An engineering bidding platform management method and system

Abstract

The application belongs to the field of platform management, and specifically discloses an engineering bidding platform management method and system, wherein the method comprises the following steps: obtaining a demand description text from a bidding project chain through a blockchain node, extracting a keyword sequence by using natural language processing, and obtaining a bidding demand feature vector; performing vector conversion processing on qualification certificate data submitted by a bidding enterprise according to the bidding demand feature vector, judging the similarity between the converted vector and the demand feature vector, and determining preliminary qualification matching if the similarity exceeds a preset threshold; extracting performance index data from historical performance records on a bidding enterprise chain, and obtaining a comprehensive qualification vector by using a vector embedding method to fuse the performance index data and the preliminary qualification matching result; and the application aims to solve the problems of low qualification examination efficiency, insufficient data reliability and non-transparent verification process in the traditional bidding process.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of platform management, and particularly relates to an engineering bidding platform management method and system. BACKGROUND

[0002] In the blockchain-based bidding and tendering scenario, the core technical problem is how to realize the dynamic and accurate matching of the bidding enterprise qualification and the tendering demand in a decentralized environment, while ensuring the transparency and non-tamperability of the verification process; this problem involves efficiently extracting the tendering demand text from the on-chain storage and converting it into a feature vector, but natural language processing may face semantic ambiguity or inaccurate keyword extraction when processing complex tendering texts, resulting in the feature vector failing to accurately reflect the demand characteristics and affecting the subsequent matching accuracy.

[0003] In addition, when converting the bidding enterprise's qualification certificates and historical performance data into vectors and calculating the similarity with the demand vectors, data format heterogeneity and non-uniform index system may cause vector space mismatch, reducing the matching precision.

[0004] Further, when calling an external oracle to obtain real-time financial statements, there is a problem of checking the data update timestamp and the bidding deadline, if the time window is too short or the network delay causes the data to be not updated in time, the financial data validity may be misjudged, affecting the reliability of the comprehensive qualification evaluation.

[0005] Finally, when the intelligent contract executes the dynamic verification score calculation and automatic approval logic, the weighted fusion algorithm may cause the score to fail to truly reflect the enterprise's ability due to unreasonable weight setting or data noise interference, and the on-chain storage of the audit log may cause the path to be incomplete when retrieving the verification path, affecting the credibility of the review result, if the node synchronization delay or the on-chain data fragmentation.

[0006] This comprehensive problem runs through the whole process from data extraction, vector conversion to dynamic verification, involving semantic analysis, data fusion, real-time verification and complex interaction of on-chain storage, directly affecting the fairness and efficiency of bidding and tendering. SUMMARY

[0007] The present application provides an engineering bidding platform management method and system, aiming to solve the problems of low qualification review efficiency, insufficient data reliability and non-transparent verification process in the traditional bidding process.

[0008] To solve the above technical problems, the technical scheme adopted by the present application is:

[0009] An engineering bidding platform management method, comprising: obtaining a demand description text from a bidding project chain storage through a blockchain node, extracting a keyword sequence using natural language processing, and obtaining a bidding demand feature vector; according to the bidding demand feature vector, performing vector conversion processing on the qualification certificate data submitted by the bidding enterprise, judging the similarity between the converted vector and the demand feature vector, and if the similarity exceeds a preset threshold, determining the preliminary matching of the qualification; extracting performance index data from the historical performance records on the bidding enterprise chain, and using a vector embedding method to fuse the performance index data and the preliminary matching result of the qualification, and obtaining a comprehensive qualification vector; obtaining real-time financial statement updates of the bidding enterprise through an intelligent contract calling external oracle interface, judging whether the financial statement update time is before the bidding deadline, and if it is before the bidding deadline, determining the validity of the financial data; according to the comprehensive qualification vector and the financial data validity result, performing weighted fusion calculation to obtain a dynamic verification score of the bidding enterprise; if the dynamic verification score is higher than a preset threshold, executing an automatic approval logic through an intelligent contract, using a hash function to store the verification process record on the chain, and obtaining an tamper-proof audit log; according to the tamper-proof audit log, retrieving the related verification path from the blockchain network, judging the path integrity, and if it is complete, determining the final output of the review result.

