Internet-based intelligent bid invitation management platform

Abstract

The invention discloses an internet-based intelligent bid invitation management platform, which comprises a data integration and analysis module configured to establish an intelligent comprehensive economic analysis model by integrating internal data and / or external data of the intelligent bid invitation platform; the standardized bid invitation file module is configured to construct an industry lexical analysis and language model by utilizing natural language processing and a knowledge graph; the intelligent question-and-answer type bidding file filling module is configured to complete compiling of a bidding file by guiding a bidder to adopt an intelligent question-and-answer mode; the intelligent bid evaluation tool set module is configured to assist the bid evaluation process by establishing an intelligent bid evaluation tool; the risk early warning module is configured to construct a knowledge graph of a supplier, a subject matter and a project; and the large model application module is configured to construct an AI bid evaluation model by using a large model technology. Through intelligent bid invitation management, bid evaluation work efficiency is improved, and fairness and accuracy of bid evaluation results are ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence application technology, and in particular to an Internet-based intelligent bidding management platform. Background Art

[0002] In the traditional bidding process, it is necessary to go through multiple links such as issuing bidding documents, evaluating qualifications, negotiating prices, and signing contracts. These links not only consume a lot of time and human resources, but also lack effective compliance supervision in offline bidding, such as the risks that may be brought about by manual operations, and are prone to induce dishonest behaviors such as bid rigging and collusion. Therefore, the complexity of the bidding process leads to information asymmetry and lack of transparency, and there is a lack of effective technical means to reduce these risks.

[0003] On the other hand, during the bid evaluation process, the judges need to spend a lot of time looking for the relevant information of the bidders, and some judges have relative discretion, which makes it easy to make mistakes and miss evaluations. As the bidding documents increase, the difficulty of bid evaluation increases, forming a vicious circle. The judges need to spend more time looking for the relevant information of the bidders, and due to insufficient management of suppliers, it also causes inefficiency.

[0004] In view of the above problems, it is necessary to provide an Internet-based intelligent bidding management platform to realize intelligent management of the bidding process. Summary of the invention

[0005] The purpose of the present invention is to provide an Internet-based intelligent bidding management platform to at least solve one or more of the technical problems mentioned above.

[0006] The purpose of the present invention can be achieved through the following technical solutions: In a first aspect, the present invention provides an Internet-based intelligent bidding management platform, comprising: a data integration and analysis module: configured to establish an intelligent comprehensive economic analysis model by integrating internal data and / or external data of the intelligent bidding platform; and to conduct a comprehensive analysis of the market, technological development and historical prices through the intelligent comprehensive economic analysis model to formulate a bidding and procurement plan; wherein the internal data includes historical bidding data, procurement contracts, and project settlement records, and the external data includes price trends, industry reports, and market analysis based on the Internet; Standardized bidding document module: It is configured to use natural language processing and knowledge graph to build industry lexical analysis and language models to achieve standardized and intelligent editing of bidding documents; Intelligent question-and-answer bidding document filling module: configured to guide bidders to complete the writing of bidding documents by using intelligent question-and-answer methods, and automatically generate the final bidding documents through online checking tools; Intelligent bid evaluation tool set module: configured to assist the bid evaluation process by establishing an intelligent bid evaluation tool; the intelligent bid evaluation tool includes an intelligent pre-scoring model and / or an objective clause verification model; Risk warning module: configured to achieve intelligent analysis and risk warning by building a knowledge graph of suppliers, subject matter, and construction projects; Big model application module: It is configured to use big model technology to build an AI bid evaluation model to achieve a dual-mode review mode of unmanned intelligent review and human-machine assisted review.

[0007] In some embodiments, the intelligent comprehensive economic analysis model includes: Data standardization processing unit: configured to process data extracted from multiple heterogeneous data sources into a unified format; including dimensionless processing of the data using a maximum-minimum normalization method to map the data into the range of [0,1], the formula is as follows: , Among them, max(x) is the maximum number in the data, and min(x) is the minimum number in the data; is the weighting coefficient; Market competition model auxiliary decision-making unit: It is configured to measure the market segment competition through the market concentration model, using the concentration index CRn and HHI. The formula is as follows: , , in, represents the winning bid amount of the i-th supplier in the market segment, and x represents the total bidding amount in the market segment; Supplier group model auxiliary decision-making unit: configured to build a supplier group classification model, including classifying suppliers into leverage type, strategic type, general type and bottleneck type according to supply risk factors and procurement amount factors; the model construction formula is as follows: Supplier Type = f(Procurement Amount,Supply Risk), Among them, f is the classification function, which is classified according to the Kalajac model; Purchase price model auxiliary decision-making unit: configured to include a linkage model between purchase price and raw material price, the formula is as follows: , Among them, P is the purchase order settlement unit price, P1 is the winning bid price of the framework agreement bidding, is the price of raw materials on the purchase order signing date, is the raw material price on the bid opening date, M is the raw material content percentage; and The linear regression model is used to construct the purchase reference price model. The formula is as follows: , Where Y is the reference price, is the impact factor, is the weight, , , Other cost factors, , , is the corresponding quantity.

