Supplier and forwarder selection method based on genetic algorithm single-objective programming model

Abstract

The present invention discloses a method for selecting suppliers and forwarders based on a single-objective programming model of a genetic algorithm, and the steps are as follows: S1: receiving the historical transaction data of each supplier for preprocessing, and scoring the preprocessed data to select the top L suppliers; S2: constructing a single-objective programming model, taking the top L suppliers as variables of the single-objective programming model, and then using a genetic algorithm to obtain the optimal solution; S3: obtaining the production volume of the enterprise in each of the next W weeks and the supply volume of the participating suppliers in each week according to the optimal solution, thereby determining the optimal supplier selection scheme. The present invention adopts a single-objective programming model based on a genetic algorithm, takes the objective function as search information, quickly searches the solution space, and avoids falling into the descent trap of the local optimal solution; at the same time, the characteristics of the genetic algorithm such as variation and selection also well simulate various uncertainties between enterprises and suppliers.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to a method for selecting suppliers and forwarders based on a genetic algorithm single-objective programming model. Background Art

[0002] In the new economic environment, the relationship between the core construction enterprises and the project participants such as suppliers and transporters is to fully understand the mutual interests, determine the common project goals, establish an organizational structure with different working groups as units, form a strategic partnership with complementary resources and advantages, shared interests and shared risks, break the traditional "vertical integration" thinking mode, and establish a "horizontal integration" strategic alliance in the supply chain. Establishing a good partnership can not only improve the comprehensive competitiveness of the enterprise, but also make the supply chain coordinated and effectively operate, improve operational efficiency, so that all enterprises in the supply chain can benefit and promote their common development. Therefore, as one of the basic activities that constitute the supply chain, the appropriateness of the choice of partners greatly affects the vital interests of every member in the chain, which is mainly determined by the accuracy of the evaluation of candidate suppliers and transporters.

[0003] For construction companies, the consumption of materials in construction projects often accounts for more than 50% of the project investment. For material suppliers, they are always unable to complete the company's order quantity on schedule, which makes the company always face the embarrassing situation of imbalance between supply and demand. It is not uncommon to cause serious problems such as project delays and poor quality, and the profit margins in the fiercely competitive construction market are getting smaller and smaller. For material transporters, there are always material loss problems in the actual transportation process due to various factors, and the transportation loss rates of different material transporters are also different. Therefore, how to reasonably select future suppliers and transporters based on the supply volume of past suppliers and the loss rate of transporters is an urgent problem for construction companies to solve.

[0004] The existing technology research adjusts the fitness function of the genetic algorithm by adding constraints, and uses the improved genetic algorithm to select the optimal supplier combination. First, the weight of the evaluation index of the supplier is determined by using AHP, and the score of each supplier under each index is obtained, and then the optimal supplier combination is selected by using the improved genetic algorithm. However, this method still has the following shortcomings:

[0005] 1) When considering the important indicators of suppliers, the supplier's production volume is not taken into account. However, for enterprises, the supplier's production volume is the top priority. No matter how good a supplier's other conditions are, it is futile if it cannot meet the production needs of the enterprise;

[0006] 2) When selecting suppliers, the total number of suppliers is too small, which is not practical for large enterprises, especially construction enterprises, and there is no idea of ​​pre-processing suppliers. Because when the number of suppliers is very large, directly using genetic algorithms to solve the optimal combination is often inefficient and cannot achieve the optimal result;

[0007] 3) At the same time, the existing method of selecting suppliers has large amount of calculation, complex calculation and low processing efficiency. Summary of the invention

[0008] In order to solve the above problems of the shortcomings and defects in the prior art, the present invention provides a method for selecting suppliers and forwarders based on a genetic algorithm single-objective programming model.