[0010] In an aspect of the present disclosure, the step of obtaining a demand description text from a bidding project chain storage through a blockchain node, extracting a keyword sequence using natural language processing, and obtaining a bidding demand feature vector comprises:

[0011] Obtaining a bidding project demand description text stored on the chain from a blockchain node, and obtaining original text data using a distributed query protocol;

[0012] Using word segmentation technology in natural language processing, performing word segmentation processing on the original text data to obtain a word sequence;

[0013] If the word sequence contains a preset stop word, filtering the stop word to obtain a filtered keyword sequence;

[0014] Converting the keyword sequence to a word vector sequence through a pre-trained word embedding model to generate an initial feature vector;

[0015] According to the initial feature vector, using principal component analysis algorithm to perform dimension reduction processing on the vector dimension to obtain an optimized bidding demand feature vector;

[0016] If the similarity between the optimized feature vector and the preset business demand template exceeds a preset threshold, it is determined that the feature vector meets the business demand, and a structured demand description is generated;

[0017] Using a clustering algorithm to classify the bidding project through the structured demand description to obtain a classification result.

[0018] In an aspect of the present disclosure, the step of judging the similarity between the converted vector and the demand feature vector, and determining the preliminary qualification matching if the similarity exceeds the preset threshold, according to the bid demand feature vector, comprises:

[0019] Obtaining the bid demand data and extracting the feature vector therefrom, using the keyword extraction and semantic analysis method to generate the demand feature vector;

[0020] Structuring the qualification certificate data submitted by the bidding enterprise, using the text parsing technology to extract the key field and generate the qualification data;

[0021] Converting the qualification data into the qualification feature vector through the pre-trained word vector model;

[0022] Calculating the cosine similarity of the qualification feature vector and the demand feature vector to obtain the similarity value;

[0023] If the similarity value exceeds the preset threshold, determining the preliminary qualification matching and generating the matching result;

[0024] Using the clustering analysis method to group the bidding enterprises according to the matching result to obtain the grouping result;

[0025] According to the grouping result, extracting the qualification feature vector of each group of bidding enterprises to generate the ranking list.

[0026] In an aspect of the present disclosure, the step of extracting the performance indicator data from the historical performance records on the bidding enterprise chain and using the vector embedding method to fuse the performance indicator data and the preliminary qualification matching result to obtain the comprehensive qualification vector comprises:

[0027] Extracting the performance indicator data from the historical performance records on the bidding enterprise chain and using the preset indicator extraction rule to obtain the structured performance indicator data set;

[0028] Calculating the weight value of each indicator through the structured performance indicator data set and using the weighted average method to generate the weighted performance indicator set;

[0029] Obtaining the enterprise qualification data from the preliminary qualification matching result and using the vector embedding method to fuse the weighted performance indicator set and the enterprise qualification data to generate the initial comprehensive qualification vector;

[0030] If the dimension of the initial comprehensive qualification vector exceeds the preset threshold, using the principal component analysis method for dimension reduction processing to obtain the optimized comprehensive qualification vector;

[0031] According to the optimized comprehensive qualification vector, the similarity between vectors is calculated, the cosine similarity algorithm is adopted, and the qualification matching degree between enterprises is determined;

[0032] Through the qualification matching degree between enterprises, combined with the preset classification rule, the qualification grade of the enterprise is judged, and the final qualification evaluation result is generated;

[0033] According to the final qualification evaluation result, a data correlation analysis method is adopted to correlate the evaluation result with the on-chain historical performance record to generate a verification data set of the comprehensive qualification vector.

[0034] In an aspect of the present disclosure, the step of obtaining real-time financial statement updates of the bidding enterprise through the intelligent contract calling external oracle interface, judging whether the financial statement update time is before the bidding deadline, and determining the validity of the financial data if it is before the bidding deadline, comprises:

[0035] Obtain the financial statement update timestamp of the bidding enterprise from the external data source through the oracle interface and store it as first time data;

[0036] Using time comparison logic, compare the first time data with the preset bidding deadline, if the first time data is earlier than the bidding deadline, determine that the financial statement update time is valid, and obtain the validity determination result;

[0037] Obtain the validity determination result through the intelligent contract interface to generate the financial data validity state;

[0038] According to the validity state, call the oracle interface to obtain the complete data set of the financial statement and store it as the first financial data;

[0039] Verify the integrity and consistency of the first financial data through the data verification algorithm to obtain the second financial data that passes the verification;

[0040] Using classification algorithm to conduct financial health assessment on the second financial data, generate financial health score;

[0041] According to the financial health score and the validity determination result, the final validity of the financial data of the bidding enterprise is obtained.

