Bulk commodity transaction risk control system based on credit assessment

Abstract

The invention relates to the technical field of transaction risk management, in particular to a bulk commodity transaction risk control system based on credit assessment, which comprises the following steps of: identifying a target order and a target analysis commodity, dividing front and rear end time by combining order creation time and a commodity characteristic transaction period, collecting front end cost and screening strong correlation factors; a back-end fluctuation curve is output through an LSTM neural network, meanwhile, historical fund settlement information of a transaction user is analyzed to determine a user risk coefficient, then a comprehensive risk coefficient is calculated in combination with the back-end fluctuation curve and the user risk coefficient, and finally a pre-estimated deposit value is determined and synchronized to terminals of the two parties. The method achieves the comprehensive quantitative evaluation of user credit, constructs a scientific comprehensive risk evaluation system, enables the pre-estimated earnest money to be precisely matched with the actual risk, and balances the interests of two transaction parties.

Description

Technical Field

[0001] This invention relates to the field of transaction risk management technology, and in particular to a commodity transaction risk control system based on credit assessment. Background Technology

[0002] Commodities refer to homogeneous, tradable goods that are widely used as basic raw materials for industry. They possess both commodity and financial attributes and are essential for economic and social development.

[0003] However, current systems mostly assess market price risk or user credit risk separately, without combining the back-end volatility risk of the product itself with the credit risk of the trading user. This results in unscientific calculation of the comprehensive risk coefficient, which in turn makes the setting of risk hedging measures such as deposits and margins lack data support. They are either too conservative, affecting trading efficiency, or too lenient, unable to cover potential risks. Summary of the Invention

[0004] The purpose of this invention is to solve the problems in the background art by proposing a risk control system for commodity trading based on credit assessment.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A credit-based risk control system for commodity trading includes:

[0007] The transaction information collection module is used to identify target orders and collect transaction information for those orders.

[0008] The product processing module is used to identify target products in transaction information, analyze the product delivery cycle of target products, and determine the characteristic transaction cycle of target products.

[0009] The cost analysis module is used to determine the front-end and back-end times based on the order creation time of the target order and the characteristic transaction cycle of the target analyzed product. It collects the unit cost within the front-end time and plots the front-end cost curve. Then, it collects cost influencing factors, calculates the key similarity value between the cost influencing factors and the front-end cost curve, identifies strongly correlated factors based on the key similarity value, sets the strongly correlated factors as input data, and uses an LSTM neural network model to output the back-end fluctuation curve of the back-end time.

[0010] The user analysis module is used to obtain the trading users in the transaction information and analyze the fund settlement information of the trading users' historical trading orders to determine the user risk coefficient of the trading users.

[0011] The risk assessment module is used to combine the back-end volatility curve of the target product with the user risk coefficient of the transaction user to determine the comprehensive risk coefficient of the target order, and based on the comprehensive risk coefficient, determine the estimated deposit value of the transaction user.

[0012] The terminal display module is used to obtain the estimated deposit value and display it on the terminal devices of both the trading user and the enterprise.

[0013] As a further aspect of the present invention, the trading commodities in the trading system are bulk commodities, and the target orders are preliminary orders, that is, initial orders that have been initially created by the trading user but have not yet been finalized. They are the preliminary state of formally completed orders, representing only the confirmation of trading intentions, and the final binding of funds and ownership of goods has not yet been completed.

[0014] As a further aspect of the present invention, when analyzing the delivery cycle of the target analysis product and determining the characteristic transaction cycle, it is necessary to first identify the transaction type of the target analysis product. The transaction type includes spot products and customized products. Customized products refer to non-standard products made on demand, that is, produced according to the buyer's exclusive requirements such as specifications, materials, and sizes, and the production cycle is relatively long. Spot products refer to products that do not require additional production and conform to industry general standards or national standards with uniform specifications.

