Agent distribution decision and inventory management optimization system based on GA-BP algorithm

Abstract

The present invention discloses an agent distribution decision and inventory management optimization system based on GA-BP algorithm. By establishing a collaborative, forecasting and replenishment (CPFR) supply chain management model, it fundamentally realizes that suppliers, agents and the like fairly share relevant information on the supply chain. Each node enterprise on the supply chain shares the information on the supply chain on time, quickly and fairly, and negotiates according to its own capabilities and the information obtained to make a reasonable production or sales plan. In the present invention, by optimizing the weights through a genetic algorithm (GA), the obtained weights are applied to the BP network for training, effectively avoiding falling into a local optimal solution, and through a specific analysis of the enterprise's raw material safety inventory, the various factors affecting the safety inventory are understood in detail. Combined with the characteristics of the enterprise's raw material inventory, a good prediction effect is achieved, the unfavorable status of the enterprise's inventory management is improved, and the enterprise's profits are expanded.

Description

Technical Field

[0001] The present invention relates to the technical field of commodity inventory management, and in particular to an agent distribution decision and inventory management optimization system based on a GA-BP algorithm. Background Art

[0002] With the increasingly fierce competition in the global market, the traditional production and operation model has been difficult to meet the needs of the market. Supplier agency distribution decision-making and large-scale inventory management have been widely used and developed rapidly in the fiercely competitive market environment. In addition, since agency distribution decision-making and large-scale inventory management are often distributed in multiple places in actual inventory management, enterprises need to centrally manage the distributed storage inventory.

[0003] The main strategic means of agent distribution decision-making and large-scale inventory management is to share real-time inventory information, and the high accuracy of safety stock prediction can effectively reduce the cost of inventory management. Since the safety stock is affected by many uncertain factors, and there is a very complex nonlinear relationship between these uncertain factors and decisions, neural networks have unique advantages in solving complex nonlinear problems compared to other methods.

[0004] Traditional BP neural network has some defects such as slow convergence speed and easy to fall into local optimum. Compared with traditional neural network algorithm, GA-BP neural network has more stable modeling model, more times to reach the preset target, fewer fitting data iteration steps, faster to reach the preset result and better data fitting effect. In order to ensure the continuous operation of normal sales activities, a certain amount of inventory is usually held. Ideally, this inventory is exactly equal to the actual demand in the time period, but this ideal situation is difficult to achieve.

[0005] The above situation is mainly caused by: 1. The inventory demand in actual management is very random and not a certain amount; 2. The transportation methods and product supply quality adopted by suppliers are also random and uncertain. Because of this, in the inventory management of enterprises, one part is used to prepare the inventory according to the expected normal sales volume, and the other part is used to prevent the supply uncertainty, unexpected interruption or delay caused by these uncertain factors. Prepared safety or buffer inventory. Summary of the invention

[0006] 1. Technical issues to be solved

[0007] The purpose of the present invention is to solve the problem in the prior art that suppliers and agents cannot fairly share relevant information on the supply chain, and that each node enterprise on the supply chain cannot share information on the supply chain on time, quickly and fairly, and thus cannot make reasonable production or sales plans. An agent distribution decision and inventory management optimization system based on the GA-BP algorithm is proposed.

[0008] 2. Technical solution

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] An agent distribution decision and inventory management optimization system based on a GA-BP algorithm comprises a data acquisition system, a data management system and a data application system, wherein the output end of the data acquisition system is connected to the input end of the data management system, and the output end of the data management system is connected to the input end of the data application system;

[0011] The data collection system includes a supplier data collection module and a CPFR supply chain management module, the data management system includes order management, production management, and GA-BP algorithm distribution decision-making, and the data application system includes agent distribution planning and customer distribution planning;

[0012] The output end of the supplier data acquisition module is connected to the input end of the CPFR supply chain management module, and the output end of the CPFR supply chain management module is respectively connected to the input ends of order management and production management, and the output ends of the order management and production management are both connected to the input end of the GA-BP algorithm distribution decision, and the output end of the GA-BP algorithm distribution decision is connected to the input end of the agent distribution plan and the customer distribution plan.