[0009] In order to achieve the above-mentioned purpose of the present invention, the technical scheme adopted is as follows:

[0010] A method for selecting suppliers and forwarders based on a genetic algorithm single-objective programming model, wherein the method steps are as follows:

[0011] S1: Receive the historical transaction data of each supplier for preprocessing, score the preprocessed data, and select the top L suppliers;

[0012] S2: Construct a single-objective programming model and use the top L suppliers as variables of the single-objective programming model.

[0013] Then use genetic algorithm to find the optimal solution;

[0014] S3: Based on the optimal solution, the production volume of the enterprise in each week in the next W weeks and the supply volume of the participating suppliers in each week are obtained to determine the optimal supplier selection plan.

[0015] Preferably, the pretreatment is as follows:

[0016] S101: Analyze various indicators based on historical transaction data and quantify each indicator;

[0017] S102: using the entropy weight method to assign weights to each indicator and determine the weights;

[0018] S103: Use the TOPSIS evaluation model to score each supplier and select the top L suppliers;

[0019] The indicators include supply completion coefficient, supply stability coefficient, supply material grade, and production capacity; the indicators are quantified by using one of the following methods: average value, weighted sum and proportion, and AHP analysis.

[0020] Furthermore, the entropy weight method comprises the following steps:

[0021] D1: Standardize the data of each indicator;

[0022] D2: Obtain the entropy value and information entropy redundancy of each indicator according to the proportion;

[0023] D3: Obtain the weights corresponding to each indicator based on the information entropy redundancy obtained in step D2.

[0024] Furthermore, the TOPSIS evaluation model comprises the following steps:

[0025] F1: normalize the indicator data and standardize the matrix;

[0026] F2: Calculate the distance between the maximum and minimum values ​​based on the weights obtained by the entropy weight method;

[0027] F3: Normalize the distance value obtained to get the total score.

[0028] Furthermore, after obtaining the score coefficients of all suppliers, the first L suppliers are selected by sorting them from large to small, or L suppliers whose score coefficients are greater than a preset score threshold are selected.

[0029] Furthermore, the single-objective programming model is constructed as follows: first determine the constraints, then use the minimum production cost as the objective function to construct the single-objective programming model, and then use the genetic algorithm to find the optimal solution of the objective function.

[0030] Furthermore, according to the different materials supplied by the selected n suppliers, the objective function Z(x) of the minimum production cost is defined as follows:

[0031]

[0032] Constraints:

[0033]

[0034] In the formula, m represents the number of suppliers of material A contained in n, p represents the number of suppliers of material B contained in n, and q represents the number of suppliers of material C contained in n; (op) imax represents the maximum production volume corresponding to the ith merchant; x i represents the supply quantity of the i-th supplier in category A, x j represents the supply quantity of the jth supplier in category B, x k represents the supply quantity of the kth supplier among the C-type suppliers, y, a represents the price required for each cubic meter of Class A materials consumed, b represents the price required for each cubic meter of Class B materials consumed, and c represents the price required for each cubic meter of Class C materials consumed.

[0035] Furthermore, a genetic algorithm is used to find the optimal solution of the objective function, as follows: For solving high-dimensional single-objective programming problems, a genetic algorithm is used to directly use the objective function value as search information and continuously iterate to find the optimal individual.

[0036] Furthermore, after step S3, step S4 is also included, which is as follows: based on the loss rate of each of the existing N forwarders in each week in the past Y years, a time series analysis method is used to predict the loss rate of these N forwarders in each of the next W weeks; then the enterprise terminal selects Q forwarders with the smallest loss rate according to the size of the loss rate predicted in the future, so as to determine the best forwarding plan for each of the next W weeks.

[0037] Furthermore, before using the time series analysis method to predict the loss rate of these N forwarders in each of the next W weeks, the linear trend method of adjacent points is used to fill in the missing values, and then the time series analysis is used to calculate the loss rate of these N forwarders in each of the next W weeks.

[0038] The beneficial effects of the present invention are as follows:

[0039] 1. The present invention adopts the TOPSIS evaluation model based on the entropy weight method, normalizes the original data, makes full use of the historical transaction data of suppliers, comprehensively and objectively reflects the gap between suppliers, and selects the best supplier according to the total score.