[0042] In an aspect of the present disclosure, the step of performing weighted fusion calculation according to the comprehensive qualification vector and the financial data validity result to obtain the dynamic verification score of the bidding enterprise comprises:

[0043] Obtain the bidding enterprise information and the financial data validity, judge the data integrity and consistency through the preset data verification rule, and obtain the verified financial data quality;

[0044] According to the verified financial data quality, extract the qualification evaluation index, use the factor analysis method to determine the comprehensive qualification vector;

[0045] If at least one index in the comprehensive qualification vector exceeds the preset threshold, then by linear weighting method, the qualification evaluation index and the financial data quality are fused to obtain a preliminary fusion score;

[0046] For the preliminary fusion score, a preset fusion weight distribution is adopted to perform weighted fusion calculation to obtain an adjusted fusion score;

[0047] According to the adjusted fusion score, a logistic regression model is applied to judge the enterprise credit level and output a dynamic verification score;

[0048] If the dynamic verification score is lower than the preset threshold, then by the score calculation model, the fusion weight distribution is adjusted to recalculate the dynamic verification score;

[0049] Get the recalculated dynamic verification score, output the evaluation result, and generate the comprehensive evaluation score of the bidding enterprise.

[0050] In an aspect of the present disclosure, if the dynamic verification score is higher than the preset threshold, the step of performing automatic approval logic by smart contract, using a hash function to store the verification process record on the chain, and obtaining an unforgeable audit log, includes:

[0051] If the verification score is higher than the preset threshold, then by the trigger logic of the smart contract, the data to be approved is obtained, and the automatic approval operation is performed to obtain the approval result;

[0052] Using a hash function to encrypt the approval result and the verification record to generate encrypted data;

[0053] Write the encrypted data into the on-chain storage through the blockchain network to generate a storage address;

[0054] According to the storage address, the on-chain data is obtained, the data integrity is verified, and the verification state is obtained;

[0055] If the verification state is complete, then by the smart contract, an audit log is generated to record the approval and storage information;

[0056] According to the content of the audit log, the decentralized network is used to distribute the log data to generate a distribution confirmation;

[0057] Through the distribution confirmation, the consensus feedback of the network node is obtained, and the final on-chain record state is obtained.

[0058] In an aspect of the present disclosure, the step of retrieving the relevant verification path from the blockchain network according to the unforgeable audit log, judging the path integrity, and if complete, determining the final output of the review result, includes:

[0059] The audit logs are obtained from the blockchain network, and a distributed node query protocol is used to extract the log dataset containing timestamps and transaction hashes to obtain the raw audit logs.

[0060] For the original audit logs, the log structure is parsed, the node sequence and hash chain in the verification path are extracted, and a path dataset is generated;

[0061] If the node sequence in the path dataset is consistent with the preset blockchain topology, the Merkle tree verification algorithm is used to calculate the hash value of each node in the path, determine the path integrity, and obtain the integrity verification result.

[0062] Based on the integrity verification results, if the integrity verification passes, the transaction timestamps and associated metadata are extracted from the path dataset to generate the review dataset;

[0063] The SHA-256 algorithm is used to perform hash calculation on the review dataset to generate review hash values ​​and determine the uniqueness of the review results;

[0064] By using network communication protocols, the audit hash value is compared with the preset audit standards to determine whether the audit result meets the standards and output the final audit result.

[0065] In another aspect, this disclosure also relates to an engineering bidding platform management system, comprising:

[0066] The requirement feature extraction module is configured to obtain the requirement description text from the blockchain storage of the bidding project chain through a blockchain node, extract the keyword sequence using natural language processing, and obtain the bidding requirement feature vector.

[0067] The qualification matching module is configured to perform vector transformation processing on the qualification certificate data submitted by the bidding company based on the bidding requirement feature vector, determine the similarity between the transformed vector and the requirement feature vector, and determine the preliminary qualification matching if the similarity exceeds a preset threshold.

[0068] The performance fusion module is configured to extract performance indicator data from historical performance records on the bidding enterprise chain, and use a vector embedding method to fuse the performance indicator data with the preliminary qualification matching results to obtain a comprehensive qualification vector.

[0069] The financial verification module is configured to obtain the real-time financial statement updates of the bidding company by calling the external oracle interface through a smart contract, determine whether the financial statement update time is before the bidding deadline, and determine the validity of the financial data if it is before the bidding deadline.