[0015] Using the target product as the search keyword, the database of the transaction system is searched to obtain the historical transaction information of the target product. The transaction quantity and product delivery cycle are extracted from the historical transaction information. The product delivery cycle refers to the length of time between the start time of order generation and the end time when the seller fully delivers the product to the buyer.

[0016] As a further aspect of the present invention, the method for determining the characteristic trading cycle of spot commodities includes:

[0017] If the target product is a spot product, then directly obtain the product delivery cycle Ti of different transaction orders in the historical transaction information, where i represents different transaction orders and i∈[1,n], indicating that there are n transaction orders in the historical transaction information. Then arrange the product delivery cycles Ti in ascending order to obtain the delivery sequence.

[0018] The positions of Q1 and Q3 are calculated using the quartile algorithm, where Q1 = k1 × (n + 1) and Q3 = k3 × (n + 1), where k1 is the proportionality coefficient of the lower quartile and k3 is the proportionality coefficient of the upper quartile. The value of k1 is set to 0.25 and the value of k2 is set to 0.75.

[0019] Obtain the delivery sequence, and identify the data Ta and Tb corresponding to positions Q1 and Q3 in the delivery sequence, respectively, where a∈i and b∈i;

[0020] Then use the calculation formula Obtain the delivery time threshold Ty. The value is set to 1.5;

[0021] The product delivery cycle Ti is then compared with the delivery time threshold Ty. The product delivery cycles with Ti ≤ Ty are selected, and the average value of all product delivery cycles with Ti ≤ Ty is calculated. The calculated average value is then marked as the characteristic transaction cycle of the target product.

[0022] As a further embodiment of the present invention, when Q1 is not an integer, the data corresponding to the integer position is identified, and the data corresponding to (integer position + 1) is obtained. Then, the weighted average of the two data is taken, and the obtained data is marked as the representative data Ta of position Q1. If Q3 is also not an integer, the same method is used to obtain the representative data Tb of position Q3.

[0023] As a further aspect of the present invention, the method for determining the characteristic transaction cycle of customized goods includes:

[0024] If the target product is a customized product, obtain the product delivery cycle Ti and the transaction quantity Ci of different transaction orders in the historical transaction information;

[0025] Set the unit production quantity, divide the transaction quantity Ci by the unit production quantity to obtain the batch production coefficient Pi, where the unit production quantity is the threshold.

[0026] Next, obtain the product delivery cycle Ti, and use the formula Ti ÷ Pi = Ri to get the unit output time Ri. Take the unit output time Ri of all transaction orders, and use the average calculation formula. Obtain the unit output representative time Rp of the target product;

[0027] Obtain the transaction quantity in the target order, divide the transaction quantity by the unit production quantity to obtain the batch production coefficient Ps of the target product, and then multiply the batch production coefficient Ps by the unit output representative time Rp. Mark the resulting product as the characteristic transaction cycle of the target product.

[0028] As a further aspect of the present invention, the method for determining the back-end fluctuation curve includes:

[0029] SS1: Obtain the order creation time of the target order. Using the order creation time as the time node, mark the time before the order creation time as the front-end time. Then obtain the characteristic transaction cycle of the target product and mark the time after the order creation time as the back-end time. The length of the back-end time is the characteristic transaction cycle of the target product.

[0030] Set time as the horizontal axis and unit cost as the vertical axis, and set up a two-dimensional coordinate system to obtain the unit cost of the front-end time. Then, set the points in the two-dimensional coordinate system and perform curve fitting on the set points to obtain the front-end cost curve.

[0031] SS2: Obtain cost influencing factors for multiple target products, including exchange rates, policy adjustments, and logistics costs;

[0032] Arbitrarily select a cost influencing factor and collect information data of this cost influencing factor in the previous time period. Align the information data with the previous cost curve according to the time dimension. Then use the correlation analysis algorithm to calculate the correlation similarity value between this cost influencing factor and the previous cost curve. If the correlation similarity value is greater than or equal to the similarity threshold A1, then this cost influencing factor is marked as a strongly correlated factor. Conversely, if the correlation similarity value is less than the similarity threshold A1, then this cost influencing factor is marked as a weakly correlated factor.