[0008] In some embodiments, the intelligent comprehensive economic analysis model further includes: Data integration unit: configured to clean the raw data of internal data and / or external data using data cleaning tools and algorithms, including noise, errors, and missing data; Data trend processing unit: It is configured to convert the reverse indicator and the moderate indicator into a positive indicator to maintain the unified trend of the indicators. The formula is as follows: , Principal component analysis unit: The configuration includes correlation analysis, which includes calculating the correlation coefficient between variables, and the formula is as follows: , in, Representation variables The correlation coefficient between , whose value is between -1 and 1; Representation variables The covariance between , to measure the extent to which two variables vary together; For variables The variance of , to measure the dispersion of the variable; For variables The variance of , to measure the dispersion of the variable; Cluster analysis unit: configured to determine the optimal number of clusters K by determining the sample classification and using the sum of squared errors within the cluster, as follows: , Among them, SSE is the sum of squared errors within the cluster, k is the number of clusters, is the number of samples contained in the i-th cluster, is the jth sample point, is the centroid of the ith cluster.

[0009] In some embodiments, using natural language processing (NLP) and knowledge graphs to construct industry lexical analysis and language models includes: Analyze the content of the tender documents using natural language processing to understand their semantics and context; this includes using lexical analysis, entity recognition, and dependency parsing to identify key words and terms, and the relationships between them; and Use knowledge graphs to integrate industry knowledge, laws and regulations, and historical bidding data to form a structured knowledge base to build an industry knowledge graph.

[0010] In some embodiments, guiding the bidders to complete the preparation of the bid document by using an intelligent question-and-answer method, and automatically generating the final bid document through an online checking tool includes: Guide bidders to fill in their bid responses one by one through a series of pre-set questions, where the pre-set questions include one or more aspects of the following bid documents: technical proposal, commercial terms, and qualification certificate; and The online inspection tool is used to intelligently check the quality of the bidding documents, including inspection of disqualified bid items, completeness inspection, and consistency inspection.

[0011] In some embodiments, the intelligent pre-scoring model includes: Data extraction and processing unit: extract objective scoring factors from the bid clearing form and bidding documents, the objective indicators of the objective scoring factors include price, technical parameters, performance, and pre-process the data extracted by the data extraction unit, the pre-processing includes data cleaning and standardization; Intelligent pre-scoring model construction unit: configured to construct a pre-scoring model according to the evaluation weights of each evaluation factor, and calculate the objective score of each bidder through the pre-scoring model for expert review and reference; wherein the pre-scoring model is constructed by a machine learning algorithm using a historical data training model, and the machine learning algorithm includes a support vector machine (SVM) or a random forest.

[0012] In some embodiments, the objective clause verification model includes: By comparing the objective clause information of machine evaluation and expert evaluation, experts are prompted to manually review the clause information of differences; when there are differences between machine evaluation and expert evaluation, the model provides intelligent prompts to help experts locate the differences.

[0013] In some embodiments, the construction of a knowledge graph of suppliers, subject matter, and construction projects to achieve intelligent analysis and risk warning includes: an entity relationship identification unit configured to extract entities and relationships from the cleaned data; A graph database storage unit: configured to store the extracted entities and relationships in a graph database; and Deep learning warning unit: configured to use deep learning for entity recognition, relationship extraction, and knowledge fusion in knowledge graph construction.

[0014] In some embodiments, the use of large model technology to build an AI bid evaluation model to achieve a dual-mode review mode of unmanned intelligent review and human-machine assisted review includes: The unmanned intelligent review includes actively learning historical bidding documents and bid evaluation reports through big model technology to form a bidding big model knowledge base exclusive to the enterprise; and On the basis of the unmanned intelligent review, experts of the human-machine assisted review conduct manual judgment and review to determine the final review result.

[0015] In a second aspect, the present invention provides a computer-readable storage medium, on which is stored a computer program for an Internet-based intelligent bidding management platform. When the computer program is executed by one or more processors, the platform module function of any embodiment as described in the first aspect is implemented.

