘ 𝑠𝑖𝑗 +𝑛 𝑗=1 (3-9) 𝑠𝑖 βˆ’ = 1 𝑛 βˆ‘ 𝑠𝑖𝑗 βˆ’π‘› 𝑗=1 (3-10) (7) Normalization of Euclidean distance and grey correlation: 𝑀 = 𝑀𝑖 π‘šπ‘Žπ‘₯𝑀𝑖 ⁄ (3-11) Where 𝑀𝑖, which can be expressed as𝑑𝑖 +, and𝑑𝑖 βˆ’, and𝑠𝑖 +, and𝑠𝑖 βˆ’. (8) Combining normalized Euclidean distance and gray correlation: 𝑓𝑖 + = πœ†1𝑑𝑖 βˆ’ + πœ†2𝑠𝑖 + (3-12) 𝑓𝑖 βˆ’ = πœ†2𝑑𝑖 + + πœ†2𝑠𝑖 βˆ’ (3-13) Among them:πœ†1 andπœ†2 denote the degree of preference of the decision maker for location and shape, and satisfyπœ†1 + πœ†2 = 1. (9) Determine the relative closeness of the gray correlation for each alternative supplier𝐷𝑖 : 𝐷𝑖 = 𝑓𝑖 + 𝑓𝑖 βˆ’+𝑓𝑖 + (3-14) 𝐷𝑖 The larger the value, the closer the candidate program is to the ideal value. 353 4. Example Analysis 4.1. Data Sources Supermarket A, in order to promote the development of fresh food operation under the new retail model and improve its market competitiveness, needs to choose the strongest one among the four strawberry fresh produce suppliers S1, S2, S3, S4 as a strategic partner. According to the information collection of the four suppliers, combined with the scoring of qualitative indicators by experts, the indicator data of the alternative suppliers are collated, see Table 2: Table 2. Alternative supplier raw data statistics norm S1 S2 S3 S4 A11 10 8 8 10 A12 97.90% 98.21% 98.54% 98.9% A13 100% 100% 100% 98% A14 8.89 8.33 9.22 9.42 A21 20 19 18 21 A22 1.32 1.29 1.26 1.34 A23 0.25 0 0 0.2 A31 98% 96.7% 99.8% 96.4% A32 8.42 7.89 9.41 7.42 A33 0.5 1 1 1.5 A34 70% 75% 79% 68% A41 9.21 9.45 9.33 9.44 A42 8.2 8.97 8.34 8.42 A43 20% 25% 22% 23% A51 98% 99% 97% 99% A52 96% 97.8% 98% 99.2% 4.2. Calculation of indicator weights Calculation of objective weights of indicators Normalization matrix from Eqs. 2-1 and 2-2 π‘Œ = [ 1.0000 0.0000 0.0000 1.0000 0.0000 0.3100 0.6400 1.0000 1.0000 1.0000 1.0000 0.0000 0.5138 0.0000 0.8165 1.0000 0.3333 0.6667 1.0000 0.0000 0.7500 0.3750 0.0000 1.0000 0.0000 1.0000 1.0000 0.2000 0.4706 0.0882 1.0000 0.0000 0.5025 0.2362 1.0000 0.0000 1.0000 0.5000 0.5000 0.0000 0.1818 0.6364 1.0000 0.0000 0.0000 1.0000 0.5000 0.9583 0.0000 1.0000 0.1818 0.2857 1.0000 0.0000 0.6000 0.4000 0.5000 1.0000 0.0000 1.0000 0.0000 0.5625 0.6250 1.0000] The entropy weight of each indicator is calculated from Equations 2-4 to 2-6 as: πœ”1 = (0.105, 0.059, 0.044, 0.049) πœ”2 = (0.057, 0.054, 0.069) πœ”3 = (0.088, 0.067, 0.053, 0.070) πœ”4 = (0.050, 0.083, 0.054) πœ”5 = (0.050, 0.049) Calculation of subjective weights for indicators the judgment matrix for the first-level indicators is shown in Table 3: Table 3. Judgement matrix for Tier 1 indicators Level 1 indicators A1 A2 A3 A4 A5 A1 1 3 2 3 5 A2 1/3 1 1/3 1/3 4 A3 1/2 3 1 1/2 3 A4 1/3 3 2 1 2 A5 1/5 1/4 1/3 1/2 1 The weights of the level 1 indicators were calculated by the sum and product method: 𝑒0 = (0.393,0.125,0.195,0.221,0.066) πœ†π‘šπ‘Žπ‘₯ = 1 5 Γ— ( 2.151 0.392 + 0.650 0.125 + 1.070 0.195 + 1.240 0.221 + 0.350 0.066 ) = 5.4 Final consistency check. 