Intelligent bidding decision-making method and system based on large model

Abstract

The invention discloses an intelligent bidding decision-making method based on a large model, and the method comprises the steps: obtaining historical purchase data, project progress and market information data, carrying out the large model analysis, predicting the material demands of a target time period, and outputting a purchase plan; based on the purchase plan, screening suppliers meeting requirements from a supplier library and sending out an electronic bidding document; receiving offers of the suppliers, performing analysis and comparison on the basis of a generative adversarial network in combination with the market quotation data and the historical purchase data, and screening an optimal supplier; and obtaining a subsequent material acceptance result, and completing delivery based on the acceptance result. Through the above mode, the decision-making process in the aspect of material purchasing is optimized and perfected, the defects of timeliness negative effect, dependence on related personnel experience and the like caused by rapid market change are effectively avoided, and the decision-making efficiency and accuracy are improved.

Description

Technical Field

[0001] The present application relates to the field of big data analysis, and in particular to an intelligent bidding decision-making method and system based on a large model. Background Art

[0002] For businesses, material procurement is a crucial daily task, often requiring price inquiries and comparisons from multiple suppliers before ultimately selecting the right option. However, the actual procurement process presents numerous challenges, including irregular procurement processes, opaque information, cumbersome approval processes, a lack of standardized procedures, price fluctuations due to market conditions, and the mismatch between price and quality. These issues lead to additional procurement and management costs. Furthermore, price negotiations based on market conditions and product prices incur further time and management costs, making the procurement process complex, inefficient, and making it difficult to reach optimal decisions.

[0003] Existing intelligent bidding technology usually automatically quotes based on the demand and rules of each product. The quotation determination stage is often limited only by a preset reasonable price range. This leads to a focus on the price itself without considering multiple factors for judgment, making the final purchasing result not the optimal solution. Summary of the Invention

[0004] In view of this, this application mainly provides an intelligent bidding decision-making method and system based on a large model to solve the above problems.

[0005] To solve the above technical problems, this application adopts a technical solution: to provide an intelligent bidding decision-making method based on a large model, including: S10: Obtain historical procurement data, project progress, and market data, perform large-scale model analysis, and predict material demand during the target period, and output a procurement plan. S20: Based on the procurement plan, select suppliers that meet the requirements from the supplier database and issue electronic bids; S30: receiving quotations from the suppliers, analyzing and comparing them based on a generative adversarial network and in combination with the market data and the historical procurement data, and selecting the best supplier; S40: Obtain subsequent material acceptance results and complete delivery based on the acceptance results.

[0006] In one possible implementation, the step of obtaining historical procurement data, project progress, and market data includes: S11: Preprocess the data, including standardization and feature extraction.

[0007] In one possible implementation, the steps of performing a large-scale model analysis and forecasting material demand during a target period and outputting a procurement plan include: S12: Perform secondary development and lightweight deployment on the large model; S13: Analyze the historical procurement data, the project progress, and the market data, divide the material demand based on time, and output a procurement plan based on the target time period.

[0008] In a possible implementation, the step of performing secondary development and lightweight deployment on the large model includes: S121: Inject procurement rules and terminology to build a high-quality annotated dataset and update model parameters through a low-rank matrix adapter; S122: Encapsulate the large model and connect it to the procurement system through the API interface.

[0009] In one possible implementation, the step of performing a large-scale model analysis and forecasting material demand for a target period, and outputting a procurement plan, further includes: S14: Perform multi-dimensional cost accounting based on the material demand and the market data, and adjust the procurement plan based on the result of the multi-dimensional cost accounting.

[0010] In a possible implementation, the step of adjusting the procurement plan based on the multi-dimensional cost accounting result includes: S15: When the market data shows an upward trend, the purchase plan outputs a long-term contract; when the market data shows a downward trend, the purchase plan outputs a short-term contract.

[0011] In a possible implementation, the step of screening suppliers that meet the requirements from the supplier database based on the procurement plan and issuing electronic bids includes: S21: Profiling all suppliers in the supplier database; S22: Score the suppliers based on the portraits, select suppliers with top scores and send electronic bids.

[0012] In a possible implementation, the step of creating a profile for all suppliers in the supplier database further includes: S211: The supplier profile includes a risk assessment of the supplier, which includes supply risk, price risk, and quality risk. When the risk assessment exceeds a preset range, a risk warning is issued and corresponding countermeasures are initiated.

[0013] In one possible implementation, the step of receiving quotations from suppliers, analyzing and comparing them based on a generative adversarial network in combination with the market data and the historical procurement data, and selecting the optimal supplier further includes: S31: If the detection finds that the quotations of more than a threshold number of suppliers deviate from the market data by a preset value, the screening and decision-making are terminated and the antitrust review process is triggered.

