Transaction decision-making method for rubber wood cross-border trade supply chain
By obtaining and processing multi-source data of rubber wood transaction decisions, a collaborative filtering transaction decision-making method generation model is built, and prediction is made using LSTM neural network and Monte Carlo simulation, the problem of insufficient accurate transaction decision-making of rubber wood cross-border trade supply chain is solved, and decision accuracy and supply chain resilience are improved.
Patent Information
- Application Number
- CN202510327377.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-19
- Publication Date
- 2025-06-17
AI Technical Summary
It is difficult for the existing technology to combine economic, production and supply data and decision-making risk data to generate predictions based on historical related data, resulting in insufficient transaction decision-making in the cross-border trade supply chain of rubber wood, which can easily lead to inventory backlog or cost overrun.
By obtaining the price volatility coefficient, weather impact factor and risk-return balance value that affects rubber wood trading decisions, perform feature processing and feature links, build a model for generating collaborative filtering transaction decision-making, and use LSTM neural network and Monte Carlo simulation for prediction and data processing to generate rubber wood trading decision-making methods.
It improves the accuracy of rubber wood transaction decisions, helps transaction decision-makers to make timely procurement and supplier adjustments, reduces cross-border procurement costs, improves supply chain resilience, and reduces labor costs and brain power consumption.
Smart Images
Figure CN120163637A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of intelligent decision-making, and particularly relates to a trading decision-making method for the cross-border trade supply chain of rubber wood. Background Art
[0002] As a sustainable resource, the demand for rubber wood has increased rapidly in the Chinese furniture and decoration industries in recent years. However, its supply chain is affected by multiple factors such as fluctuations in the Thai source supply (such as weather and shipping costs), market price fluctuations, and international trade policies, making it difficult for purchasers to accurately grasp the best trading opportunity. The traditional supply chain management model and the output of the trading decision-making method rely extremely on manual experience, lacking real-time data analysis and dynamic prediction capabilities, which are prone to inventory backlogs or cost overruns.
[0003] As a digital platform for the wood industry, Wood Cloud has built digital tools covering rubber wood procurement, trading, and warehousing (such as the "Rubber Wood Ordering System", "Wood Cloud Collection", "Wood Cloud Warehouse", "Wood Cloud Commerce", etc.), and integrated global supplier information through big data. However, the existing system has not been deeply combined with dynamic variables such as weather and production efficiency for intelligent decision-making support. In terms of trading decision-making, there are certain impacts caused by natural laws, and the impact of nature on rubber wood supply is often ignored in the existing technology. Trading decisions are often considered from the human economic and social levels but only from one aspect.
[0004] Through the above analysis, the problems existing in the prior art that need to be solved are as follows:
[0005] How to combine data at the economic level, production supply level, and decision-making risk level, and generate predictions based on historical relevant data. Combine the determined correct relevant prediction data and historical data to generate scenarios for trading decisions in the cross-border trade supply chain of rubber wood. Build a personalized simple and practical machine learning collaborative filtering model based on multi-source data, so as to generate a rubber wood trading decision-making method to assist trading decision-makers in purchasing rubber wood on time and adjusting suppliers, and scientifically manage the rubber wood supply chain. How to perform targeted effective feature processing on the various data parameters that affect the formulation of the trading decision-making method, and how to improve the simple and practical machine learning model to improve the accuracy of the trading decision-making method formulation, and determine the most suitable application scenario model after training. Generate corresponding trading decision-making methods by improving the model applicability and accuracy to assist in reducing cross-border procurement costs and enhancing supply chain resilience, and reducing labor costs and mental consumption, and improving the accuracy of procurement timing and supplier adaptation rate. For this reason, a trading decision-making method for the cross-border trade supply chain of rubber wood is proposed. Summary of the Invention
[0006] To solve the above technical problems, the present invention proposes a transaction decision-making method for the cross-border trade supply chain of rubber wood, and the method includes:
[0007] Obtain the price fluctuation coefficient, weather impact factor, and risk-return balance value that affect the rubber wood transaction decision-making; and obtain the corresponding rubber wood transaction decision-making method;
[0008] Process the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubber wood transaction decision-making to obtain the rubber wood transaction decision-making characteristics;
[0009] Construct a collaborative filtering transaction decision-making method generation model according to the rubber wood transaction decision-making characteristics and the corresponding transaction decision-making method;
[0010] Use the historical transaction decision-making method and corresponding parameters to test the transaction decision-making method generation error rate of the collaborative filtering transaction decision-making method generation model;
[0011] And generate a timely transaction decision-making method based on the collaborative filtering transaction decision-making method generation model with an error rate less than the error value, and output the recommended best procurement timing and the matching supplier list according to the timely transaction decision-making method.