[0033] Each cost-influencing factor is classified into strongly correlated factors and weakly correlated factors according to the above method;

[0034] SS3: Obtain all strongly correlated factors, collect information data corresponding to the strongly correlated factors in the current time period, and set this information data as input data. Based on the front-end cost curve, use machine learning models to predict the cost curve of the target product in the back-end time period, and mark the cost curve generated after prediction as the back-end fluctuation curve.

[0035] As a further aspect of the present invention, the method for determining the user risk coefficient includes:

[0036] Identify the transaction user in the target order and collect the fund settlement information of the transaction user's historical transaction orders. The fund settlement information includes the agreed delivery time and the actual delivery time. The agreed delivery time refers to the settlement time of the final payment agreed between the transaction user and the enterprise, and the actual delivery time refers to the time when the transaction user actually settles the final payment with the enterprise.

[0037] Obtain the fund settlement information of the trading user, and mark the agreed delivery time in the fund settlement information as TDj and the actual delivery time as TSj, where j represents different trading orders of the trading user, and j∈[1,J], indicating that the trading user has a total of J trading orders;

[0038] The settlement time difference TCj is obtained by using the formula TDj-TSj=TCj. If TCj≥0, the corresponding transaction order is marked as a credit order; otherwise, if TCj<0, the corresponding transaction order is marked as an overdue order.

[0039] Obtain all overdue orders and their settlement time difference TCq, where q represents a distinct overdue order and q∈[1, v1], and v1 represents the total number of existing overdue orders. Then, use the calculation formula... Obtain the user risk coefficient FY for this transaction user. Represent the base, and >1.

[0040] As a further aspect of the present invention, the method for determining the comprehensive risk coefficient and the estimated deposit value includes:

[0041] Obtain the back-end fluctuation curve of the target product and identify the interval peak data Hm, initial data Hc, and termination data Hz in the back-end fluctuation curve. Here, m represents different peak points, the interval peak data represents the unit cost corresponding to the maximum or minimum value in any interval of the back-end fluctuation curve, the initial point refers to the unit cost corresponding to the first valid data point in the back-end fluctuation curve, and the termination point refers to the unit cost corresponding to the last valid data point in the back-end fluctuation curve.

[0042] Using calculation formula The overall risk coefficient Fz of the target order is obtained, where, M represents the total number of peak data points, and Ha represents the mean of the peak data points Hm. This means that when Hc < Hz, the value of this term is 0. This means that when CV > 1, this item takes the value 1. , and All are proportionality coefficients;

[0043] Obtain the overall transaction price of the target order, multiply the overall risk coefficient Fz of the target order by the overall transaction price, and mark the resulting product as the estimated deposit value.

[0044] Compared with existing technologies, the advantages of this invention are:

[0045] This invention identifies target orders and target analyzed commodities, divides the front-end and back-end timelines by combining order creation time and commodity characteristics with the transaction cycle, collects front-end costs and filters strongly correlated factors, outputs a back-end volatility curve through an LSTM neural network, and simultaneously analyzes the historical fund settlement information of trading users to determine user risk coefficients. Then, it combines the back-end volatility curve and user risk coefficients to calculate a comprehensive risk coefficient, ultimately determining the estimated deposit value and synchronizing it to both parties' terminals. This invention significantly improves the accuracy of back-end volatility prediction, achieves comprehensive quantitative assessment of user credit, constructs a scientific comprehensive risk assessment system, ensures accurate matching between estimated deposits and actual risks, balances the interests of both parties, and forms a closed-loop control mechanism throughout the entire process. It also improves transaction information transparency and collaborative efficiency, effectively reduces losses from defaults and market fluctuations, and ensures the safety and stability of bulk commodity transactions. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0047] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0048] Reference Figure 1 A risk control system for bulk commodity transactions based on credit assessment includes a transaction information collection module, a commodity processing module, a cost analysis module, a user analysis module, a risk assessment module, and a terminal display module.