[0016] Through the above-provided intelligent bidding management platform based on the Internet, the beneficial effects of the present invention are: By integrating advanced information technology, including big data technology, artificial intelligence technology and cloud computing technology, we can improve the efficiency and transparency of the bidding process, reduce costs and risks, and improve the quality of bid evaluation through intelligent means. By building an AI bid evaluation model, we can achieve a dual-mode evaluation mode of unmanned intelligent evaluation and human-machine assisted evaluation, aiming to improve the objectivity and work efficiency of bid evaluation, while ensuring the fairness and accuracy of the bid evaluation results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will be further described below in conjunction with the accompanying drawings.

[0018] Figure 1 It is a structural diagram of an Internet-based intelligent bidding management platform in some embodiments of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Figure 1 It is a schematic diagram of the structure 100 of an Internet-based intelligent bidding management platform according to some embodiments of the present invention.

[0021] like Figure 1As shown, in structure 100, an Internet-based intelligent bidding management platform is exemplarily described through an embodiment, specifically including: Data integration and analysis module 101: configured to establish an intelligent comprehensive economic analysis model by integrating the internal data and / or external data of the intelligent bidding platform; and to conduct a comprehensive analysis of the market, technological development and historical prices through the intelligent comprehensive economic analysis model to formulate a bidding and procurement plan; wherein the internal data includes historical bidding data, procurement contracts, and project settlement records, and the external data includes price trends, industry reports, and market analysis based on the Internet.

[0022] In some embodiments, the intelligent comprehensive economic analysis model in the above module 101 may include: a data standardization processing unit: configured to process the data extracted from multiple heterogeneous data sources into a unified format; including using the maximum-minimum normalization method to perform dimensionless processing on the data to map the data into the range of [0,1]. The formula is as follows: , Among them, max(x) is the maximum number in the data, and min(x) is the minimum number in the data; is the weighting coefficient.

[0023] The principle of the above module 101 is further explained below through an example process of a project bidding implementation.

[0024] First, data collection and integration are required; the data sources mainly include internal data, including historical bidding data, procurement contracts, project settlement records, etc. External data includes price trends, industry reports, market analysis, etc. based on the Internet. In terms of technical implementation, the use of big data to integrate data sources mainly involves collecting internal and external data related to the project, including security event data, vulnerability data, threat intelligence, etc. Integrate these data sources to establish a comprehensive data set for analysis and modeling.

[0025] Furthermore, through the ETL (Extract, Transform, Load) process, data is extracted from multiple heterogeneous data sources, and after cleaning, transformation and other processing, it is finally loaded into the target database.

[0026] Furthermore, data cleaning and preprocessing are required. Data cleaning can use data cleaning tools and algorithms to remove noise, errors and missing data in the original data to improve the quality and accuracy of the data. Data standardization can convert data from different sources into a unified format for easy comparison and analysis.

[0027] Secondly, using the data processed by data cleaning, data normalization, feature extraction and other technologies, we establish corresponding early warning models, including statistical models (such as time series analysis, regression analysis), machine learning models (such as decision trees, support vector machines, neural networks), deep learning models (such as CNN, RNN), etc. Then, through the intelligent comprehensive economic analysis model constructed, we conduct a comprehensive market, technology development and historical price analysis to formulate a bidding and procurement plan.

[0028] By integrating internal and external data to establish an economic analysis model, we conduct comprehensive market, technological development and historical price analysis, providing data support and decision-making basis for the entire bidding platform.

[0029] When building an intelligent comprehensive economic analysis model, it is also necessary to verify and test the established early warning model, evaluate its accuracy and reliability, and optimize it. Use appropriate evaluation indicators (such as accuracy, recall, F1 value, etc.) to evaluate the model and judge the performance and effectiveness of the model. According to the evaluation results, optimize and adjust the model, including feature selection, parameter adjustment, model structure adjustment, etc. As new data is continuously added, the model needs to be dynamically updated to maintain its accuracy and timeliness.

[0030] Finally, according to actual needs and risk preferences, the early warning threshold of economic risks is determined. By setting appropriate thresholds, when the monitoring indicators exceed or approach the thresholds, the early warning mechanism is triggered, and relevant personnel are notified in time and corresponding response measures are taken.

[0031] Through the above steps, we have built a complete process from data collection to analysis, demonstrated the feasibility of the "data integration and analysis module", and explained in detail the technical details involved in each step. This implementation not only covers the entire process of data integration and analysis, but also shows how to use intelligent technology to improve the efficiency and accuracy of tendering and procurement plans.

[0032] Furthermore, module 101 also includes a market competition model auxiliary decision-making unit: configured to measure the market segment competition through a market concentration model, using the concentration index CRn and HHI, and the formula is as follows: , , in, represents the winning bid amount of the i-th supplier in the market segment, and x represents the total bidding amount in the market segment.