[0012] In this embodiment, as shown in Figure 1 Before constructing and processing the relevant feature data into a model, the data will be preprocessed and cleaned.
[0013] Further, the price fluctuation coefficient is calculated using the price fluctuation trend in the first time period predicted by the LSTM neural network and the historical price fluctuation trend in the second time period.
[0014] Further, the weather impact factor is obtained by correcting the felling rate based on the rubber wood growth cycle rate and climate index in the region.
[0015] Further, the risk-return balance value is obtained by integrating the sea freight cost, tariff fluctuation, inventory turnover rate, and supplier credit risk through Monte Carlo simulation. For the Monte Carlo simulation of the sea freight cost, the tariff fluctuation, and the inventory turnover rate, a normal distribution function model is used to generate the simulation data set.
[0016] Further, the supplier credit risk uses a lognormal distribution model to generate the simulation data set.
[0017] Further, perform mean processing on the data in the generated simulation dataset that is lower than the error value to obtain the average shipping cost, the average tariff fluctuation, the average inventory turnover rate, and the average supplier credit risk. Process the average shipping cost, the average tariff fluctuation, the average inventory turnover rate, and the average supplier credit risk to obtain the risk-return balance value.
[0018] Further, before processing the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubberwood trading decision to obtain the rubberwood trading decision characteristics, perform removal and correction on the rubberwood trading decision characteristics to remove outliers and improve the generalization ability of the model.
[0019] Further, the processing of the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubberwood trading decision to obtain the rubberwood trading decision characteristics is carried out in a feature linking manner, and the feature linking manner includes feature horizontal splicing and feature vertical splicing.
[0020] Further, use the historical trading decision method and corresponding parameters to test the error rate generated by the trading decision method of the collaborative filtering trading decision method generation model;
[0021] Among them, the historical trading decision method and corresponding parameters include the historical price fluctuation coefficient, the historical weather impact factor, and the historical risk-return balance point used outside the training model, as well as the historical actual rubberwood trading decision method;
[0022] And generate the historical trading decision method based on the collaborative filtering trading decision method generation model with an error rate less than the error value. Calculate according to the error value between the historical trading decision method and the historical actual rubberwood trading decision method, and the error value is taken as 0.15. When those skilled in the art make rubberwood trading decisions, the error rate given by the model cannot exceed a certain value to ensure the monitoring and reasonable use of the trained model.
[0023] In this embodiment, the output of the rubberwood trading decision method is carried out according to the value of T(R). There are a total of five types of rubberwood trading decision methods, which are respectively for the cases where the value of T(R) is less than -3, the value of T(R) is greater than or equal to -3 and less than 0, the value of T(R) is 0, the value of T(R) is greater than 0 and less than 3, and the value of T(R) is greater than or equal to 3. The five types of rubberwood trading decision methods respectively correspond to different procurement postponement times, and the recommendation and selection of suppliers with lower ratings, low ratings, medium ratings, high ratings, and higher ratings are set in the wood cloud. Here, the procurement postponement time and the supplier rating are set by technicians according to specific actual situations to meet the needs.