[0049] The transaction information collection module is used to identify real-time generated transaction orders in the transaction system and mark these transaction orders as target orders. Then, it collects the transaction information of the target orders. At the same time, the transaction information collection module establishes a one-way communication connection with the commodity processing module and the user analysis module respectively, and transmits the transaction information of the target orders to the commodity processing module. In this embodiment, the transaction commodity in the transaction system is a bulk commodity, and the transaction information includes the name of the transaction commodity, the transaction customer, the transaction unit price, the overall transaction price, and the transaction quantity.

[0050] It should be further explained that the target order here is a preliminary order, which is an initial order that the trading user has completed the initial creation of but has not yet reached a final transaction. It is a preliminary state of a formal transaction order, which only represents the confirmation of the transaction intention and has not yet completed the final binding of funds and ownership of the goods.

[0051] The product processing module is used to identify target products in target orders and analyze historical data of these products to determine their characteristic transaction cycles. Specific methods for determining these characteristic transaction cycles include:

[0052] S1: Extract the name of the traded commodity from the transaction information and mark this traded commodity name as the target analysis commodity. First, identify the transaction type of the target analysis commodity. The transaction type includes spot commodities and customized commodities. Customized commodities refer to non-standard products made on demand, that is, produced according to the buyer's exclusive requirements such as specifications, materials, and sizes, and the production cycle is relatively long. Examples include industrial steel of specific grades, chemical raw materials of customized specifications, and metal profiles of exclusive sizes. Spot commodities refer to commodities that do not require additional production and conform to industry general standards or national standards with uniform specifications. Examples include crude oil, gold, and soybeans listed on exchanges, as well as general steel and coal circulating in the market.

[0053] S2: Using the target product as the search keyword, search the data information database of the transaction system to obtain the historical transaction information of the target product, and extract the transaction quantity and product delivery cycle from the historical transaction information. The product delivery cycle refers to the length of time between the start time of the order generation time and the end time of the seller's complete delivery of the product to the buyer.

[0054] Based on the transaction type of the target product, if the target product is a spot product, the product delivery cycle Ti of different transaction orders in the historical transaction information is directly obtained, where i represents different transaction orders and i∈[1,n], indicating that there are n transaction orders in the historical transaction information. The product delivery cycles Ti are then arranged in ascending order to obtain the delivery sequence.

[0055] The positions of Q1 and Q3 are calculated using the quartile algorithm, where Q1 = k1 × (n + 1) and Q3 = k3 × (n + 1), k1 is the proportionality coefficient of the lower quartile and k3 is the proportionality coefficient of the upper quartile. In this embodiment, k1 is set to 0.25 and k2 is set to 0.75.

[0056] Obtain the delivery sequence, and identify the data Ta and Tb corresponding to positions Q1 and Q3 in the delivery sequence, respectively, where a∈i and b∈i;

[0057] It should be further explained that when Q1 is not an integer, the data corresponding to the integer part is identified, and the data corresponding to (integer part + 1) is obtained. Then, the weighted average of the two data is taken, and the obtained data is marked as the representative data Ta of position Q1. If Q3 is also not an integer, it is processed in the same way to obtain the representative data Tb of position Q3.

[0058] Then use the calculation formula The delivery time threshold Ty is obtained, where, Constant coefficients, in this embodiment The value is set to 1.5;

[0059] Then, the product delivery cycle Ti is compared with the delivery time threshold Ty. The product delivery cycles with Ti≤Ty are selected, and the average value of all product delivery cycles with Ti≤Ty is calculated. The calculated average value is marked as the characteristic transaction cycle of the target product.