[0070] The score calculation module is configured to perform weighted fusion calculation according to the comprehensive qualification vector and the financial data validity result, and obtain a dynamic verification score of the bidding enterprise.

[0071] The contract execution module is configured to perform automatic approval logic through the smart contract if the dynamic verification score is higher than a preset threshold, perform on-chain storage of the verification process record by using a hash function, and obtain an unforgeable audit log.

[0072] The audit verification module is configured to retrieve a relevant verification path from the blockchain network according to the unforgeable audit log, judge the path integrity, and determine the final output of the review result if the path is complete.

[0073] Compared with the prior art, the present application has the following beneficial effects:

[0074] The blockchain node obtains the demand description text from the bidding project chain, extracts keywords to generate a demand feature vector by using natural language processing, and vectorizes the bidding enterprise qualification certificate data, calculates the similarity with the demand feature vector to preliminarily match the qualification. At the same time, the historical performance indicators of the enterprise are extracted from the chain, and the preliminary matching result is fused by using a vector embedding method to generate a comprehensive qualification vector. The present application verifies the financial statement update time by calling an external oracle interface through a smart contract, ensures the validity of the financial data, and performs weighted fusion calculation combined with the comprehensive qualification vector to obtain a dynamic verification score. If the score exceeds the threshold, the smart contract automatically executes the approval logic, the verification process generates an unforgeable audit log by using a hash function, and the path integrity is verified based on the blockchain network to ensure that the review result is traceable. The present application significantly improves the efficiency, transparency and credibility of the bidding qualification verification, and optimizes the bidding process. BRIEF DESCRIPTION OF DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0076] Fig. 1 The flowchart of the bidding platform management method in the present application.

[0077] Fig. 2 One of the schematic diagrams of the bidding platform management method in the present application.

[0078] Fig. 3 The second schematic diagram of the bidding platform management method in the present application. DETAILED DESCRIPTION

[0079] The application will be further described below in conjunction with the embodiments, and the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0080] Embodiment one

[0081] Please refer to Figs. 1-3 As shown in the figure, the embodiment discloses an engineering bidding platform management method, which comprises the following steps: obtaining demand description text from a bidding project chain through a blockchain node, extracting a keyword sequence by using natural language processing, and obtaining a bidding demand feature vector; according to the bidding demand feature vector, performing vector conversion processing on the qualification certificate data submitted by a bidding enterprise, judging the similarity between the converted vector and the demand feature vector, and if the similarity exceeds a preset threshold, determining that the qualification is preliminarily matched; extracting performance index data from historical performance records on a bidding enterprise chain, and obtaining a comprehensive qualification vector by using a vector embedding method to fuse the performance index data and the preliminary matching result of the qualification; obtaining real-time financial statement updates of the bidding enterprise by calling an external oracle interface through a smart contract, judging whether the financial statement update time is before the bidding deadline, and if it is before the bidding deadline, determining that the financial data is valid; according to the comprehensive qualification vector and the financial data validity result, performing weighted fusion calculation to obtain a dynamic verification score of the bidding enterprise; if the dynamic verification score is higher than a preset threshold, executing automatic approval logic through a smart contract, storing the verification process record on the chain by using a hash function to obtain an unalterable audit log; and according to the unalterable audit log, retrieving a related verification path from a blockchain network, judging the path integrity, and if the path is complete, determining that the review result is finally output.

[0082] The application obtains demand description text from a bidding project chain through a blockchain node, generates a demand feature vector by using natural language processing to extract keywords, and vectorizes the qualification certificate data of a bidding enterprise to preliminarily match the qualification by calculating the similarity with the demand feature vector. Meanwhile, the application extracts enterprise historical performance indicators from the chain, generates a comprehensive qualification vector by using a vector embedding method to fuse the preliminary matching result, verifies the financial statement update time through a smart contract calling an external oracle interface to ensure the validity of the financial data, and combines the comprehensive qualification vector to perform weighted fusion calculation to obtain a dynamic verification score. If the score exceeds a threshold, the smart contract automatically executes approval logic, the verification process generates an unalterable audit log by using a hash function, and the path integrity is verified based on the blockchain network to ensure that the review result is traceable. The application significantly improves the efficiency, transparency and credibility of bidding qualification verification, and optimizes the bidding process.