[0040] 2. The present invention adopts a single-objective programming model based on a genetic algorithm, takes the objective function as search information, and quickly searches the solution space to avoid falling into the trap of local optimal solutions; at the same time, the variation, selection and other characteristics of the genetic algorithm also well simulate the various uncertainties between enterprises and suppliers.

[0041] 3. The present invention uses time series analysis to predict the future loss rate of the forwarder based on the past loss rate change trend, which is in line with the law of the development of the forwarder's forwarding capacity; it pays attention to the seasonal and cyclical changes in the loss rate of the forwarder; and it recognizes that random variables may affect the final prediction results.

[0042] 4. The present invention constructs a single-objective programming model, takes the top L suppliers as variables of the single-objective programming model, and then uses a genetic algorithm to obtain the optimal solution, which greatly reduces the calculation volume, reduces the complexity of the algorithm, and improves computer processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the supplier and forwarder selection method described in Example 1.

[0044] Figure 2It is a flow chart of the supplier and forwarder selection method described in Example 3. DETAILED DESCRIPTION

[0045] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0046] Example 1

[0047] like Figure 1 As shown, a method for selecting suppliers and forwarders based on a genetic algorithm single-objective programming model, wherein the method steps are as follows:

[0048] The steps of the method are as follows:

[0049] S1: Receive the historical transaction data of each supplier for preprocessing, score the preprocessed data, and select the top L suppliers;

[0050] In a specific embodiment, the pre-processing is as follows:

[0051] S101: Analyze various indicators based on historical transaction data and quantify each indicator;

[0052] S102: using the entropy weight method to assign weights to each indicator and determine the weights;

[0053] S103: Use the TOPSIS evaluation model to score each supplier and select the top L suppliers;

[0054] The indicators include supply completion coefficient, supply stability coefficient, supply material grade, and production capacity; the indicators are quantified by using one of the following methods: average value, weighted sum ratio, and AHP analysis. The historical transaction data described in this embodiment includes historical transaction orders and corresponding completion statistics.

[0055] In a specific embodiment, in order to enable the TOPSIS scoring model to be effectively weighed by various indicators, the "entropy weight method" is first used to assign weights to various indicators. Further, the entropy weight method includes the following steps:

[0056] D1: Standardize the data of each indicator;

[0057] D2: Obtain the entropy value and information entropy redundancy of each indicator according to the proportion;

[0058] D3: Obtain the weights corresponding to each indicator based on the information entropy redundancy obtained in step D2.

[0059] Furthermore, the TOPSIS evaluation model comprises the following steps:

[0060] F1: normalize the indicator data and standardize the matrix;

[0061] F2: Calculate the distance between the maximum and minimum values ​​based on the weights obtained by the entropy weight method;

[0062] F3: Normalize the distance value obtained to get the total score.

[0063] In a specific embodiment, after obtaining the score coefficients of all suppliers, the top L suppliers are selected from hundreds, thousands, or even tens of thousands of suppliers through sorting from large to small, or the L suppliers whose score coefficients are greater than a preset score threshold are selected. The extraction and selection of suppliers can simplify the variables required by the mathematical model when selecting suppliers and forwarders in the subsequent process, improve the speed of the solution algorithm, and also meet the actual needs of enterprises.

[0064] S2: Use the enterprise terminal to build a single-objective programming model, use the top L suppliers as variables of the single-objective programming model, and then use the genetic algorithm to find the optimal solution;

[0065] In a specific embodiment, the single-objective programming model is constructed as follows: first determine the constraint conditions, then use the minimum production cost as the objective function, thereby constructing the single-objective programming model, and then use the genetic algorithm to find the optimal solution of the objective function. The optimal solution of the objective function is found by using the genetic algorithm as follows: For solving the high-dimensional single-objective programming problem, the genetic algorithm is used to directly use the objective function value as the search information, and iterates to find the optimal individual.