[0060] If the target product is a customized product, obtain the product delivery cycle Ti and the transaction quantity Ci of different transaction orders in the historical transaction information;

[0061] Set the unit production quantity, divide the transaction quantity Ci by the unit production quantity to obtain the batch production coefficient Pi, where the unit production quantity is a threshold, and the specific value is set by those skilled in the art based on the company's production parameters;

[0062] Next, obtain the product delivery cycle Ti, and use the formula Ti ÷ Pi = Ri to get the unit output time Ri. Take the unit output time Ri of all transaction orders, and use the average calculation formula. Obtain the unit output representative time Rp of the target product;

[0063] Obtain the transaction quantity in the target order, divide the transaction quantity by the unit production quantity to obtain the batch production coefficient Ps of the target product, and then multiply the batch production coefficient Ps by the unit output representative time Rp. Mark the product result as the characteristic transaction cycle of the target product.

[0064] Then, a one-way communication connection is established between the commodity processing module and the cost analysis module, and the characteristic transaction cycle of the target commodity is transmitted to the cost analysis module.

[0065] The cost analysis module is used to analyze the unit cost of the target product and determine its back-end volatility curve. Specific methods for determining the back-end volatility curve include:

[0066] SS1: Obtain the order creation time of the target order. Using the order creation time as the time node, mark the time before the order creation time as the front-end time. Then obtain the characteristic transaction cycle of the target product and mark the time after the order creation time as the back-end time. The length of the back-end time is the characteristic transaction cycle of the target product.

[0067] Set time as the horizontal axis and unit cost as the vertical axis, and set up a two-dimensional coordinate system to obtain the unit cost of the front-end time. Then, set the points in the two-dimensional coordinate system and perform curve fitting on the set points to obtain the front-end cost curve.

[0068] SS2: Obtain cost influencing factors for multiple target products, including exchange rates, policy adjustments, and logistics costs;

[0069] Arbitrarily select a cost influencing factor and collect information data of this cost influencing factor in the previous time period. Align the information data with the previous cost curve according to the time dimension. Then use the correlation analysis algorithm to calculate the correlation similarity value between this cost influencing factor and the previous cost curve. If the correlation similarity value is greater than or equal to the similarity threshold A1, then this cost influencing factor is marked as a strongly correlated factor. Conversely, if the correlation similarity value is less than the similarity threshold A1, then this cost influencing factor is marked as a weakly correlated factor.

[0070] Each cost-influencing factor is classified into strongly correlated factors and weakly correlated factors according to the above method;

[0071] Furthermore, the correlation analysis algorithm used in this embodiment is the Pearson correlation coefficient algorithm, and the specific calculation process of the correlation similarity value using the Pearson correlation coefficient is existing technology, which will not be described in detail here. Meanwhile, the specific value of the similarity threshold A1 is obtained by those skilled in the art based on big data calculation.

[0072] SS3: Obtain all strongly correlated factors, collect information data corresponding to the strongly correlated factors in the current time, and set this information data as input data. Based on the front-end cost curve, use machine learning models to predict the cost curve of the target product in the back-end time. Mark the cost curve generated after prediction as the back-end fluctuation curve.

[0073] Furthermore, the machine learning model used in this embodiment is the LSTM neural network model, which is an existing technology and will not be described in detail here.

[0074] Then, a one-way communication connection is established between the cost analysis module and the risk assessment module, and the back-end fluctuation curve is transmitted to the risk assessment module.

[0075] The user analysis module is used to obtain historical transaction information of trading users and to conduct user risk assessment based on this information. Specific methods for user risk assessment include:

[0076] Identify the transaction user in the target order and collect the fund settlement information of the transaction user's historical transaction orders. The fund settlement information includes the agreed delivery time and the actual delivery time. The agreed delivery time refers to the settlement time of the final payment agreed between the transaction user and the enterprise, and the actual delivery time refers to the time when the transaction user actually settles the final payment with the enterprise.