[0083] Embodiment two

[0084] Referring to Figs. 1-3 As shown in the figure, the embodiment is further optimized on the basis of embodiment one. In the embodiment, the bidding platform management method can specifically include:

[0085] S1, obtaining the demand description text from the blockchain node from the bidding project chain storage, extracting the keyword sequence by natural language processing, and obtaining the bidding demand feature vector.

[0086] Obtain the bidding project demand description text stored on the chain from the blockchain node, obtain the original text data by using the distributed query protocol, and obtain the word sequence by using the word segmentation technology in natural language processing. If the word sequence contains a preset stop word, filter the stop word to obtain the filtered keyword sequence;

[0087] Convert the keyword sequence to a word vector sequence by using a pre-trained word embedding model to generate an initial feature vector. According to the initial feature vector, use principal component analysis algorithm to reduce the dimension of the vector, and obtain the optimized bidding demand feature vector. If the similarity of the optimized feature vector and the preset business demand template exceeds the preset threshold, it is determined that the feature vector meets the business demand, and a structured demand description is generated.

[0088] By using the structured demand description, a clustering algorithm is used to classify the bidding project to obtain a classification result.

[0089] For example, when obtaining the bidding project demand description text from the blockchain node, the text data stored on the chain can be called by the smart contract. Assuming that Ethereum blockchain is used, the node calls the ImmGetDescription function of the smart contract through the JSON-RPC interface to obtain the demand description file stored in IPFS.

[0090] The node downloads the file using the get method of IPFS, which takes about 0.8 seconds, and obtains the text content as "This project needs to develop a supply chain management system, which needs to support real-time inventory tracking, data analysis module, compatible with MySQL database, and response time less than 0.5 seconds."

[0091] Subsequently, the natural language processing technology is used to extract the keyword sequence, the BERT model is used to perform word segmentation and semantic analysis on the text, the maximum sequence length is set to 128, and the word embedding dimension is set to 768.

[0092] After word segmentation, the word sequence is obtained: {“supply chain management”, “real-time inventory tracking”, “data analysis module”, “MySQL database”, “response time”}. The term frequency-inverse document frequency (TF-IDF) algorithm is used to calculate the term frequency-inverse document frequency. Assuming that the term frequency of supply chain management is 3, the TF value is 0.15, and the IDF value is 2.3, the keyword weight is 0.345.

[0093] The keyword sequence is vectorized, and a 300-dimensional feature vector is generated using the Word2Vec model. Assuming that the vector value of supply chain management is [0.12, -0.45, 0.67,], the cosine similarity between vectors is calculated to obtain the similarity between real-time inventory tracking and data analysis module, which is 0.82, indicating that there is a strong association between the demands.

[0094] Finally, all keyword feature vectors are integrated to form a 768-dimensional comprehensive feature vector of the bidding demand, which is stored in the local database for subsequent matching analysis. The entire process is realized through an automated script without human intervention, forming a closed loop from data acquisition to feature extraction.

[0095] S2, according to the bidding demand feature vector, the qualification certificate data submitted by the bidding enterprise is vectorized and processed, and the similarity between the transformed vector and the demand feature vector is judged. If the similarity exceeds the preset threshold, it is determined that the qualification is preliminarily matched.

[0096] Obtain the bidding demand data, extract the feature vector from it, use keyword extraction and semantic analysis methods to generate the demand feature vector. The qualification certificate data submitted by the bidding enterprise is structured, and the key fields are extracted using text parsing technology to generate the qualification data. The qualification data is converted into a qualification feature vector through a pre-trained word vector model. Calculate the cosine similarity between the qualification feature vector and the demand feature vector to obtain the similarity value. If the similarity value exceeds the preset threshold, it is determined that the qualification is preliminarily matched, and the matching result is generated. For the matching result, use clustering analysis method to group the bidding enterprises, and obtain the grouping result. According to the grouping result, extract the qualification feature vector of each group of bidding enterprises to generate a ranking list.

[0097] For example, based on the bidding demand feature vector, the qualification certificate data of the bidding enterprise is vectorized and the similarity is judged. This process can be realized through the following information technology. Assuming that the bidding demand is a construction engineering project, the enterprise is required to have a first-class general construction contractor qualification, complete 5 similar projects in the past three years, and the registered capital is not less than 50 million yuan; the demand feature vector can be defined as [1, 5, 5000], representing the qualification level (1 for first-class), the number of project experience, and the registered capital.