[0024] The present invention combines data at the economic level, production supply level, and decision-making risk level through feature fusion, uses the LSTM neural model to generate predictions based on historical relevant data, combines Monte Carlo simulation to generate and determine low-error relevant simulation data and historical data based on set thresholds, generates scenarios for transaction decisions in the cross-border trade supply chain of rubber wood, and constructs a personalized simple and practical Fisher criterion machine learning collaborative filtering model based on multi-source data, thereby generating a rubber wood transaction decision-making method to assist transaction decision-makers in purchasing rubber wood on time and adjusting suppliers, and scientifically managing the rubber wood supply chain. By performing targeted effective feature linking processing on various data parameters affecting the formulation of the transaction decision-making method, and improving the personalized scenario application of the Fisher criterion machine learning collaborative filtering model by correcting the classification result values, the accuracy of formulating the transaction decision-making method is improved, and the classification results are tested using historical relevant data. Based on the error value, the applicability of the trained model is determined. By improving the model applicability and accuracy, corresponding transaction decision-making methods are generated to assist in reducing cross-border procurement costs, enhancing supply chain resilience, reducing labor costs and mental consumption, and improving the accuracy of procurement timing and supplier adaptation rate.
[0025] The present invention combines weather and climate conditions with supply chain-related data for the first time, breaks through the limitations of single-market analysis, and uses an improved collaborative filtering model based on machine learning to accurately generate a transaction decision-making method.
[0026] More embodiments and improvement effects of the present invention will be further introduced in combination with the drawings and specific embodiments. Description of the Drawings
[0027] Figure 1 is a flowchart of a transaction decision-making method for the cross-border trade supply chain of rubber wood according to the present invention;
[0028] Figure 2 is a schematic diagram of the rubber wood price trend in a certain time period for calculating the price fluctuation coefficient in the present invention;
[0029] Figure 3 is a schematic diagram of the principle of an improved classifier based on the Fisher criterion in the present invention;
[0030] Figure 4 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Embodiments
[0031] Next, in combination with the drawings and specific embodiments, the invention will be further described.
[0032] In the first aspect of the present invention, a transaction decision-making method for the cross-border trade supply chain of rubber wood is provided, and the method includes:
[0033] Obtain the price fluctuation coefficient, weather impact factor, and risk-return balance value that affect the rubberwood trading decision; and obtain the corresponding rubberwood trading decision method;
[0034] Process the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubberwood trading decision to obtain rubberwood trading decision characteristics;
[0035] Construct a collaborative filtering trading decision method generation model according to the rubberwood trading decision characteristics and the corresponding trading decision method;
[0036] Use the historical trading decision method and corresponding parameters to test the trading decision method generation error rate of the collaborative filtering trading decision method generation model;
[0037] And generate an immediate trading decision method based on the collaborative filtering trading decision method generation model with an error rate less than the error value, and output the recommended best purchase timing and the matching supplier list according to the immediate trading decision method.
[0038] Furthermore, the price fluctuation coefficient is calculated by using the price fluctuation trend in the first time period predicted by the LSTM neural network and the historical price fluctuation trend in the second time period. By taking the difference between the predicted future price fluctuation trend and the historical price fluctuation trend in the same time period, it provides an objective and accurate data basis for the generation of the decision method, improves the generalization ability of the model in subsequent model training and the accurate generation of the decision method, and enhances the applicability of the model to the scenario:
[0039]
[0040] In the formula, P c is the price fluctuation coefficient, the first time period and the second time period are different time periods with the same number of days, N represents the total number of days in the first time period and the second time period, t is the t-th day within the time period, is the rubberwood price on the t-th day of the first time period, is the rubberwood price on the t-th day of the second time period, μ1 is the average rubberwood price in the first time period, and μ2 is the average rubberwood price in the second time period.