[0013] Preferably, the CPFR supply chain management module obtains qualitative indicators and constitutes sample data by analyzing the comprehensive supervision and evaluation management records of the industrial supply chain on the production and operation performance of enterprises and agents. In order to facilitate network testing, the data is standardized using a formula, thereby normalizing the original data to the [0.1, 0.9] interval;

[0014]

[0015] Where A is the evaluation result, a is the parameter, i and j are the order management quantity and production quantity respectively.

[0016] Preferably, the CPFR supply chain management module prioritizes coordinating inventory management and distribution decisions by adding data from node enterprises, and selects the optimal number of nodes by observing the training effects of different numbers of nodes. The network input data can be normalized in the following two ways:

[0017]

[0018]

[0019] Preferably, in the GA-BP algorithm distribution decision, the GA genetic algorithm is used to optimize the initial weights of the BP neural network.

[0020] Preferably, the algorithm process of optimizing the initial weights of the BP neural network using the GA genetic algorithm includes the following steps:

[0021] Step 1: First, the connection weights and node thresholds between the network topology structures determined by the neural network are used as the parameters of the genetic algorithm, and the real number encoding method is adopted. Therefore, assuming that the number of input layer nodes is p, the number of output layer nodes is q, and the number of hidden layer nodes is r, the length n of the BP neural network weight threshold length can be obtained by the following calculation formula: n = (p + 1) × r + r + (r × 1) × q;

[0022] Step 2: Determine the range of the connection weights of the GA encoding the length n of each node of the initial value neural network: E∈[x min , x max ], and then assign the calculated uniformly distributed random number of the corresponding range to the gene value, which will generate initial data and predict the market situation based on the initial data;

[0023] Step 3: Then the fitness values ​​of individuals in the group need to be evaluated, the individuals are decoded and assigned to the corresponding connection weights and node thresholds, the learning samples are introduced and the learning error E is calculated. The fitness calculation formula is as follows: The error is obtained as the fitness value;

[0024] Step 4: Then predict the product based on the fitness f;

[0025] Step 5: Use the optimal retention strategy method to select p according to the GA selection in step 4 i , Cross X i and variant P m Search the global and near the solution point with a high probability in the whole variable space, and use the BP neural network to converge quickly and accurately, integrate the advantages of the two, and organically combine the two. Use the genetic algorithm to train the neural network weights and topological structure at the same time, which can avoid falling into the local minimum, improve the algorithm convergence speed, and quickly get the global optimal solution to the problem, ensuring the accuracy of the company's prediction of the market situation, and realizing the full agent distribution of inventory management and the optimal sales plan;

[0026] Step 6: The optimal solution obtained by the genetic algorithm is then assigned to the weight of the neural network. The BP algorithm is used to optimize the weight and threshold of the neural network. Through the weight and threshold, a specific analysis of the agent distribution decision and large-scale inventory management optimization is performed.

[0027] Preferably, the product prediction operation based on fitness f in step 4 includes the following steps:

[0028] S1: Select operation, select the roulette wheel selection method. If the population size is M, the probability that an individual with a high fitness value is selected and passed to the next generation is: where p i is the probability of selection in each generation, f i is the fitness value of each generation, and i is the number of generations.

[0029] S2: Crossover operation, select the arithmetic crossover operator, individuals X1 and X2 in the parent generation, with a crossover probability P c After the crossover operation, the generated offspring individuals are X′1=aX1+(1+a)X2, X′2=aX1+(1+a)X2, a=∈(0,1), where a is a parameter.

[0030] S3: mutation operation, using uniform mutation operator, individual X i The probability of mutation of a gene is P m , according to this probability, it will be uniformly distributed in [x min , x max ]Replace the original value with a random number in the interval.