[0014] To solve the above technical problems, another technical solution adopted by this application is to provide an intelligent bidding decision system based on a large model, which is applicable to the intelligent bidding decision method based on a large model as described above, including: Interactive module, used to input data and output procurement plans, screening results and analysis results; Large model module for data analysis, comparison and supplier screening; Inspection module, used to process acceptance results and delivery process.

[0015] The beneficial effect of this application is: different from the existing technology, this application discloses an intelligent bidding decision-making method and system based on a large model. By accessing the large model and realizing intelligent and automated analysis and decision-making based on multi-dimensional data, it optimizes and improves the decision-making process in material procurement, effectively avoiding the negative time effects brought about by rapid market changes, dependence on the experience of relevant personnel and other defects, and improving decision-making efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. Those skilled in the art can also derive other drawings based on these drawings without inventive work, among which: Figure 1 This is a flow chart of an intelligent bidding decision-making method based on a large model in one embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0018] The terms "first", "second" and "third" in the embodiments of the present application are only used for descriptive purposes and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second" and "third" may explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but may optionally also include steps or units that are not listed, or may optionally also include other steps or units inherent to these processes, methods, products or devices.

[0019] References to "embodiments" herein mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0020] See also Figure 1 , the embodiment of the present application includes: an intelligent bidding decision method based on a large model, including: S10: Obtain historical procurement data, project progress, and market data, perform large-scale model analysis, and predict material demand during the target period, and output a procurement plan. S20: Based on the procurement plan, select suppliers that meet the requirements from the supplier database and issue electronic bids; S30: Receive quotations from suppliers, analyze and compare them based on a generative adversarial network and combined with market data and historical procurement data to select the best supplier; S40: Obtain subsequent material acceptance results and complete delivery based on the acceptance results.

[0021] Specifically, in step S10, historical procurement data includes the purchase quantity of the materials to be purchased over a period of time, with the time span covering at least one year and multiple projects to ensure the reliability of subsequent data analysis. Project progress can be divided into several phases based on the project's construction progress, or by time or quarter. Market data refers to historical price fluctuations of the materials to be purchased.

[0022] Subsequent analysis involves inputting the aforementioned data into a locally deployed large model, which has preset rules for processing the data. This includes comparing historical procurement data with market data based on project progress, identifying a reasonable price fluctuation range and the procurement volume required for each project stage, and specifying a procurement plan based on this.

[0023] Large-scale model prediction requires building a predictive model. Using time series models and machine learning, it automatically generates forecasts based on input data (historical procurement data, project progress, and market data), integrating a range of material forecasts. Training data can be obtained through project management systems, supply chain management systems, and more specifically, can be directly connected to ERP systems for data sharing.

[0024] In step S20, based on the material type, price and usage in the procurement plan, a supplier that meets the requirements is selected from the supplier database, and an electronic bid document is issued for bidding.

[0025] In step S30, after receiving supplier quotations, they are analyzed and compared using a generative adversarial network (GAN) along with market data and historical procurement data. The GAN generates virtual supplier proposals and compares them against existing suppliers, thereby identifying suppliers and determining the authenticity of their quotations based on the loss function. Furthermore, the GAN evaluates each supplier's loss function results and uses this as a benchmark for supplier screening. If the loss function of all suppliers in this screening round exceeds the threshold, an anomaly is considered, the screening process is terminated, and a warning is triggered. If necessary, the procurement plan and related parameters can be adjusted and the bidding process can be renewed.

[0026] In step S40, the subsequent materials are inspected and accepted, and evaluated from aspects including but not limited to the speed of arrival of the materials, the quality of the materials, and the efficiency of cooperation. The delivery is completed according to the evaluation until the entire procurement project is completed.

[0027] The entire process includes the complete process from designated planning, bidding and evaluation to acceptance, and realizes automated workflow through various API interfaces, which greatly improves the integrity of the process and enhances the efficiency and accuracy of decision-making. In one embodiment, the step of obtaining historical procurement data, project progress, and market data includes: S11: Preprocess the data, including standardization and feature extraction.

[0028] Specifically, data is preprocessed to facilitate subsequent analysis. Standardization includes processing heterogeneous data (text, time series, transaction records, price curves, etc.) according to corresponding formats (such as aligned time, unified units). Extracting key features includes capturing the cyclical characteristics of historical procurement data, the stage characteristics of project progress, and the cyclical characteristics of price trends in market data. Cyclic characteristics can be seasonal, project stage, or based on year-on-year or month-on-month growth rates.

[0029] In one embodiment, the steps of performing a large-scale model analysis and forecasting material demand during a target period and outputting a procurement plan include: S12: Secondary development and lightweight deployment of large models; S13: Analyze historical procurement data, project progress, and market data, divide material demand based on time, and output procurement plans based on target time periods.