[0033] The above-mentioned concentration indexes CRn and HHI can help tenderers understand the competitive situation in the market and adjust their procurement strategies accordingly.

[0034] Supplier group model auxiliary decision-making unit: configured to build a supplier group classification model, including classifying suppliers into leverage type, strategic type, general type and bottleneck type according to supply risk factors and procurement amount factors; the model construction formula is as follows: Supplier Type = f(Procurement Amount,Supply Risk), Among them, f is the classification function, and classification is performed according to the Karajek model.

[0035] This unit can help tenderers develop differentiated procurement strategies and optimize supplier management and procurement costs.

[0036] Purchase price model auxiliary decision-making unit: configured to include a linkage model between purchase price and raw material price, the formula is as follows: , Among them, P is the purchase order settlement unit price, P1 is the winning bid price of the framework agreement bidding, is the price of raw materials on the purchase order signing date, is the raw material price on the bid opening date, M is the raw material content percentage; and The linear regression model is used to construct the purchase reference price model. The formula is as follows: , Where Y is the reference price, is the impact factor, is the weight, , , Other cost factors, , , is the corresponding quantity.

[0037] This unit can be used to analyze the correlation between historical purchase prices and raw material prices, so that a purchase price linkage model can be established.

[0038] Furthermore, in some embodiments, the intelligent comprehensive economic analysis model further includes: Data integration unit: configured to clean the raw data of internal data and / or external data using data cleaning tools and algorithms, including noise, errors, and missing data; Data trend processing unit: It is configured to convert the reverse indicator and the moderate indicator into a positive indicator to maintain the unified trend of the indicators. The formula is as follows: , Principal component analysis unit: The configuration includes correlation analysis, which includes calculating the correlation coefficient between variables, and the formula is as follows: , in, Representation variables The correlation coefficient between , whose value is between -1 and 1; Representation variables The covariance between , to measure the extent to which two variables vary together; For variables The variance of , to measure the dispersion of the variable; For variables The variance is used to measure the dispersion of the variable.

[0039] Through correlation analysis, the correlation between different data can be identified. For example, the correlation between supplier quotations and raw material prices can be analyzed to predict the impact of price fluctuations on procurement costs. It can also be used in risk early warning models to predict potential performance risks and supply risks by analyzing suppliers' historical performance and market behavior. This helps tenderers take measures in advance to reduce procurement risks.

[0040] The results of correlation analysis can also be used to continuously optimize procurement strategies. By continuously collecting new data and feedback, the correlation analysis model is updated to make procurement strategies more scientific, reasonable and accurate.

[0041] In one embodiment, a cluster analysis unit is further included: configured to determine the optimal cluster number K by determining the sample classification and using the sum of squared errors within the cluster, and the formula is as follows: , Among them, SSE is the sum of squared errors within the cluster, k is the number of clusters, is the number of samples contained in the i-th cluster, is the jth sample point, is the centroid of the ith cluster.

[0042] The above cluster analysis can help the bidding management platform identify natural groupings of suppliers or bidding projects, thereby providing data support for bidding strategies. For example, through cluster analysis, we can find out which suppliers often bid at similar prices and quality levels, or which projects often attract similar types of suppliers. This information is very useful for optimizing the bidding process, selecting partners, and controlling costs.

[0043] Furthermore, in the standardized bidding document module 102, it can be configured to utilize natural language processing and knowledge graphs to construct industry lexical analysis and language models to achieve standardized and intelligent editing of bidding documents.

[0044] In one embodiment, the use of natural language processing (NLP) and knowledge graphs to build industry lexical analysis and language models includes: using natural language processing to analyze the content of bidding documents to understand their semantics and context; using lexical analysis, entity recognition, and dependency syntax analysis to identify key words and terms, and the relationships between them; and using knowledge graphs to integrate industry knowledge, laws and regulations, and historical bidding data to form a structured knowledge base to build an industry knowledge graph.

[0045] In the above lexical analysis, the lexical analysis technology in NLP can be used to decompose the bidding document text into terms, identify nouns, verbs, adjectives, etc., so as to build a sentence dependency graph and analyze the dependency relationship between terms to understand the sentence structure. The text in the bidding document is segmented and POS tagged to identify key words and terms. The formula can be expressed as: , where T is a text in the bidding document, It is the result of word segmentation, each of which is a term.

[0046] In one embodiment, industry lexical analysis and language models can also be developed, and knowledge graph technology can be used to integrate industry knowledge, laws and regulations, historical bidding data, etc. to form a structured knowledge base. For example, entities in the bidding documents (such as company names, project names) are linked to corresponding entities in the knowledge graph to provide accurate contextual information.