[0041] Furthermore, the weather impact factor is obtained by correcting the harvesting rate with the rubberwood growth cycle rate and climate index in the region:
[0042]
[0043] In the formula, F w is the weather impact factor, δ CIt is the growth cycle rate of Hevea brasiliensis, obtained by the ratio of the average felling age and the optimal felling age of Hevea brasiliensis at the right time in the region, F r is the felling rate, represents the climate index. Since rainfall and temperature have a great impact on the felling of Hevea brasiliensis and their dimensions are different, in order to represent the positive and negative impacts of the weather impact factors on it, it is processed into a dimensionless form, and the absolute value of the logarithmic function is used to correct the numerical value to reduce the occurrence of overly large data differences, so as to improve the accuracy for subsequent model training and decision-making method generation.
[0044] In this embodiment, since the weather will affect the supply of Hevea brasiliensis, for example, when the rainfall exceeds 200 mm, the felling efficiency will decrease by about 30%. The present invention takes into account various natural influencing factors on the supply of Hevea brasiliensis and obtains a corrected value of the comprehensive influence to correspondingly correct and adjust the input features and model parameters of the subsequent trading decision model, so as to obtain a more accurate decision-making trading method.
[0045] Furthermore, the risk-return balance value is obtained by integrating the sea freight cost, tariff fluctuations, inventory turnover rate, and supplier credit risk through Monte Carlo simulation. The Monte Carlo simulation uses the normal distribution function model to generate simulation data sets for the sea freight cost, the tariff fluctuations, and the inventory turnover rate:
[0046]
[0047] In the formula, m is the characteristic value of the sea freight cost or the tariff fluctuation or the inventory turnover rate, F(m) is the normal distribution function, σ is the standard deviation of the existing data, μ m is the mean value of m, and e is the natural constant.
[0048] Furthermore, the supplier credit risk uses the lognormal distribution model to generate simulation data sets:
[0049]
[0050] In the formula, lnm is the supplier credit risk indication number, F(lnm) is the lognormal distribution function, σ is the standard deviation of the pre-set supplier credit risk data, μ m is the mean value of m, and e is the natural constant. This is because the sea freight cost, the tariff fluctuations, and the inventory turnover rate are obtained by using the corresponding objective data, while the supplier credit risk is usually set by those skilled in the art. In order to ensure the reduction of the dispersion degree of this type of data, the lognormal distribution model is used to generate simulation data sets.
[0051] Further, perform mean processing on the data in the generated simulation dataset that is lower than the error value to obtain the mean of shipping costs, the mean of tariff fluctuations, the mean of inventory turnover rates, and the mean of supplier credit risks. Process the mean of shipping costs, the mean of tariff fluctuations, the mean of inventory turnover rates, and the mean of supplier credit risks to obtain the risk-return balance value:
[0052]
[0053] In the formula, R s is the risk-return balance value, T l is the mean of inventory turnover rates, S c is the mean of shipping costs, T f is the mean of tariff fluctuations, S r is the mean of the supplier credit risks. The higher the risk-return balance value, the more balanced the risk and return. The inventory turnover rate is the number of times the inventory goods are turned over within a certain period of time and is an indicator reflecting the speed of inventory turnover. The larger the turnover rate, the better the sales situation. The higher the supplier credit risk, the greater the risk. Similarly, the greater the fluctuations in shipping costs and tariffs, the higher the corresponding risk. Those skilled in the art should be aware, so the positive and negative proportional impact calculations are performed in the formula.
[0054] In this embodiment, since there is a situation where the relevant characteristic parameters affecting the risk-return balance value have a small amount of data, simulation data and error value judgment are used to perform data expansion generation to cope with the large data required for model training, thereby improving the generation accuracy of the model for subsequent transaction decision-making methods. The error value in this embodiment is defined within a certain range of errors based on the historical mean of shipping costs, the historical mean of tariff fluctuations, the historical mean of inventory turnover rates, and the historical mean of supplier credit risks. The error value is generally set to 0.5 times the ratio of each characteristic mean according to the mean situation.
[0055] Further, before processing the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubberwood transaction decision to obtain the rubberwood transaction decision characteristics, perform removal and correction on the rubberwood transaction decision characteristics to remove outliers to improve the generalization ability of the model.