[0030] Specifically, after the prediction model is built and used to output predicted values ​​in the previous steps, the large model undergoes secondary development and lightweight deployment to improve its usability and processing efficiency. This secondary development involves infusing domain knowledge to build a high-quality annotation set. Domain knowledge includes enterprise procurement rules, bidding processes, supply chain terminology, and more. Furthermore, the large model is fine-tuned by using a low-rank matrix adapter to update model parameters to reduce video memory usage. A toolchain can also be used to fine-tune the model. Lightweight deployment involves converting the native FP8 model to INT8 mode using block-wise or channel-wise quantization to reduce video memory requirements and increase inference throughput. The model is then processed using the SGlang framework for asynchronous execution to accelerate execution. Alternatively, a distilled version of the large model, such as DeepSeek-R1-Distill-Qwen-7B, can be used, which has a smaller footprint and hardware requirements, enabling lightweight deployment.

[0031] In one embodiment, the steps of secondary development and lightweight deployment of a large model include: S121: Inject procurement rules and terminology to build a high-quality annotated dataset and update model parameters through a low-rank matrix adapter; S122: Encapsulate the large model and connect it to the procurement system through the API interface.

[0032] Injecting procurement rules and belonging sets in step S121 is a further explanation of secondary development. After secondary development is completed, step S122 is performed to encapsulate the interface of the large model into a RESTful API, and connect it with the relevant systems of the enterprise to realize direct collection and analysis of data.

[0033] In one embodiment, the steps of performing a large-scale model analysis and forecasting material demand during a target period and outputting a procurement plan further include: S14: Conduct multi-dimensional cost accounting based on material demand and market data, and adjust procurement plans based on the results of multi-dimensional cost accounting.

[0034] Specifically, in addition to the purchase cost of the materials themselves, additional costs need to be considered, such as transportation costs, warehousing costs, and capital liquidity costs. The multi-dimensional costs can be calculated using the following formula: S+F+R Where P refers to multidimensional cost, m refers to material demand, u refers to the market benchmark unit price of materials, L refers to transportation cost, S refers to warehousing cost, F refers to capital liquidity cost, and R refers to risk premium. The risk ratio of the risk premium can be obtained based on the product of material demand and material benchmark price multiplied by the risk ratio.

[0035] In one embodiment, the step of adjusting the procurement plan based on the multi-dimensional cost accounting results includes: S15: When the market data shows an upward trend, the procurement plan outputs a long-term contract; when the market data shows a downward trend, the procurement plan outputs a short-term contract.

[0036] Market trends can be determined by directly monitoring and comparing the market data itself, or by analyzing overall changes in multi-dimensional costs. When the market is trending upward, it indicates that the current price is likely to remain low for a period of time, so it is necessary to prioritize long-term contracts to save money. Conversely, if the market is trending downward, it indicates that the current price is likely to remain high for a period of time, so it is necessary to prioritize short-term contracts to reduce overall expenses and achieve cost optimization.

[0037] Furthermore, after outputting a long-term or short-term contract, the supplier may not fully respond. Therefore, the supplier's response and time factors can be taken into consideration. When a supplier refuses the corresponding contract, subsequent suppliers should be invited and negotiated according to their priority, and the time weight should be increased during these negotiations. If the market data changes further during the negotiation period and exceeds the threshold, the choice of continuing negotiations, screening suppliers, or recalculating costs should be made based on the scope of the change.

[0038] In one embodiment, based on the procurement plan, the steps of screening suppliers that meet the requirements from the supplier database and issuing electronic bids include: S21: Profiling all suppliers in the supplier database; S22: Score suppliers based on the profile, select suppliers with the highest scores and send electronic bids.

[0039] Specifically, suppliers in the supplier database should be profiled in advance, with clear indicators for each supplier extracted, aggregated, and then scored uniformly based on supplier scoring criteria. This scoring criteria can include multiple criteria to address different types of substances and projects, and will be related to the company's rules for supplier inventory management.

[0040] In this embodiment, the supplier profile includes indicators such as the past material quality qualification rate, delivery timeliness rate, after-sales service quality, and quotation deviation rate. Suppliers are comprehensively scored based on these indicators. Furthermore, this score should be updated regularly to ensure its timeliness and indirectly improve the level of supplier management.

[0041] In one embodiment, the step of generating a profile for all suppliers in the supplier database further includes: S211: The supplier profile includes a risk assessment of the supplier, which includes supply risk, price risk, and quality risk. When the risk assessment exceeds the preset range, a risk warning will be issued and corresponding countermeasures will be initiated.