[0047] When analyzing the content of the tender documents, the semantic similarity model can be used to compare the similarity of clauses in different tender documents to ensure the consistency and standardization of the tender documents. The semantic similarity can be calculated using the cosine similarity formula: , where A and B are two vectors representing two different terms, A×B is the dot product of the vectors, and |A| and |B| are the norms of the vectors.

[0048] In the knowledge graph, ontology construction is the process of describing entity types and their relationships. In bidding management, the following ontology can be constructed: Ontology = {C, R, I}, where C is a concept set, R is a relationship set, and I is an instance set. For example, C can contain concepts such as "bidder" and "bidding project", and R can contain relationships such as "participation" and "belonging to".

[0049] In knowledge fusion, the process of resolving conflicts and redundant information by integrating knowledge from different sources. In bidding management, this can be achieved through entity linking and disambiguation technology: , where E is the extracted entity set, is a linked entity set, each Is a specific entity instance.

[0050] The above-mentioned NLP and knowledge graph technology of the present invention can be used to realize the automatic generation, review and update of bidding documents. Through NLP technology, semantic analysis and similarity comparison of bidding documents can be performed to ensure the standardization and consistency of bidding documents. Knowledge graph technology can help build the ontology of the bidding field, realize the organization and retrieval of knowledge, and improve the efficiency and quality of bidding.

[0051] Furthermore, the intelligent question-and-answer bidding document filling module 103 of the present invention is configured in one embodiment to guide bidders to complete the writing of bidding documents by adopting an intelligent question-and-answer method, and to automatically generate the final bidding documents through online checking tools.

[0052] In one embodiment scenario, it includes: guiding the bidders to fill in the bid response content one by one through a series of preset questions, wherein the preset questions include one or more aspects of the following bid documents: technical solutions, commercial terms, and qualification certificates; and using the online inspection tool to intelligently check the quality of the bid documents, including invalid bid item inspection, completeness inspection, and consistency inspection.

[0053] Through the above-mentioned guided writing method, machine learning models can be used to check grammatical errors, format errors, etc. in the bidding documents in real time based on a predefined rule base, so as to make intelligent recommendations and corrections for errors that may occur during the writing of the bidding documents. For example, the following classification model can be used to identify error types: , in, is the probability of an error occurring given the context, f(context) is the feature function of the context, and n is the total number of categories.

[0054] The working principle of module 103 is described below through an implementation scenario. In a specific bidding process, the following steps are performed: Step 1: The bidder logs in to the intelligent bidding management platform and selects the corresponding bidding project.

[0055] Step 2: The platform system guides bidders to answer questions one by one through an intelligent question-and-answer interface. The system background uses NLP technology to parse the answers and fill them into the bidding document template.

[0056] Step 3: During the filling process, bidders can use online checking tools at any time to conduct real-time quality checks and error corrections on the bidding documents.

[0057] Step 4: After the bidder completes all questions and answers, the system automatically generates the final bidding documents and provides preview and download functions.

[0058] Step 5: After the bidder submits the bid documents, the system conducts a final completeness and compliance check to ensure that the bid documents meet the bidding requirements.

[0059] Through the intelligent question-and-answer bidding document filling module 103 of the present invention, the above intelligent recommendation and error correction functions can assist tendering personnel to quickly identify and correct errors in the bidding documents, reduce manual errors, and improve the efficiency and quality of the writing of bidding documents. It can also improve the standardization and compliance of bidding documents, and improve the efficiency and accuracy of preparation.

[0060] Furthermore, in one embodiment, the intelligent bid evaluation tool set module 104 is configured to assist the bid evaluation process by establishing an intelligent bid evaluation tool; the intelligent bid evaluation tool includes an intelligent pre-scoring model and / or an objective clause verification model.

[0061] In one embodiment, the intelligent pre-scoring model of module 104 includes: Data extraction and processing unit: extract objective scoring factors from the bid clearing form and bidding documents, the objective indicators of the objective scoring factors include price, technical parameters, performance, and pre-process the data extracted by the data extraction unit, the pre-processing includes data cleaning and standardization; Intelligent pre-scoring model construction unit: configured to construct a pre-scoring model according to the evaluation weights of each evaluation factor, and calculate the objective score of each bidder through the pre-scoring model for expert review and reference; wherein the pre-scoring model is constructed by a machine learning algorithm using a historical data training model, and the machine learning algorithm includes a support vector machine (SVM) or a random forest.