[0066] S3: The enterprise terminal obtains the production volume of the enterprise in each week of the next W weeks and the supply volume of the participating suppliers in each week based on the optimal solution, thereby determining the optimal supplier selection plan.

[0067] In a specific embodiment, according to the different materials supplied by the selected n suppliers, the objective function Z(x) of the minimum production cost is defined as follows:

[0068]

[0069] Constraints:

[0070]

[0071] In the formula, m represents the number of suppliers of material A contained in n, p represents the number of suppliers of material B contained in n, and q represents the number of suppliers of material C contained in n; (op) imax represents the maximum production volume corresponding to the ith merchant; x i represents the supply quantity of the i-th supplier in category A, x j represents the supply quantity of the jth supplier in category B, xk represents the supply quantity of the kth supplier among the C-type suppliers, y, a represents the price required for each cubic meter of Class A materials consumed, b represents the price required for each cubic meter of Class B materials consumed, and c represents the price required for each cubic meter of Class C materials consumed.

[0072] Example 2

[0073] Based on the supplier and forwarder selection method based on the genetic algorithm single-objective programming model described in Example 1, this embodiment provides a specific embodiment as follows:

[0074] When companies choose suppliers, they should first determine their own demand. At the same time, the supply of each supplier varies, and companies almost always choose from multiple suppliers. Taking into account multiple factors, companies will first select the minimum number of suppliers that meet their needs from merchants with high scoring coefficients through indicators such as average supply.

[0075] The reselected suppliers at least meet the needs of the enterprise, but how to choose among these suppliers to meet the optimization task that meets the needs of the enterprise based on the different types of materials and the upper limit of the quantity provided by each merchant is a difficult point. The most classic optimization task is to select the material quantity selection plan of each supplier with the lowest cost. The specific model establishment and solution of the optimization task are as follows:

[0076] Assume that the production capacity of an enterprise in a certain period of time is y, and the required materials are material A, material B, and material C. Each cubic meter of y consumes d cubic meters of material A, e cubic meters of material B, or f cubic meters of material C. The corresponding unit prices of the three materials are a, b, and c.

[0077] (1) Establishment of constraint conditions

[0078] The material quantity allocation problem is affected and constrained by various factors. According to the cost optimization problem, this embodiment summarizes the complex influencing factors as follows:

[0079] ① Supplier maximum production capacity constraint, that is, the material quantity allocated to each supplier must not exceed its corresponding maximum production capacity.

[0080] ② Supplier quantity constraint, that is, the total number of material suppliers shall not exceed the number of selected suppliers.

[0081] ③ Production capacity constraints, that is, the corresponding supply quantity of each assigned supplier must meet the company's production capacity needs.

[0082] (2) Objective function establishment

[0083] According to the different materials supplied by the n suppliers selected in the above stage, the objective function Z(x) with the lowest cost can be defined as follows:

[0084]

[0085] Constraints:

[0086]

[0087] In the formula, m represents the number of suppliers of material A contained in n, p represents the number of suppliers of material B contained in n, and q represents the number of suppliers of material C contained in n; (op) imax represents the maximum production volume corresponding to the ith merchant; x i represents the supply quantity of the i-th supplier in category A, x j represents the supply quantity of the jth supplier in category B, x k represents the supply quantity of the kth supplier among the C-type suppliers, y, a represents the price required for each cubic meter of Class A materials consumed, b represents the price required for each cubic meter of Class B materials consumed, and c represents the price required for each cubic meter of Class C materials consumed.

[0088] (3) Genetic algorithm solution

[0089] For solving high-dimensional single-objective programming problems, genetic algorithms can directly use the objective function value as search information and continuously iterate to find the optimal individual.