[0042] Supplier profiling should also include risk assessment as an additional consideration. This includes assessing risks such as timely delivery, price stability, and quality compliance. After analyzing these trends, a comprehensive supplier risk assessment should be conducted. If the risk assessment exceeds the preset range, it indicates that the supplier's risks are uncontrollable and the probability of failing to deliver on time and with quality is too high. In this case, an early warning should be issued to alert personnel to take appropriate countermeasures or remedial measures.

[0043] In one embodiment, the step of receiving quotations from suppliers, analyzing and comparing them based on a generative adversarial network in combination with market data and historical procurement data, and selecting the best supplier further includes: S31: If the detection finds that the quotations of more than a threshold number of suppliers deviate from the market data by a preset value, the screening and decision-making will be terminated and the antitrust review process will be triggered.

[0044] Specifically, in this embodiment, if more than five suppliers are found to have bids that exceed market data by a preset value (15%), the rules will identify suspected monopoly activity and trigger a review process to ensure compliance and proper operation of the procurement process. Furthermore, in subsequent steps, if collusion is identified, the supplier's score will be lowered and recorded.

[0045] A large-model-based intelligent bidding decision system, applicable to the large-model-based intelligent bidding decision method described above, includes: Interactive module, used to input data and output procurement plans, screening results and analysis results; Large model module for data analysis, comparison and supplier screening; Inspection module, used to process acceptance results and delivery process.

[0046] The above description is merely an embodiment of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An intelligent bidding decision-making method based on a large model, characterized in that: include: S10: Obtain historical procurement data, project progress, and market data, perform large-scale model analysis, and predict material demand during the target period, and output a procurement plan. S20: Based on the procurement plan, select suppliers that meet the requirements from the supplier database and issue electronic bids; S30: receiving quotations from the suppliers, analyzing and comparing them based on a generative adversarial network and in combination with the market data and the historical procurement data, and selecting the best supplier; S40: Obtain subsequent material acceptance results and complete delivery based on the acceptance results.

2. The intelligent bidding decision-making method based on a large model according to claim 1 is characterized in that: The steps of obtaining historical procurement data, project progress and market data include: S11: Preprocess the data, including standardization and feature extraction.

3. The intelligent bidding decision-making method based on a large model according to claim 1 is characterized in that: The steps of performing large-scale model analysis and forecasting material demand during the target period and outputting a procurement plan include: S12: Perform secondary development and lightweight deployment on the large model; S13: Analyze the historical procurement data, the project progress, and the market data, divide the material demand based on time, and output a procurement plan based on the target time period.

4. The intelligent bidding decision-making method based on a large model according to claim 3 is characterized in that: The steps of performing secondary development and lightweight deployment on the large model include: S121: Inject procurement rules and terminology to build a high-quality annotated dataset and update model parameters through a low-rank matrix adapter; S122: Encapsulate the large model and connect it to the procurement system through the API interface.

5. The intelligent bidding decision-making method based on a large model according to claim 1 is characterized in that: The steps of performing large-scale model analysis and forecasting material demand during the target period and outputting a procurement plan also include: S14: Perform multi-dimensional cost accounting based on the material demand and the market data, and adjust the procurement plan based on the result of the multi-dimensional cost accounting.

6. The intelligent bidding decision-making method based on a large model according to claim 5 is characterized in that: The step of adjusting the procurement plan based on the multi-dimensional cost accounting result includes: S15: When the market data shows an upward trend, the purchase plan outputs a long-term contract; when the market data shows a downward trend, the purchase plan outputs a short-term contract.

7. The intelligent bidding decision-making method based on a large model according to claim 1 is characterized in that: The step of screening suppliers that meet the requirements from the supplier database based on the procurement plan and issuing electronic bids includes: S21: Profiling all suppliers in the supplier database; S22: Score the suppliers based on the portraits, select suppliers with top scores and send electronic bids.

8. The intelligent bidding decision-making method based on a large model according to claim 7 is characterized in that: The step of profiling all suppliers in the supplier database further includes: S211: The supplier profile includes a risk assessment of the supplier, which includes supply risk, price risk, and quality risk. When the risk assessment exceeds a preset range, a risk warning is issued and corresponding countermeasures are initiated.

9. The intelligent bidding decision-making method based on a large model according to claim 1 is characterized in that: The step of receiving the supplier's quotation, analyzing and comparing it based on a generative adversarial network and in combination with the market data and the historical procurement data to select the best supplier also includes: S31: If the detection finds that the quotations of more than a threshold number of suppliers deviate from the market data by a preset value, the screening and decision-making are terminated and the antitrust review process is triggered.

10. An intelligent bidding decision system based on a large model, applicable to the intelligent bidding decision method based on a large model according to any one of claims 1 to 9, characterized in that: include: Interactive module, used to input data and output procurement plans, screening results and analysis results; Large model module for data analysis, comparison and supplier screening; Inspection module, used to process acceptance results and delivery process.

Images