[0098] The data submitted by bidding enterprise A is: first-level general construction contracting qualification, completion of four similar projects in the past three years, registered capital of 60 million yuan, which needs to be converted into vector [1, 4, 6000]; to unify the dimension, first normalize the vector, using Min-Max normalization algorithm.

[0099] Assuming that the maximum values of each dimension are respectively 2 (two levels are 2), the number of projects is 10, and the registered capital is 100 million yuan, and the minimum values are respectively 0, 0, and 0, then the demand vector is normalized to [(1-0) / (2-0), (5-0) / (10-0), (5000-0) / (10000-0)] = [0.5, 0.5, 0.5], and the vector of enterprise A is normalized to [0.5, 0.4, 0.6]. Next, calculate the similarity of the two vectors, using cosine similarity algorithm;

[0100] The formula is: ;

[0101] Calculate the dot product: 0.5*0.5+0.5*0.4+0.5*0.6=0.75;

[0102] Modulus = ;

[0103] = ;

[0104] The cosine similarity is ≈0.987. If the preset threshold is 0.9, because 0.987>0.9, it is determined that the qualification of enterprise A is preliminarily matched.

[0105] To ensure logical rigor, weights can be introduced, such as the weight of qualification level 0.5, project experience 0.3, and capital 0.2. Calculate the weighted similarity to further refine the matching accuracy, and the system automatically outputs the matching result and records the log for subsequent audit.

[0106] S3, extract performance index data from the chain of bidding enterprises' historical performance records, and use vector embedding method to integrate performance index data and preliminary qualification matching results to obtain comprehensive qualification vector.

[0107] S3-1, extract performance indicator data from the bidding enterprise chain historical performance record, adopt the preset index extraction rule, obtain the structured performance indicator data set. Through the structured performance indicator data set, the weight value of each indicator is calculated, and the weighted performance indicator set is generated by using the weighted average method. From the preliminary matching result of the qualification, the enterprise qualification data is obtained, and the weighted performance indicator set and the enterprise qualification data are fused by using the vector embedding method to generate the initial comprehensive qualification vector. If the dimension of the initial comprehensive qualification vector exceeds the preset threshold, the principal component analysis method is used for dimension reduction processing to obtain the optimized comprehensive qualification vector. According to the optimized comprehensive qualification vector, the similarity between vectors is calculated, and the cosine similarity algorithm is used to determine the qualification matching degree between enterprises. Through the qualification matching degree between enterprises, combined with the preset classification rule, the qualification grade of the enterprise is judged, and the final qualification evaluation result is generated. According to the final qualification evaluation result, the data correlation analysis method is used to associate the evaluation result with the chain historical performance record to generate a verification data set of the comprehensive qualification vector.

[0108] For example, from the bidding enterprise chain historical performance record, performance indicator data can be extracted through blockchain smart contract and data parsing technology. Suppose the chain stores the project records of enterprise A in the past 5 years, including project amount, completion time, customer rating and other information. The smart contract calls the API interface to extract data from the chain to obtain the performance indicators of enterprise A: cumulative project amount 500 million yuan, average project cycle 6 months, customer average rating 4.8, assuming full score is 5.

[0109] S3-2, the data parsing module standardizes these indicators, for example, normalizing the amount to the [0,1] interval, 500 million yuan corresponds to 0.85, assuming the maximum amount is 100 million yuan, and the cycle and rating are normalized to 0.67 and 0.96 respectively.

[0110] Then, the performance indicator data and the preliminary matching result of the qualification are fused by using the vector embedding method. The qualification matching result comes from the enterprise qualification database, including the registered capital, qualification level, number of technical personnel, etc.

[0111] For example, the registered capital of enterprise A is 80 million yuan, which is normalized to 0.8, the qualification level is first class, which is assigned to 1, and the number of technical personnel is 50, which is normalized to 0.5.

[0112] S3-3, use Word2Vec algorithm to convert performance indicators (0.85, 0.67, 0.96) and qualification data (0.8, 1, 0.5) into high-dimensional vectors respectively, with a dimension of 128.

[0113] Word2Vec maps each indicator into a semantic vector through a pre-trained model, capturing the potential correlation between indicators, such as the correlation between the amount and the qualification level. When fused, a weighted average method is used, with the performance vector weight set to 0.6 and the qualification vector weight set to 0.4, to calculate the comprehensive vector V = 0.6 × V 业绩 + 0.4 × V 资质 , obtaining a 128-dimensional comprehensive qualification vector.