𝐢𝐼 = πœ†π‘šπ‘Žπ‘₯βˆ’π‘› π‘›βˆ’1 = 5.4βˆ’5 5βˆ’1 = 0.1 𝐢𝑅 = 𝐢.𝐼. 𝑅.𝐼. = 0.1 1.12 = 0.089 < 0.10 Inspection passed. According to the above steps, the judgment matrix and solution of the secondary indicators are shown in Table 4 to Table 8: Table 4. Judgment matrix for secondary indicator A1 indicators A11 A12 A13 A14 consistency test A11 1 2 1 3/2 πœ†π‘šπ‘Žπ‘₯ =4.09 𝐢𝐼 = 0.03 𝐢𝑅 = 0.03 < 0.10 A12 1/2 1 1/2 3/4 A13 3/4 2 1 2/3 A14 2/3 4/3 2/3 1 Table 5. Judgment matrix for secondary indicator A2 indicators A21 A22 A23 consistency test A21 1 5/6 5/8 πœ†π‘šπ‘Žπ‘₯ = 3.08 𝐢𝐼 = 0.04 𝐢𝑅 = 0.06 < 0.10 A22 6/5 1 3/4 A23 8/5 5/3 1 Table 6. Judgment matrix for secondary indicator A3 indicators A31 A32 A33 A34 consistency test A31 1 3/2 3/2 6/5 πœ†π‘šπ‘Žπ‘₯ =4.03 𝐢𝐼 = 0.01 𝐢𝑅 = 0.01 < 0.10 A32 2/3 1 1 4/5 A33 2/3 6/7 1 4/5 A34 7/8 5/4 5/4 1 Table 7. Judgment matrix for secondary indicator A4 Secondary indicators A41 A42 A43 consistency test A41 1 9/10 9/10 πœ†π‘šπ‘Žπ‘₯ = 3.04 𝐢𝐼 = 0.02 𝐢𝑅 = 0.03 < 0.10 A42 10/9 1 10/9 A43 9/8 1 1 Table 8. Judgment matrix for secondary indicator A5 Secondary indicators A51 A52 consistency test A51 1 7/5 πœ†π‘šπ‘Žπ‘₯ =2 𝐢𝑅 = 0 < 0.10 A52 5/7 1 The calculation gives the weights of the secondary 354 indicators: 𝑒1 = (0.328,0.199,0.254,0.219) 𝑒2 = (0.255,0.306,0.439) 𝑒3 = (0.317,0.211,0.204,0.268) 𝑒4 = (0.306,0.352,0.342) 𝑒5 = (0.583,0.417) The combined weights are derived from Equation 2-9, see Table 9: Table 9. Combined weights of indicators norm Subjective weights𝑒𝑗 objective weightingπœ”π‘— 𝑒𝑗 Γ— πœ”π‘— Combined weightsπœƒπ‘— A11 0.328 0.105 0.186 0.085 A12 0.199 0.059 0.108 0.049 A13 0.254 0.044 0.105 0.048 A14 0.219 0.049 0.104 0.047 A21 0.255 0.057 0.121 0.055 A22 0.306 0.054 0.129 0.059 A23 0.439 0.069 0.174 0.079 A31 0.317 0.088 0.167 0.076 A32 0.211 0.067 0.119 0.054 A33 0.204 0.053 0.104 0.047 A34 0.268 0.070 0.137 0.062 A41 0.306 0.050 0.124 0.056 A42 0.352 0.083 0.171 0.078 A43 0.342 0.054 0.136 0.062 A51 0.583 0.050 0.171 0.078 A52 0.417 0.049 0.143 0.065 4.3. Gray TOPSIS-based Model Solving According to Equations 3-1 and 3-2, each indicator is weighted after standardization, and then according to Equations 3-3 and 3-4, the positive ideal solution is calculated 𝐡+ and negative ideal solution π΅βˆ’ which are shown in Table 10: Table 10. Weighted normalized matrices and positive and negative ideal solutions norm S1 S2 S3 S4 𝐡+ π΅βˆ’ A11 0.0469 0.0375 0.0375 0.0469 0.0469 0.0375 A12 0.0244 0.0245 0.0245 0.0246 0.0246 0.0244 A13 0.0241 0.0241 0.0241 0.0236 0.0241 0.0236 A14 0.0233 0.0218 0.0241 0.0247 0.0247 0.0218 A21 0.0147 0.0294 0.0441 0.0000 0.0441 0.0000 A22 0.0299 0.0292 0.0285 0.0303 0.0303 0.0285 A23 0.0000 0.0553 0.0553 0.0111 0.0553 