[0031] Preferably, in the process of obtaining the individual evolution in this individual in S2, the individual closest to the optimal solution in the population is obtained, the above operation is continuously performed, and finally the optimal individual in the last generation population is decoded, so as to obtain the optimal prediction of inventory management that meets the requirements.

[0032] Preferably, every payment of the customers and agents in S3 will be recorded in detail. For some controversial financial records, the system will automatically remind you to eliminate the interference of abnormal data on the prediction results.

[0033] 3. Beneficial effects

[0034] Compared with the prior art, the advantages of the present invention are:

[0035] (1) In the present invention, the GA-BP algorithm is applied to effectively solve the complex nonlinear problems affecting safety inventory factors under the traditional inventory management model, so that the functions of the inventory management system can be utilized to the maximum extent, thereby providing decision support for the inventory management of logistics enterprises to improve the decision-making ability, efficiency and accuracy of enterprises in inventory management.

[0036] (2) In the present invention, relatively accurate market demand information can be obtained through the integration of planning, forecasting and replenishment through the electronic information platform, so that enterprises can better forecast demand and manage inventory, establish common goals with upstream and downstream enterprises, strive for comprehensive benefits on the basis of mutual trust and commitment, and share business information to coordinate and arrange replenishment plans.

[0037] (3) In the present invention, by effectively controlling the logistics costs of each link in the supply chain, the risk of squeeze is greatly reduced, the efficiency of commodity allocation is improved, and the inventory management capabilities of enterprises are optimized. By building a mature logistics system, enterprises can cooperate with each other to reduce costs and share risks, continuously improve their own warehousing management capabilities, reduce inventory backlogs of commodities, achieve full utilization of resources, and improve the level of logistics services. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a general structure diagram of the agent distribution decision and inventory management optimization system based on the GA-BP algorithm proposed by the present invention;

[0039] Figure 2 This is a flow chart of the GA-BP algorithm in the agent distribution decision and inventory management optimization system based on the GA-BP algorithm proposed in the present invention. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0041] Embodiment 1:

[0042] Reference Figure 1 , an agent distribution decision and inventory management optimization system based on GA-BP algorithm, including a data acquisition system, a data management system and a data application system, the output end of the data acquisition system is connected to the input end of the data management system, and the output end of the data management system is connected to the input end of the data application system;

[0043] The data collection system includes supplier data collection module and CPFR supply chain management module; the data management system includes order management, production management and GA-BP algorithm distribution decision; the data application system includes agent distribution plan and customer distribution plan;

[0044] The output end of the supplier data acquisition module is connected to the input end of the CPFR supply chain management module. The output end of the CPFR supply chain management module is connected to the input ends of order management and production management respectively. The output ends of order management and production management are both connected to the input end of the GA-BP algorithm distribution decision. The output end of the GA-BP algorithm distribution decision is connected to the input end of the agent distribution plan and the customer distribution plan.

[0045] In the present invention, by establishing a collaborative, forecasting and replenishment (CPFR) supply chain management model, it is fundamentally achieved that suppliers, agents, etc. fairly share relevant information on the supply chain. Each node enterprise on the supply chain shares the information on the supply chain on a timely, fast and fair basis, and negotiates based on its own capabilities and the information obtained to make reasonable production or sales plans.

[0046] In the present invention, the weights are optimized by using a genetic algorithm (GA), and the obtained weights are applied to a BP network for training, thereby effectively avoiding falling into a local optimal solution.

[0047] In the present invention, through the specific analysis of the enterprise's raw material safety inventory, various factors affecting the safety inventory are understood in detail, and combined with the characteristics of the enterprise's raw material inventory, a good prediction effect is achieved, the unfavorable status quo of the enterprise's inventory management is improved, and the enterprise's profits are expanded.