[0077] Obtain the fund settlement information of the trading user, and mark the agreed delivery time in the fund settlement information as TDj and the actual delivery time as TSj, where j represents different trading orders of the trading user, and j∈[1,J], indicating that the trading user has a total of J trading orders;

[0078] It should be further explained that when a company delays the delivery time of goods, the transaction customer needs to extend the agreed delivery time based on the company's delayed delivery time, and then set the final agreed delivery time as the agreed delivery time of the corresponding transaction order.

[0079] The settlement time difference TCj is obtained by using the formula TDj-TSj=TCj. If TCj≥0, the corresponding transaction order is marked as a credit order; otherwise, if TCj<0, the corresponding transaction order is marked as an overdue order.

[0080] Obtain all overdue orders and their settlement time difference TCq, where q represents a distinct overdue order and q∈[1, v1], and v1 represents the total number of existing overdue orders. Then, use the calculation formula... Obtain the user risk coefficient FY for this transaction user. Represent the base, and >1, The specific values ​​were obtained by those skilled in the art based on big data calculations.

[0081] Then, a one-way communication connection is established between the user analysis module and the risk assessment module, and the user risk coefficient of the trading user is transmitted to the risk assessment module;

[0082] The risk assessment module combines the back-end volatility curve of the target product with the user risk coefficient of the trading user to determine the comprehensive risk coefficient of the target order. The specific methods for determining the comprehensive risk coefficient include:

[0083] Obtain the back-end fluctuation curve of the target product and identify the interval peak data Hm, initial data Hc, and termination data Hz in the back-end fluctuation curve, where m represents different peak points;

[0084] Furthermore, the interval peak data represents the unit cost corresponding to the maximum or minimum value within any interval of the back-end fluctuation curve. The initial point refers to the unit cost corresponding to the first valid data point in the back-end fluctuation curve, and the termination point refers to the unit cost corresponding to the last valid data point in the back-end fluctuation curve.

[0085] Using calculation formula The overall risk coefficient Fz of the target order is obtained, where, M represents the total number of peak data points, and Ha represents the mean of the peak data points Hm. This means that when Hc < Hz, the value of this term is 0. This means that when CV > 1, this item takes the value 1. , and All are proportionality coefficients, and , and The specific value was obtained by those skilled in the art based on big data calculations;

[0086] The overall transaction price of the target order is obtained, the comprehensive risk coefficient Fz of the target order is multiplied by the overall transaction price, and the resulting product is marked as the estimated deposit value. Then the risk assessment module transmits the estimated deposit value to the terminal display module.

[0087] Furthermore, the larger the overall risk coefficient Fz of the target order, the greater the transaction risk for the corresponding trading user, and the greater the estimated deposit value that the trading user needs to pay for the target order.

[0088] The terminal display module is used to obtain the estimated deposit value and display it on the terminal devices of both the transaction user and the enterprise. The transaction user then pays the corresponding transaction deposit based on the displayed estimated deposit value.

[0089] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A system for controlling risk in commodity trading based on credit assessment, characterized in that, include: The transaction information collection module is used to identify target orders and collect transaction information for those orders. The product processing module is used to identify target products in transaction information, analyze the product delivery cycle of target products, and determine the characteristic transaction cycle of target products. The cost analysis module is used to determine the front-end and back-end times based on the order creation time of the target order and the characteristic transaction cycle of the target analyzed product. It collects the unit cost within the front-end time and plots the front-end cost curve. Then, it collects cost influencing factors, calculates the key similarity value between the cost influencing factors and the front-end cost curve, identifies strongly correlated factors based on the key similarity value, sets the strongly correlated factors as input data, and uses an LSTM neural network model to output the back-end fluctuation curve of the back-end time. The user analysis module is used to obtain the trading users in the transaction information and analyze the fund settlement information of the trading users' historical trading orders to determine the user risk coefficient of the trading users. The risk assessment module is used to combine the back-end volatility curve of the target product with the user risk coefficient of the transaction user to determine the comprehensive risk coefficient of the target order, and based on the comprehensive risk coefficient, determine the estimated deposit value of the transaction user. The terminal display module is used to obtain the estimated deposit value and display it on the terminal devices of both the trading user and the enterprise.