0.0000 A31 0.0381 0.0376 0.0388 0.0375 0.0388 0.0375 A32 0.0273 0.0256 0.0305 0.0241 0.0305 0.0241 A33 0.0384 0.0192 0.0192 0.0000 0.0384 0.0000 A34 0.0297 0.0318 0.0335 0.0288 0.0335 0.0288 A41 0.0276 0.0283 0.0279 0.0282 0.0283 0.0276 A42 0.0377 0.0412 0.0383 0.0387 0.0412 0.0377 A43 0.0503 0.0000 0.0302 0.0201 0.0503 0.0000 A51 0.0389 0.0393 0.0385 0.0393 0.0393 0.0385 A52 0.0319 0.0325 0.0326 0.0330 0.0226 0.0319 Calculated from equations 3-5 and 3-6: 𝑑𝑖 + = (0.0630, 0.0569, 0.0296, 0.0797) 𝑑𝑖 βˆ’ = (0.0657, 0.0657, 0.0797, 0.0251) Calculated according to equations 3-7 and 3-8: 𝑠𝑖𝑗 + = [ 1.0000 0.3333 0.3333 1.0000 0.3333 0.4202 0.5814 1.0000 1.0000 1.0000 1.0000 0.3333 0.5070 0.3333 0.7315 1.0000 0.4286 0.6000 1.0000 0.3333 0.6667 0.4444 0.3333 1.0000 0.3333 1.0000 1.0000 0.3846 0.4857 0.3524 1.0000 0.3333 0.5013 0.3956 1.0000 0.3333 1.0000 0.5000 0.5000 0.3333 0.3793 0.5789 1.0000 0.3333 0.3333 1.0000 0.5000 0.9231 0.3333 1.0000 0.3793 0.4118 1.0000 0.3333 0.5556 0.4545 0.5000 1.0000 0.3333 1.0000 0.3333 0.5333 0.5714 1.0000] 𝑆𝑖𝑗 βˆ’ = [ 0.3333 1.0000 1.0000 0.3333 1.0000 0.6173 0.4386 0.3333 0.3333 0.3333 0.3333 1.0000 0.4932 1.0000 0.3798 0.3333 0.6000 0.4286 0.3333 1.0000 0.4000 0.5714 1.0000 0.3333 1.0000 0.3333 0.3333 0.7143 0.5152 0.8500 0.3333 1.0000 0.4987 0.6792 0.3333 1.0000 0.3333 0.5000 0.5000 1.0000 0.7333 0.4400 0.3333 1.0000 1.0000 0.3333 0.5000 0.3429 1.0000 0.3333 0.7333 0.6364 0.3333 1.0000 0.4545 0.5556 0.5000 0.3333 1.0000 0.3333 1.0000 0.4706 0.4444 0.3333] Calculated from equations 3-9 and 3-10: 𝑠𝑖 + = (0.5709, 0.6142, 0.6762, 0.6359) 𝑠𝑖 βˆ’ = (0.2445, 0.5765, 0.5282, 0.6406) When takingπœ†1 = πœ†2 When it is taken, it can be obtained according to Eqs. 3-11 to 3-13: 𝑓𝑖 + = (0.8345, 0.8663, 1.0000, 0.6276) 𝑓𝑖 βˆ’ = (0.5857, 0.8069, 0.5976, 1.0000) The relative closeness of each alternative supplier is calculated according to Equation 3-14: 𝐷𝑖 = (0.5876, 0.5177, 0.6259, 0.3856) From the results, it can be seen that the relative closeness of gray correlation for the four alternative suppliers results in: 𝐷3 > 𝐷1 > 𝐷2 > 𝐷4, therefore supplier S3 is the best choice. 5. Conclusion Under the new retail model, selecting high-quality suppliers and cooperating with them in depth is conducive to fresh produce supermarkets to solve various problems of retail terminals, such as product quality, food safety, and customer satisfaction and so on, and to improve their own competitiveness. Through literature research, expert interviews and other methods, we constructed a supplier selection index system for fresh supermarkets consisting of 5 primary indicators and 16 secondary indicators. The entropy 355 weight method and hierarchical analysis method are used to determine the comprehensive weights of the indicators to overcome the defects of the single weight solving method. The