[0090] ①Individual coding

[0091] Each individual has n characteristics, and each characteristic corresponds to the amount of selected materials of each supplier. where x n Indicates the value corresponding to the nth feature of an individual in the i-th generation. Initialize the number of populations and the number of iterations at the same time. The larger the values ​​of the two, the richer the population, and the more phenotypical the excellent individuals produced, but the algorithm speed will also slow down. You can consider it comprehensively according to the actual situation.

[0092] ②Fitness function design

[0093] The setting of the fitness function will affect the entire solution. The larger the fitness value, the better the quality of the solution. This embodiment directly uses the constrained objective function as the fitness function.

[0094] ③Select

[0095] The selection operation allows individuals to undergo a process of survival of the fittest. The genetic algorithm described in this embodiment uses a sorting selection method instead of a general roulette selection method. After the individuals in the population are sorted by fitness, the best individual is N, and the probability of being selected is p max , the worst individual number is 1, and the probability of selection is p min , the probability that the i-th individual is selected is pxi for:

[0096]

[0097] ④ Cross

[0098] After the selection, there is a historically optimal individual P i gen =(p1,p2,p3,…p n ), each individual after selection Perform discrete crossover with it to generate the next generation of individuals The discrete crossover formula is:

[0099]

[0100] In the formula, rand(0,1) is a random number that satisfies uniform distribution, and p is the crossover probability. Then the generated next generation of new individuals is compared with the previous optimal individual to select the new optimal individual again. The crossover probability p is selected according to the actual situation. A smaller p can retain more of its own information and maintain the diversity of the group; while a larger p can better learn the excellent information of the group and accelerate the convergence of the algorithm.

[0101] ⑤Variation

[0102] This embodiment divides the iterated generations into three parts: early stage, middle stage and late stage. In each stage, different positions of individuals are reversed according to probability.

[0103] (1) Classify individual attributes according to the different materials, that is, divide individual chromosomes into three categories: A, B, and C. Determine the distribution of ABC positions according to the actual situation. In this example, assuming that the unit price ranking of the materials is a>b>c, the ABC position distribution of the chromosome is:

[0104] [x1,x m ],(x m ,x m+p ],(x m+n ,x m+n+q ]

[0105] In the formula, the amount of materials purchased by m+n+q (m Class A suppliers, n Class B suppliers, q Class C suppliers) suppliers is taken as an individual, and the individual contains m+n+q features (corresponding to the supply quantity of each supplier). m ],(x m ,x m+p ],(x m+n ,x m+n+q ] represents the characteristic range of an individual in the A, B, and C material suppliers. Random numbers are selected for individuals. If the probability number is less than the mutation probability p, a mutation operation is performed.

[0106] [1] If the generation is 1 to gen1, a random number is taken from the A position of the chromosome and a random number change is performed.

[0107] [2] If it is gen1+1~gen2 generation, a random number is taken from the B position of the chromosome and a random number change is performed.

[0108] [3] If gen2+1~gen max In each generation, the C position of the chromosome randomly takes a number for random number change.

[0109] By performing mutation operations at different positions, the diversity of the population can be maintained while preventing premature maturity. At the same time, optimization of different generations can speed up the convergence of the algorithm and improve the accuracy of finding the optimal individual.

[0110] At this point, all steps of the genetic procedure have been completed, and the optimal solution of the model can be obtained by iterating the program.

[0111] Example 3

[0112] Based on Example 1 or Example 2, Figure 2 As shown, after step S3, this embodiment further includes step S4, which is as follows: based on the loss rate of each of the existing N forwarders in each week in the past Y years, a time series analysis method is used to predict the loss rate of the N forwarders in each of the next W weeks; then the enterprise terminal determines the best forwarding plan for each of the next W weeks based on the size of the loss rate predicted in the future.

[0113] In a specific embodiment, before using the time series analysis method to predict the loss rate of the N forwarders in each of the next W weeks, the linear trend method of the adjacent points (using the period number as x and the time series value as y for regression) is used to fill in the missing values, and then the time series analysis is used to calculate the loss rate of the N forwarders in each of the next W weeks.