[0048] Embodiment 2:

[0049] It has the implementation content of the above-mentioned embodiment, wherein the specific implementation of the above-mentioned embodiment can refer to the above description, and the embodiment here will not be repeated in detail; and in the embodiment of the present application, it is different from the above-mentioned embodiment in that:

[0050] In this embodiment, CPFR, namely coordinated planning, forecasting and replenishment, is essentially coordination and cooperation. Through collaborative planning, forecasting and replenishment inventory management, each node enterprise can obtain relatively accurate market demand information through the electronic information platform, so that the enterprise can better predict demand and manage inventory.

[0051] In this embodiment, based on the deep collaboration (CPFR) solution with the upstream and downstream of the supply chain, this requires the e-commerce platform to strengthen data sharing with suppliers at the data levels of sales volume, inventory, traffic and conversion rate, product after-sales evaluation, industry public opinion, etc., and enhance the service quality of suppliers. In this way, the entire process of "collaboration" can be transparent, visible, and feedback can be immediately provided, further forming a more sticky supply chain alliance.

[0052] In this embodiment, the CPFR supply chain management module obtains qualitative indicators and constitutes sample data by analyzing the comprehensive supervision and evaluation management records of the industrial supply chain on the production and operation performance of enterprises and agents. In order to facilitate network testing, the data is standardized using a formula, thereby normalizing the original data to the [0.1, 0.9] interval;

[0053]

[0054] Where A is the evaluation result, a is the parameter, i and j are the order management quantity and production quantity respectively.

[0055] In this embodiment, the CPFR supply chain management module prioritizes coordinating inventory management and distribution decisions by adding data from node companies, and selects the optimal number of nodes by observing the training effects of different numbers of nodes. The network input data can be normalized in the following two ways:

[0056]

[0057]

[0058] In this embodiment, based on the above normalization, each node enterprise of the present invention can obtain relatively accurate market demand information through the electronic information platform, so that the enterprise can better predict demand and manage inventory.

[0059] In this embodiment, big data and intelligent AI algorithms are used to analyze product sales hotspots and trends across the entire network, predict changes in commodity sales, and guide retail plans that match the market. This in turn guides suppliers’ inventory management. At the same time, by establishing a CPFR supply chain strategic cooperation model, data can generate value flows, suppliers can share analysis data from e-commerce platforms in a timely manner, adjust production plans and replenishment plans in a timely manner, and build comprehensive collaboration from planning to forecasting and replenishment processes.

[0060] Embodiment 3:

[0061] It has the implementation content of the above-mentioned embodiment, wherein the specific implementation of the above-mentioned embodiment can refer to the above description, and the embodiment here will not be repeated in detail; and in the embodiment of the present application, it is different from the above-mentioned embodiment in that:

[0062] Reference Figure 2 In the GA-BP algorithm distribution decision, the GA genetic algorithm is used to optimize the initial weights of the BP neural network. The algorithm process of optimizing the initial weights of the BP neural network by the GA genetic algorithm includes the following steps:

[0063] Step 1: First, the connection weights and node thresholds between the network topology structures determined by the neural network are used as the parameters of the genetic algorithm, and the real number encoding method is adopted. Therefore, assuming that the number of input layer nodes is p, the number of output layer nodes is q, and the number of hidden layer nodes is r, the length n of the BP neural network weight threshold length can be obtained by the following calculation formula: n = (p + 1) × r + r + (r × 1) × q;

[0064] Step 2: Determine the range of the connection weights of the GA encoding the length n of each node of the initial value neural network: E∈[x min , x max ], and then assign the calculated uniformly distributed random number of the corresponding range to the gene value, which will generate initial data and predict the market situation based on the initial data;

[0065] Step 3: Then the fitness values ​​of individuals in the group need to be evaluated, the individuals are decoded and assigned to the corresponding connection weights and node thresholds, the learning samples are introduced and the learning error E is calculated. The fitness calculation formula is as follows: The error is obtained as the fitness value;

[0066] Step 4: Then predict the product based on the fitness f, where the prediction operation includes the following steps:

[0067] S1: Select operation, select the roulette wheel selection method. If the population size is M, the probability that an individual with a high fitness value is selected and passed to the next generation is: where p i is the probability of selection in each generation, f i is the fitness value of each generation, and i is the number of generations.