[0056] Further, the processing of the price fluctuation coefficient, the weather impact factor, and the risk-return balance point that affect the rubberwood transaction decision to obtain the rubberwood transaction decision characteristics is performed in a feature linking manner. The feature linking manner includes feature horizontal splicing and feature vertical splicing.
[0057] Further, use the historical transaction decision-making method and corresponding parameters to test the transaction decision-making method generation error rate of the collaborative filtering transaction decision-making method generation model;
[0058] The historical trading decision-making method and corresponding parameters include the historical price volatility coefficient, historical weather impact factor, and historical risk-return balance point used outside the training model, as well as the historical actual rubberwood trading decision-making method;
[0059] And a model for generating a historical trading decision-making method is generated based on a collaborative filtering trading decision-making method with an error rate less than an error value. The error value is calculated based on the difference between the historical trading decision-making method and the historical actual rubberwood trading decision-making method, and the error value is set to 0.15. When a person skilled in the art makes a rubberwood trading decision, the error rate of the trading decision given by the model cannot exceed a certain value to ensure the monitoring and reasonable use of the trained model.
[0060] Further, the collaborative filtering trading decision-making method generation model uses a collaborative filtering algorithm improved based on machine learning:
[0061] The calculation formula of the collaborative filtering algorithm improved based on machine learning is as follows:
[0062]
[0063] In the formula, T(R) is the output rubberwood trading decision-making method, W T is the normal vector perpendicular to the hyperplane, obtained through model training, R is the rubberwood trading decision-making feature, R s is the risk-return balance value, P c is the price volatility coefficient, F w is the weather impact factor.
[0064] The present invention combines data at the economic level, production and supply level, and decision-making risk level through feature fusion, and uses the LSTM neural model to generate predictions based on historical relevant data. By combining Monte Carlo simulation to generate and determine low-error relevant simulation data and historical data based on set thresholds, for the transaction decision-making scenario of the rubberwood cross-border trade supply chain, a personalized construction of a simple and practical Fisher criterion machine learning collaborative filtering model is carried out based on multi-source data, so as to generate a rubberwood transaction decision-making method to assist transaction decision-makers in purchasing rubberwood on time and adjusting suppliers, and scientifically manage the rubberwood supply chain. By performing targeted and effective feature linking processing on various data parameters affecting the formulation of the transaction decision-making method, and improving the personalized scenario application of the Fisher criterion machine learning collaborative filtering model by correcting the classification result values, the accuracy of formulating the transaction decision-making method is improved, and the classification results are tested using historical relevant data. Based on the error value, the applicability of the trained model is determined. By improving the model applicability and accuracy, corresponding transaction decision-making methods are generated to assist in reducing cross-border procurement costs and enhancing the resilience of the supply chain, and reducing labor costs and mental consumption, and improving the accuracy of procurement timing and the supplier adaptation rate.
[0065] Of course, it can be understood that each embodiment of the present invention can achieve one of the effects alone, and the combination of multiple embodiments of the present invention can achieve all of the above effects. However, it is not required that each embodiment of the present invention achieve all of the above advantages and effects, because each embodiment of the present invention can constitute an independent technical solution and make one or more contributions to the prior art.
[0066] For the part of the module structure not specifically defined in the present invention, it shall be subject to the content recorded in the prior art. The prior art mentioned in the foregoing background art part and specific embodiment part of the present invention can be used as a part of the present invention to understand the meaning of some technical features or parameters. The protection scope of the present invention shall be subject to the content actually recorded in the claims.
Claims
1. A transaction decision method for a cross-border rubber wood trade supply chain, characterized in that: The method comprises: Obtain the price volatility coefficient, weather influencing factors and risk-return balance value that affect rubber wood trading decisions; and obtain the corresponding rubber wood trading decision-making method; The price fluctuation coefficient, weather influencing factor and risk-return balance point that affect the rubber wood trading decision are processed to obtain the rubber wood trading decision feature; Constructing a collaborative filtering transaction decision method generation model based on the rubber wood transaction decision characteristics and the corresponding transaction decision method; Using historical transaction decision methods and corresponding parameters to test the transaction decision method generation error rate of the collaborative filtering transaction decision method generation model; A timely transaction decision method is generated based on a collaborative filtering transaction decision method generation model in which the error rate is less than the error value, and a recommended optimal purchasing time and a matching supplier list are output according to the timely transaction decision method.