[0114] Example 4

[0115] Based on the above embodiments 1, 2, and 3, this embodiment further provides an enterprise terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method steps implemented are as follows:

[0116] S1: Receive the historical transaction data of each supplier for preprocessing, score the preprocessed data, and select the top L suppliers;

[0117] S2: Construct a single-objective programming model, use the top L suppliers as variables of the single-objective programming model, and then use genetic algorithms to find the optimal solution;

[0118] S3: Based on the optimal solution, the production volume of the enterprise in each week in the next W weeks and the supply volume of the participating suppliers in each week are obtained to determine the optimal supplier selection plan.

[0119] After step S3, the enterprise terminal also includes predicting the loss rate of the N existing forwarders in each week of the next W weeks using a time series analysis method based on the loss rate of each forwarder in each week of the past Y years; and then the enterprise terminal determines the best forwarding plan for each of the next W weeks based on the predicted loss rate in the future.

[0120] Obviously, the above embodiments of the present invention are only examples for clearly explaining the present invention, and are not intended to limit the implementation methods of the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A method for selecting suppliers and forwarders based on a genetic algorithm single-objective programming model, characterized in that: The steps of the method are as follows: S1: Receive the historical transaction data of each supplier for preprocessing, score the preprocessed data, and select the top L suppliers; S2: Construct a single-objective programming model, use the top L suppliers as variables of the single-objective programming model, and then use genetic algorithms to find the optimal solution; S3: Based on the optimal solution, the production volume of the enterprise in each week in the next W weeks and the supply volume of the participating suppliers in each week are obtained, so as to determine the optimal supplier selection plan; S4: Based on the loss rate of each of the N existing forwarders in each week in the past Y years, the time series analysis method is used to predict the loss rate of the N forwarders in each of the next W weeks; then, based on the predicted loss rate in the future, Q forwarders with the smallest loss rate are selected to determine the best forwarding plan for each of the next W weeks; The pre-processing is specifically as follows: S101: Analyze various indicators based on historical transaction data and quantify each indicator; S102: using the entropy weight method to assign weights to each indicator and determine the weights; S103: Use the TOPSIS evaluation model to score each supplier and select the top L suppliers; The indicators include supply completion coefficient, supply stability coefficient, supply material grade, and production capacity; the indicators are quantified by using one of the following methods: average value, weighted sum ratio, and AHP analysis; The entropy weight method comprises the following steps: D1: Standardize the data of each indicator; D2: Obtain the entropy value and information entropy redundancy of each indicator according to the proportion; D3: Obtain the weights corresponding to each indicator based on the information entropy redundancy obtained in step D2; The TOPSIS evaluation model includes the following steps: F1: normalize the indicator data and standardize the matrix; F2: Calculate the distance between the maximum and minimum values ​​based on the weights obtained by the entropy weight method; F3: Normalize the distance value obtained to get the total score; After obtaining the score coefficients of all suppliers, sort them from largest to smallest and select the top L suppliers, or select L suppliers whose score coefficients are greater than a preset score threshold; The specific steps of constructing a single-objective programming model are as follows: first determine the constraints, then use the minimum production cost as the objective function to construct a single-objective programming model, and then use a genetic algorithm to find the optimal solution of the objective function.

2. The method for selecting suppliers and forwarders based on the genetic algorithm single-objective programming model according to claim 1 is characterized by: The genetic algorithm is used to find the optimal solution of the objective function, as follows: For solving high-dimensional single-objective programming problems, the genetic algorithm is used to directly use the objective function value as the search information and continuously iterate to find the optimal individual.

3. The supplier and forwarder selection method based on the genetic algorithm single-objective programming model according to claim 1 is characterized by: Before using the time series analysis method to predict the loss rate of these N forwarders in each of the next W weeks, the linear trend method of adjacent points is used to fill in the missing values, and then the time series analysis is used to calculate the loss rate of these N forwarders in each of the next W weeks.

Images