[0068] S2: Crossover operation, select the arithmetic crossover operator, individuals X1 and X2 in the parent generation, with a crossover probability P c After the crossover operation, the generated offspring individuals are X′1=aX1+(1+a)X2, X′2=aX1+(1+a)X2, a=∈(0,1), where a is a parameter.

[0069] S3: mutation operation, using the uniform mutation operator. i The probability of mutation of a gene is P m , according to this probability, it will be uniformly distributed in [x min , x max ] The random numbers in the interval replace the original values, thereby reducing the interference of random data in the prediction. Every payment from customers and agents will also be recorded in detail. For some controversial financial records, the system will automatically remind you to eliminate the interference of abnormal data on the prediction results.

[0070] Step 5: Use the optimal retention strategy method to select p according to the GA selection in step 4 i , Cross X i and variant P m Search the global and near the solution point with a high probability in the whole variable space, and use the BP neural network to converge quickly and accurately, integrate the advantages of the two, and organically combine the two. Use the genetic algorithm to train the neural network weights and topological structure at the same time, which can avoid falling into the local minimum, improve the algorithm convergence speed, and quickly get the global optimal solution to the problem, ensuring the accuracy of the company's prediction of the market situation, and realizing the full agent distribution of inventory management and the optimal sales plan;

[0071] Step 6: The optimal solution obtained by the genetic algorithm is assigned to the weight of the neural network, and the BP algorithm is used to optimize the weight and threshold of the neural network. Through the weight and threshold, the specific analysis of the agent distribution decision and large-scale inventory management optimization is carried out. The difference between the inventory forecast value and the actual value is small, and the prediction accuracy is high. It can well predict the future safety inventory, and then provide decision support for the inventory management of logistics enterprises to improve the decision-making ability, efficiency and accuracy of enterprises in inventory management.

[0072] In this example, the financial operation goals are achieved through the CPFR supply chain management model. In order to better control cash flow, by optimizing the entire supply chain and demand management, with the assistance of relevant information tools, the inventory level is reduced as much as possible while ensuring timely delivery, and the risks of inventory backlog, depreciation and scrapping are reduced.

[0073] In this example, the BP network is optimized by the GA genetic algorithm to achieve scientific management and control of agent distribution decisions and inventory procurement, ensuring the healthy operation and continuous production of the enterprise, allowing the enterprise to focus on managing a few important inventory items, which not only strengthens management but also saves costs, and is of great help in helping merchants improve warehouse management efficiency.