[0062] In one embodiment, the objective clause verification model includes: By comparing the objective clause information of machine evaluation and expert evaluation, the model provides experts with clause information that differs from manual review; when there is a difference between machine evaluation and expert evaluation, the model provides intelligent prompts to help experts locate the difference. The specific implementation process includes: First, NLP technology is used to perform semantic analysis and key information extraction on the bidding documents to identify and verify objective terms.

[0063] Secondly, build a rule engine and define verification rules for various objective terms. The system will automatically verify the terms based on these rules.

[0064] Finally, NLP technology is used to perform semantic analysis and comparison on the evaluation results of machines and experts to identify the differences.

[0065] Furthermore, by using online assisted generation of bid evaluation reports, with the structured data of bidding documents, tender documents, and bid clearance forms as input, part of the bid evaluation report content is automatically generated according to the format requirements of the bid evaluation report. After manual review and modification, the final bid evaluation report is generated, thereby improving the efficiency of report writing by the bid evaluation report writers.

[0066] Through the intelligent bid evaluation tool set module provided by the present invention, a technical bid auxiliary evaluation tool is constructed to realize intelligent preliminary evaluation. The system automatically analyzes the evaluation elements of the evaluation points, corresponds to the corresponding content of the bidding documents, and obtains the evaluation results such as qualified and risk.

[0067] Furthermore, in the risk warning module 105, an embodiment is configured to achieve intelligent analysis and risk warning by constructing a knowledge graph of suppliers, subject matter, and construction projects. Specifically, it may include: an entity relationship identification unit configured to extract entities and relationships from the cleaned data; A graph database storage unit: configured to store the extracted entities and relationships in a graph database; and Deep learning warning unit: configured to use deep learning for entity recognition, relationship extraction, and knowledge fusion in knowledge graph construction.

[0068] Specifically in entity relationship recognition, first, extract entities and relationships from the cleaned data. For example, implement a simple named entity recognition model in Python to identify entities in text.

[0069] class NERModel(nn.Module): def __init__(self, vocab_size, embedding_dim, hidden_dim): super(NERModel, self).__init__() self.embedding = nn.Embedding(vocab_size, embedding_dim) self.lstm = nn.LSTM(embedding_dim, hidden_dim, batch_first=True) self.fc = nn.Linear(hidden_dim, vocab_size) def forward(self, x): embedded = self.embedding(x) lstm_out, _ = self.lstm(embedded) out = self.fc(lstm_out) return out In graph database storage, the extracted entities and relationships are stored in a graph database, such as Neo4j graph database, using Cypher language to create nodes and relationships CREATE (p1:Person {name: 'Alice'}) CREATE (p2:Person {name: 'Bob'}) CREATE (p1)-[:KNOWS]->(p2) To improve query efficiency, create an index in the graph database: CREATE INDEX ON :Person(name) In deep learning applications, including in knowledge graph construction, deep learning technology is mainly used for entity recognition, relationship extraction and knowledge fusion. For example, a model for relationship extraction using deep learning. Take Python code implementation as an example: class RelationExtractionModel(nn.Module): def __init__(self, input_dim, hidden_dim): super(RelationExtractionModel, self).__init__() self.lstm = nn.LSTM(input_dim, hidden_dim, batch_first=True) self.fc = nn.Linear(hidden_dim, 2) # Assume there are two types of relationships def forward(self, x): lstm_out, _ = self.lstm(x) out = self.fc(lstm_out[:, -1, :]) return out Risk warning also includes the setting and execution of rules. The preset rule base consists of specific rule conditions, and a single condition can be expressed as a left operand, a right operand, and a relational operator. Conditions are generally combined through logical relationships such as AND and OR. When all conditions of a rule are met, subsequent risk prevention and control actions need to be performed. According to the different risk levels of the hit rules, corresponding actions will be performed, mainly including blocking, waiting, and prompting warnings.

[0070] Through the above-mentioned risk warning module, intelligent analysis and risk warning of suppliers, subject matter, and construction projects can be achieved, helping enterprises to promptly discover and deal with potential risks.

[0071] In the big model application module 106, the AI ​​bid evaluation model is constructed by configuring to utilize the big model technology to realize the dual-mode review mode of unmanned intelligent review and human-machine assisted review. Specifically in one embodiment, the unmanned intelligent review may include actively learning historical bidding documents and bid evaluation reports through the big model technology to form a bidding big model knowledge base exclusive to the enterprise; and on the basis of the unmanned intelligent review, the human-machine assisted review experts perform manual judgment and review to determine the final review result.