2. A transaction decision-making method for a cross-border trade supply chain of rubber wood as claimed in claim 1, characterized in that: The price volatility coefficient is calculated by using the price volatility trend of the first time period predicted by the LSTM neural network and the historical price volatility trend of the second time period. By taking differential values between the predicted future price volatility trend and the historical price volatility trend in the same time period, an objective and accurate data basis is provided for the generation of a decision-making method, which improves the generalization ability of the model in subsequent model training and accurate generation of decision-making methods, and enhances the applicability of the model to scenario applications.
3. A transaction decision-making method for a cross-border trade supply chain of rubber wood as claimed in claim 1, characterized in that: The weather influencing factor is obtained by correcting the harvesting rate based on the growth cycle rate of regional rubber trees and the climate index.
4. A transaction decision method for a cross-border rubber wood trade supply chain as claimed in claim 2, characterized in that: The risk-return balance value is obtained by integrating shipping costs, tariff fluctuations, inventory turnover rate and supplier credit risk processing through Monte Carlo simulation. The Monte Carlo simulation of the shipping costs, tariff fluctuations and inventory turnover rate uses a normal distribution function model to generate a simulated data set.
5. A transaction decision method for a cross-border rubber wood trade supply chain as claimed in claim 4, characterized in that: The supplier credit risk is simulated using a logarithmic normal distribution model to generate a data set.
6. A transaction decision method for a cross-border rubber wood trade supply chain as claimed in claim 5, characterized in that: The data below the error value in the generated simulation data set are averaged to obtain the average shipping cost, the average tariff fluctuation, the average inventory turnover rate and the average supplier credit risk. The average shipping cost, the average tariff fluctuation, the average inventory turnover rate and the average supplier credit risk are processed to obtain the risk-return balance value.
7. A transaction decision-making method for a cross-border rubber wood trade supply chain as claimed in claim 1, characterized in that: Before processing the price volatility coefficient, the weather influencing factor and the risk-return balance point that affect the rubber wood trading decision to obtain the rubber wood trading decision features, the rubber wood trading decision features are removed and corrected to remove outliers to improve the generalization ability of the model.
8. A transaction decision-making method for a cross-border rubber wood trade supply chain as claimed in claim 6, characterized in that: The price volatility coefficient, the weather influencing factor and the risk-return balance point that will affect the rubber wood trading decision are processed to obtain the rubber wood trading decision features using a feature linking method, and the feature linking method includes feature horizontal splicing and feature vertical splicing.
9. A transaction decision method for a cross-border rubber wood trade supply chain as claimed in claim 1, characterized in that: The use of historical transaction decision methods and corresponding parameters to test the transaction decision method generation error rate of the collaborative filtering transaction decision method generation model; The historical transaction decision method and corresponding parameters include the historical price volatility coefficient, historical weather impact factor and historical risk-return balance point used outside the training model, as well as the historical actual rubber wood transaction decision method; And based on the collaborative filtering transaction decision method generation model with an error rate less than the error value, the historical transaction decision method is generated, and the error value between the historical transaction decision method and the historical actual rubber wood transaction decision method is calculated, and the error value is 0.
15. When those skilled in the art make rubber wood transaction decisions, the transaction decision error rate given by the model cannot exceed a certain value to ensure the monitoring and rational use of the trained model.
10. A transaction decision method for a cross-border rubber wood trade supply chain as claimed in claim 2, 3 or 5, characterized in that: The collaborative filtering transaction decision-making method generation model utilizes a collaborative filtering algorithm based on an improvement of machine learning.