TOPSIS model improved by gray correlation analysis is constructed, which considers the correlation between indicators and makes the results more scientific and reasonable. The example application shows that the method is easy to operate, the selection results are intuitive, and it provides a reference for the selection of fresh supermarket suppliers. References [1] Du Ruiyun, Jiang Kan. New retail:connotation,development motives and key issues [J]. Price theory and practice, 2017, (2):139-141. [2] Wang Kun, XIANG Feng. Theoretical structure and research paradigm of "new retail" [J]. China Circulation Economy, 2018, (1):3-11. [3] Thomas L. Saaty. the analytic hierarchy process :planning, priority setting, resource alloction [M]. 1980. [4] Rajnish Kumar. Supplier Selection of an Indian Heavy Locomotive Manufacturer:An Integrated Approach using Taguchi Loss Function,TOPSIS,and AHP [J]. IIMB Management Review, 2018, 31(1):78-90. [5] Gaballa A. Minimum Cost Allocation of Tenders [J]. Journal of the Operational Research Society, 1974, (25): 389-398. [6] Ghodsypour S H, O'Brien C. The total cost of logistics in supplier selection, under conditions of multiple sourcing, multiple criteria and capacity constraint [J]. International Journal of Production Economics, 2001, 73(1):15-27. [7] Rajesh R, Ravi V . Supplier selection in resilient supply chains: a grey relational analysis approach [J]. Journal of Cleaner Production, 2015, 86(13): 343. [8] Ying Zhang. Dang Yaoguo. Supplier evaluation method based on improved gray correlation analysis [J]. Statistics and Decision Making, 2007(21):50-52. [9] Zhang Xianfeng, Zheng Jianguo. Research on information system supplier evaluation based on fuzzy hierarchy analysis and fuzzy evaluation [J]. China Management Informatization, 2009, 12(02):86-89. [10] Cui T. Research on Green Supplier Evaluation Method for Manufacturing Enterprises Based on Hierarchical Analysis Method and Fuzzy Comprehensive Evaluation [J]. China Business Trade, 2009(05):149-150. [11] Zhu Xuezhen. Supplier selection in green agricultural supply chain [J]. Agricultural Economic Issues, 2007, (04):67-70. [12] Song Bao'e, Zhu Wen'in et al. Research on supermarket fresh food supplier selection model based on TOPSIS method [J]. Food and Machinery, 2013, 29(4):223-232. [13] Su Zongrong. Research on optimization of Metro's fresh product procurement management under new retail [C]. Beijing, China University of Mining and Technology, 2019:41- 51. [14] Feng Mengge. Research on the evaluation problem of the selection of agricultural products supplier in fresh e-commerce [C]. Zhengzhou, Henan University of Finance and Law, 2021:48-54. [15] Chengli, Zhao, Yanheng, et al. Evaluation for Survivable Networked System Based on Grey Correlation and Improved TOPSIS [J]. Journal of Networks, 2011, 6(10):1514-1520.