[0074] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. Agent distribution decision and inventory management optimization system based on GA-BP algorithm, including data collection system, data management system and data application system, It is characterized in that The output end of the data acquisition system is connected to the input end of the data management system, and the output end of the data management system is connected to the input end of the data application system; The data collection system includes a supplier data collection module and a CPFR supply chain management module, the data management system includes order management, production management, and GA-BP algorithm distribution decision-making, and the data application system includes agent distribution planning and customer distribution planning; The output end of the supplier data acquisition module is connected to the input end of the CPFR supply chain management module, the output end of the CPFR supply chain management module is respectively connected to the input ends of the order management and production management, the output ends of the order management and production management are both connected to the input end of the GA-BP algorithm distribution decision, and the output end of the GA-BP algorithm distribution decision is connected to the input end of the agent distribution plan and the customer distribution plan; In the GA-BP algorithm distribution decision, the initial weights of the BP neural network are optimized by using the GA genetic algorithm; The algorithm process of optimizing the initial weights of the BP neural network by the GA genetic algorithm includes the following steps: Step 1: First, the connection weights and node thresholds between the network topology structures determined by the neural network are used as the parameters of the genetic algorithm, and the real number encoding method is adopted. Therefore, assuming that the number of input layer nodes is p, the number of output layer nodes is q, and the number of hidden layer nodes is r, the length n of the BP neural network weight threshold length can be obtained by the following calculation formula: n = (p + 1) × r + r + (r × 1) × q; Step 2: Determine the range of the connection weights of the length n encoding of each node of the initial value neural network by GA: W∈[x min ,x max ], and then assign the calculated uniformly distributed random number of the corresponding range to the gene value, which will generate initial data and predict the market situation based on the initial data; Step 3: Then the fitness values ​​of individuals in the group need to be evaluated, the individuals are decoded and assigned to the corresponding connection weights and node thresholds, the learning samples are introduced and the learning error E is calculated. The fitness calculation formula is as follows: The error is obtained as the fitness value; Step 4: Then predict the product based on the fitness f. The prediction operation includes the following steps: S1: Selection operation, select the roulette wheel selection method. If the population size is M, the probability that an individual with a high fitness value is selected and passed to the next generation is: where p i is the probability of selection in each generation, f i is the fitness value of each generation, i is the number of generations; S2: Crossover operation, select the arithmetic crossover operator, individual X in the parent generation 1 and X 2 , with the crossover probability P c Perform crossover operation, parent individual X 1 and X 2 The offspring individuals generated by the crossover operation are X' 1 =aX 1 +(1+a)X 2 , X' 2 =aX 1 +(1+a)X 2 , a=∈(0,1), a is a parameter; S3: mutation operation, using the uniform mutation operator, individual X i The probability of mutation of a gene is P m , according to this probability, it will be uniformly distributed in [x min ,x max ]The random number in the interval replaces the original value; In S2, the individual closest to the optimal solution in the population is obtained during the individual evolution process. The above operations are continuously performed, and finally the optimal individual in the last generation of the population is decoded, thereby obtaining the optimal prediction for inventory management that meets the requirements; Step 5: Adopt the optimal retention strategy method, search the area near the point of the global optimal solution in the whole variable space according to the selection probability, crossover individual and mutation probability of GA in step 4, and use the characteristics of BP neural network that can converge quickly and accurately, integrate the advantages of the two, and organically combine the two. Use genetic algorithm to train the weights and topological structure of neural network at the same time, which can avoid falling into local minimum, improve the convergence speed of algorithm, and quickly obtain the global optimal solution of the problem, so as to ensure the accuracy of enterprise's prediction of market situation and realize the full agent distribution of inventory management and the optimal sales plan; Step 6: The optimal solution obtained by the genetic algorithm is then assigned to the weight of the neural network. The BP algorithm is used to optimize the weight and threshold of the neural network. Through the weight and threshold, a specific analysis of the agent distribution decision and large-scale inventory management optimization is performed.

2. The agent distribution decision and inventory management optimization system based on GA-BP algorithm according to claim 1, It is characterized in that The CPFR supply chain management module obtains qualitative indicators and constitutes sample data by analyzing the comprehensive supervision and evaluation management records of the industrial supply chain on the production and operation performance of enterprises and agents. In order to facilitate network testing, the data is standardized using formulas, thereby normalizing the original data to the [0.1, 0.9] interval.

3. The agent distribution decision and inventory management optimization system based on GA-BP algorithm according to claim 2, It is characterized in that In the CPFR supply chain management module, inventory management and distribution decisions are coordinated by adding data of node enterprises, and the optimal number of nodes is selected by observing the training effects of different numbers of nodes.

4. The agent distribution decision and inventory management optimization system based on GA-BP algorithm according to claim 1, It is characterized in that In the S3, every payment made by customers and agents will be recorded in detail. For some controversial financial records, the system will automatically remind you to eliminate the interference of abnormal data on the prediction results.

Images