2. The credit assessment based risk control system for commodity trading according to claim 1, wherein, The trading products in the trading system are bulk commodities. The target orders are preliminary orders, which are initial orders that have been initially created by the trading user but have not yet been finalized. They are the preliminary state of formally executed orders, representing only the confirmation of trading intentions, and the final binding of funds and ownership of goods has not yet been completed.

3. The credit assessment based risk control system for commodity trading according to claim 1, wherein, When analyzing the delivery cycle of the target product and determining the characteristic transaction cycle, it is necessary to first identify the transaction type of the target product. The transaction type includes spot products and customized products. Customized products refer to non-standard products made on demand, that is, produced according to the buyer's exclusive requirements such as specifications, materials, and sizes, and the production cycle is relatively long. Spot products refer to products that do not require additional production and conform to industry general standards or national standards with uniform specifications. Using the target product as the search keyword, the database of the transaction system is searched to obtain the historical transaction information of the target product. The transaction quantity and product delivery cycle are extracted from the historical transaction information. The product delivery cycle refers to the length of time between the start time of order generation and the end time when the seller fully delivers the product to the buyer.

4. The commodity trading risk control system based on credit assessment according to claim 3, characterized in that, Methods for determining the characteristic trading cycle of spot commodities include: If the target product is a spot product, then directly obtain the product delivery cycle Ti of different transaction orders in the historical transaction information, where i represents different transaction orders and i∈[1,n], indicating that there are n transaction orders in the historical transaction information. Then arrange the product delivery cycles Ti in ascending order to obtain the delivery sequence. The positions of Q1 and Q3 are calculated using the quartile algorithm, where Q1 = k1 × (n + 1) and Q3 = k3 × (n + 1), where k1 is the proportionality coefficient of the lower quartile and k3 is the proportionality coefficient of the upper quartile. The value of k1 is set to 0.25 and the value of k2 is set to 0.

75. Obtain the delivery sequence, and identify the data Ta and Tb corresponding to positions Q1 and Q3 in the delivery sequence, respectively, where a∈i and b∈i; Then use the calculation formula Obtain the delivery time threshold Ty. The value is set to 1.5; The product delivery cycle Ti is then compared with the delivery time threshold Ty. The product delivery cycles with Ti ≤ Ty are selected, and the average value of all product delivery cycles with Ti ≤ Ty is calculated. The calculated average value is then marked as the characteristic transaction cycle of the target product.

5. The commodity trading risk control system based on credit assessment according to claim 4, characterized in that, When Q1 is not an integer, the data corresponding to the integer position is identified, and the data corresponding to (integer position + 1) is obtained. The weighted average of these two data is then taken, and the resulting data is marked as the representative data Ta of position Q1. If Q3 is also not an integer, the same method is used to obtain the representative data Tb of position Q3.

6. The commodity trading risk control system based on credit assessment according to claim 3, characterized in that, Methods for determining the characteristic transaction cycle of customized products include: If the target product is a customized product, obtain the product delivery cycle Ti and the transaction quantity Ci of different transaction orders in the historical transaction information; Set the unit production quantity, divide the transaction quantity Ci by the unit production quantity to obtain the batch production coefficient Pi, where the unit production quantity is the threshold. Next, obtain the product delivery cycle Ti, and use the formula Ti ÷ Pi = Ri to get the unit output time Ri. Take the unit output time Ri of all transaction orders, and use the average calculation formula. Obtain the unit output representative time Rp of the target product; Obtain the transaction quantity in the target order, divide the transaction quantity by the unit production quantity to obtain the batch production coefficient Ps of the target product, and then multiply the batch production coefficient Ps by the unit output representative time Rp. Mark the resulting product as the characteristic transaction cycle of the target product.