[0072] Through the AI ​​big model, the key content in the lengthy tender documents can be quickly locked, reducing the time for experts to read and search, and improving the accuracy of the review. Using the AI ​​+ expert collaborative review method, the bidding documents are automatically evaluated for objective content, and subjective risks are predicted by AI. Experts only need to focus on review and objections, and humans and machines work together to build a review defense line.

[0073] Through the above-mentioned technical implementation, the large model application module can build an AI bid evaluation model in the intelligent bidding management platform, and realize a dual-mode review mode of unmanned intelligent review and human-machine assisted review, thereby improving the efficiency and accuracy of bid evaluation and reducing corporate procurement costs.

[0074] In a certain energy group that we have cooperated with, we built a comprehensive knowledge graph around suppliers, realized various business applications such as automatic verification of supplier registration information, industrial and commercial information verification and resident identity verification, relationship exploration and discovery of dishonest behavior, and classified management of group suppliers. Through measures such as identification of bidding document feature codes and machine ID consistency, key information similarity verification, supplier relationship analysis, and deposit remittance account inspection and control, we identified and warned of bid rigging and collusion, providing an important basis for the implementation of efficient and intelligent supervision. Through the practice of intelligent review of non-bidding business and the automated identification and analysis of procurement documents and quotation documents, we achieved a 50% increase in procurement document writing efficiency and a 70% increase in review efficiency.

[0075] In summary, through the Internet-based intelligent bidding management platform provided above, by integrating advanced information technology, including big data technology, artificial intelligence technology and cloud computing technology, the efficiency and transparency of the bidding process can be improved, costs and risks can be reduced, and the quality of bid evaluation can be improved through intelligent means. By building an AI bid evaluation model, a dual-mode evaluation mode of unmanned intelligent evaluation and human-machine assisted evaluation is realized, aiming to improve the objectivity and work efficiency of bid evaluation, while ensuring the fairness and accuracy of the bid evaluation results.

[0076] Based on the above, the present invention also discloses a computer-readable storage medium, which stores a computer program for an Internet-based intelligent bidding management platform. When the computer-readable program is executed by one or more processors, it implements the module functions described in the previous multiple embodiments or implementation methods.

[0077] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An Internet-based intelligent bidding management platform, characterized in that: include: Data integration and analysis module: configured to establish an intelligent comprehensive economic analysis model by integrating internal data and / or external data of the intelligent bidding platform; And through the intelligent comprehensive economic analysis model, a comprehensive analysis of the market, technological development and historical prices is conducted to formulate a bidding and procurement plan; wherein the internal data includes historical bidding data, procurement contracts, and project settlement records, and the external data includes price trends, industry reports, and market analysis based on the Internet; Standardized bidding document module: It is configured to use natural language processing and knowledge graph to build industry lexical analysis and language models to achieve standardized and intelligent editing of bidding documents; Intelligent question-and-answer bidding document filling module: configured to guide bidders to complete the writing of bidding documents by using intelligent question-and-answer methods, and automatically generate the final bidding documents through online checking tools; Intelligent bid evaluation tool set module: configured to assist the bid evaluation process by establishing intelligent bid evaluation tools; The intelligent bid evaluation tool includes an intelligent pre-scoring model and / or an objective clause verification model; Risk warning module: configured to achieve intelligent analysis and risk warning by building a knowledge graph of suppliers, subject matter, and construction projects; Big model application module: It is configured to use big model technology to build an AI bid evaluation model to achieve a dual-mode review mode of unmanned intelligent review and human-machine assisted review.

2. The intelligent bidding management platform according to claim 1 is characterized in that: The intelligent comprehensive economic analysis model includes: Data standardization processing unit: configured to process data extracted from multiple heterogeneous data sources into a unified format; including using the maximum-minimum normalization method to perform dimensionless processing on the data to map the data into the range of [0,1], the formula is as follows: ， Among them, max(x) is the maximum number in the data, and min(x) is the minimum number in the data; is the weighting coefficient; Market competition model auxiliary decision-making unit: It is configured to measure the market segment competition through the market concentration model, using the concentration index CRn and HHI. The formula is as follows: ， ， in, represents the winning bid amount of the i-th supplier in the market segment, and x represents the total bidding amount in the market segment; Supplier group model auxiliary decision-making unit: configured to build a supplier group classification model, including classifying suppliers into leverage type, strategic type, general type and bottleneck type according to supply risk factors and procurement amount factors; the model construction formula is as follows: Supplier Type = f(Procurement Amount,Supply Risk), Among them, f is the classification function, which is classified according to the Kalajac model; Purchase price model auxiliary decision-making unit: configured to include a linkage model between purchase price and raw material price, the formula is as follows: ， Among them, P is the purchase order settlement unit price, P1 is the winning bid price of the framework agreement bidding, is the price of raw materials on the purchase order signing date, is the raw material price on the bid opening date, M is the raw material content percentage; and The linear regression model is used to construct the purchase reference price model. The formula is as follows: ， Where Y is the reference price, is the impact factor, is the weight, , , Other cost factors, , , is the corresponding quantity.

3. The intelligent bidding management platform according to claim 2 is characterized in that: The intelligent comprehensive economic analysis model also includes: Data integration unit: configured to clean the raw data of internal data and / or external data using data cleaning tools and algorithms to remove noise, errors and missing data; Data trend processing unit: It is configured to convert the reverse indicator and the moderate indicator into a positive indicator to maintain the unified trend of the indicators. The formula is as follows: , Principal component analysis unit: The configuration includes correlation analysis, which includes calculating the correlation coefficient between variables, and the formula is as follows: ， in, Representation variables The correlation coefficient between , whose value is between -1 and 1; Representation variables The covariance between , to measure the extent to which two variables vary together; For variables The variance of , to measure the dispersion of the variable; For variables The variance of , to measure the dispersion of the variable; Cluster analysis unit: configured to determine the optimal number of clusters K by determining the sample classification and using the sum of squared errors within the cluster, as follows: ， Among them, SSE is the sum of squared errors within the cluster, k is the number of clusters, is the number of samples contained in the i-th cluster, is the jth sample point, is the centroid of the ith cluster.

4. The intelligent bidding management platform according to claim 1, characterized in that: The use of natural language processing (NLP) and knowledge graphs to build industry lexical analysis and language models includes: Analyze the content of the tender documents using natural language processing to understand their semantics and context; this includes using lexical analysis, entity recognition, and dependency parsing to identify key words and terms, and the relationships between them; and Use knowledge graphs to integrate industry knowledge, laws and regulations, and historical bidding data to form a structured knowledge base to build an industry knowledge graph.

5. The intelligent bidding management platform according to claim 1 is characterized in that: The method of guiding the bidders to complete the preparation of the bidding documents by using the intelligent question-and-answer method and automatically generating the final bidding documents through the online checking tool includes: Guide bidders to fill in their bid responses one by one through a series of pre-set questions, where the pre-set questions include one or more aspects of the following bid documents: technical proposal, commercial terms, and qualification certificate; and The online inspection tool is used to intelligently check the quality of the bidding documents, including inspection of disqualified bid items, completeness inspection, and consistency inspection.

6. The intelligent bidding management platform according to claim 1, characterized in that: The intelligent pre-scoring model includes: Data extraction and processing unit: extract objective scoring factors from the bid clearing form and bidding documents, the objective indicators of the objective scoring factors include price, technical parameters, performance, and pre-process the data extracted by the data extraction unit, the pre-processing includes data cleaning and standardization; Intelligent pre-scoring model construction unit: configured to construct a pre-scoring model according to the evaluation weights of each evaluation factor, and calculate the objective score of each bidder through the pre-scoring model for expert review and reference; wherein the pre-scoring model is constructed by a machine learning algorithm using a historical data training model, and the machine learning algorithm includes a support vector machine (SVM) or a random forest.

7. The intelligent bidding management platform according to claim 1, characterized in that: The objective clause verification model includes: By comparing the objective clause information of machine evaluation and expert evaluation, experts are prompted to manually review the clause information of differences; when there are differences between machine evaluation and expert evaluation, the model provides intelligent prompts to help experts locate the differences.

8. The intelligent bidding management platform according to claim 1, characterized in that: The construction of a knowledge graph of suppliers, subject matter, and construction projects to achieve intelligent analysis and risk warning includes: an entity relationship identification unit configured to extract entities and relationships from the cleaned data; A graph database storage unit: configured to store the extracted entities and relationships in a graph database; and Deep learning warning unit: configured to use deep learning for entity recognition, relationship extraction, and knowledge fusion in knowledge graph construction.

9. The intelligent bidding management platform according to claim 1, characterized in that: The use of big model technology to build an AI bid evaluation model to achieve a dual-mode evaluation mode of unmanned intelligent evaluation and human-machine assisted evaluation includes: The unmanned intelligent review includes actively learning historical bidding documents and bid evaluation reports through big model technology to form a bidding big model knowledge base exclusive to the enterprise; and On the basis of the unmanned intelligent review, experts of the human-machine assisted review conduct manual judgment and review to determine the final review result.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program of an Internet-based intelligent bidding management platform. When the computer program is executed by one or more processors, the platform module function as described in any one of claims 1 to 9 is implemented.

Images