7. The commodity trading risk control system based on credit assessment according to claim 1, characterized in that, Methods for determining the back-end fluctuation curve include: SS1: Obtain the order creation time of the target order. Using the order creation time as the time node, mark the time before the order creation time as the front-end time. Then obtain the characteristic transaction cycle of the target product and mark the time after the order creation time as the back-end time. The length of the back-end time is the characteristic transaction cycle of the target product. Set time as the horizontal axis and unit cost as the vertical axis, and set up a two-dimensional coordinate system to obtain the unit cost of the front-end time. Then, set the points in the two-dimensional coordinate system and perform curve fitting on the set points to obtain the front-end cost curve. SS2: Obtain cost influencing factors for multiple target products, including exchange rates, policy adjustments, and logistics costs; Arbitrarily select a cost influencing factor and collect information data of this cost influencing factor in the previous time period. Align the information data with the previous cost curve according to the time dimension. Then use the correlation analysis algorithm to calculate the correlation similarity value between this cost influencing factor and the previous cost curve. If the correlation similarity value is greater than or equal to the similarity threshold A1, then this cost influencing factor is marked as a strongly correlated factor. Conversely, if the correlation similarity value is less than the similarity threshold A1, then this cost influencing factor is marked as a weakly correlated factor. Each cost-influencing factor is classified into strongly correlated factors and weakly correlated factors according to the above method; SS3: Obtain all strongly correlated factors, collect information data corresponding to the strongly correlated factors in the current time period, and set this information data as input data. Based on the front-end cost curve, use machine learning models to predict the cost curve of the target product in the back-end time period, and mark the cost curve generated after prediction as the back-end fluctuation curve.

8. The commodity trading risk control system based on credit assessment according to claim 1, characterized in that, Methods for determining user risk coefficients include: Identify the transaction user in the target order and collect the fund settlement information of the transaction user's historical transaction orders. The fund settlement information includes the agreed delivery time and the actual delivery time. The agreed delivery time refers to the settlement time of the final payment agreed between the transaction user and the enterprise, and the actual delivery time refers to the time when the transaction user actually settles the final payment with the enterprise. Obtain the fund settlement information of the trading user, and mark the agreed delivery time in the fund settlement information as TDj and the actual delivery time as TSj, where j represents different trading orders of the trading user, and j∈[1,J], indicating that the trading user has a total of J trading orders; The settlement time difference TCj is obtained by using the formula TDj-TSj=TCj. If TCj≥0, the corresponding transaction order is marked as a credit order; otherwise, if TCj<0, the corresponding transaction order is marked as an overdue order. Obtain all overdue orders and their settlement time difference TCq, where q represents a distinct overdue order and q∈[1, v1], and v1 represents the total number of existing overdue orders. Then, use the calculation formula... Obtain the user risk coefficient FY for this transaction user. Represent the base, and >

1.

9. The commodity trading risk control system based on credit assessment according to claim 1, characterized in that, The methods for determining the comprehensive risk factor and the estimated deposit value include: Obtain the back-end fluctuation curve of the target product and identify the interval peak data Hm, initial data Hc, and termination data Hz in the back-end fluctuation curve. Here, m represents different peak points, the interval peak data represents the unit cost corresponding to the maximum or minimum value in any interval of the back-end fluctuation curve, the initial point refers to the unit cost corresponding to the first valid data point in the back-end fluctuation curve, and the termination point refers to the unit cost corresponding to the last valid data point in the back-end fluctuation curve. Using calculation formula The overall risk coefficient Fz of the target order is obtained, where, M represents the total number of peak data points, and Ha represents the mean of the peak data points Hm. This means that when Hc < Hz, the value of this term is 0. This means that when CV > 1, this item takes the value 1. , and All are proportionality coefficients; Obtain the overall transaction price of the target order, multiply the overall risk coefficient Fz of the target order by the overall transaction price, and mark the resulting product as the estimated deposit value.

Images