[0114] The analysis process ensures semantic consistency between vectors, and the fusion effect is verified by cosine similarity. Assuming that the similarity between V and the ideal qualification vector is 0.92, it indicates that the fusion result is reliable.

[0115] On the logic chain, on-chain data extraction ensures data authenticity, standardized processing unifies dimensions, vector embedding captures semantics, and fusion algorithm balances performance and qualification weights, finally outputting a comprehensive qualification vector for subsequent bid evaluation.

[0116] Step S3-4, if the dimension of the initial comprehensive qualification vector exceeds the preset threshold, principal component analysis method is used for dimension reduction processing to obtain the optimized comprehensive qualification vector. The initial comprehensive qualification vector usually contains multiple dimensions of qualification data and performance indicators, such as 128 dimensions or higher, which may increase the calculation complexity and storage requirements.

[0117] Therefore, through the principal component analysis method, the high-dimensional vector is projected into a low-dimensional space, retaining the main information. For example, assuming that the initial comprehensive qualification vector is 128-dimensional, the principal component analysis retains the first 32 dimensions as the optimized comprehensive qualification vector. These dimensions usually correspond to the core features of qualification and performance, such as technical certification level or project delivery efficiency.

[0118] Specifically, principal component analysis first calculates the covariance matrix of the initial comprehensive qualification vector to determine the main variation direction of the data; then, through eigenvalue decomposition, the principal components with larger eigenvalues are selected as the projection basis vectors. The initial comprehensive qualification vector is projected onto these basis vectors to generate a low-dimensional optimized vector. This method effectively reduces redundant information while retaining the core features of enterprise qualification and performance. For example, in the Internet of Things device procurement scenario, the optimized comprehensive qualification vector may highlight Internet of Things certification and delivery time, and weaken the influence of secondary information, improving the efficiency of subsequent matching calculations.

[0119] Step S3-5, according to the optimized comprehensive qualification vector, the similarity between vectors is calculated, and the cosine similarity algorithm is used to determine the matching degree of the qualifications between enterprises. The optimized comprehensive qualification vector reflects the comprehensive ability of the bidding enterprise. By comparing the similarity between the vectors of different enterprises, the matching degree of the qualifications is determined. For example, assume that the demand feature vector of a certain bidding project emphasizes 5G technology and fast delivery, and the comprehensive qualification vector of a certain enterprise highlights 5G certification and 6-month delivery record. The angle between the two in the vector space is small, and the similarity value may be 0.88, indicating a high matching degree.

[0120] In one possible implementation, the cosine similarity calculation is realized by the ratio of vector dot product and module length, focusing on the similarity of vector direction; the calculation result is between 0 and 1, and the value closer to 1 indicates that the matching degree of enterprise qualification and bidding demand is higher;

[0121] It should be noted that the similarity calculation can also be combined with other methods, such as Euclidean distance, as a supplementary verification to ensure the robustness of the result. The similarity value serves as the basis for subsequent qualification level judgment, directly affecting the evaluation result of the enterprise.

[0122] Step S3-6, through the matching degree of the qualifications between enterprises, combined with the preset classification rules, the qualification level of the enterprise is judged, and the final qualification evaluation result is generated; the classification rules are customized based on the business demand of the bidding project, for example, in the scene of Internet of Things device procurement, the rule may stipulate that the similarity greater than 0.85 is excellent qualification, and 0.7 to 0.85 is qualified qualification.

[0123] For example, if the similarity of the comprehensive qualification vector of a certain enterprise and the demand feature vector is 0.88, then according to the classification rule, it is determined to be excellent qualification; the qualification evaluation result is recorded in a structured form, for example, enterprise A: excellent qualification; similarity: 0.88.

[0124] In one embodiment, the classification rule can also be combined with other business factors, such as the registered capital or industry experience of the enterprise. For example, if the registered capital of the enterprise exceeds a certain scale, it may obtain a higher qualification level under the same similarity. This method improves the comprehensiveness of the evaluation and adapts to the needs of different bidding scenes. The generated qualification evaluation result provides an important basis for subsequent comprehensive evaluation.

[0125] Step S3-7, according to the final qualification evaluation result, the data association analysis method is used to associate the evaluation result with the on-chain historical performance record, and the verification data set of the comprehensive qualification vector is generated. Data association analysis verifies the accuracy of the evaluation by comparing the evaluation result with the on-chain historical performance record.

[0126] For example, if the qualification evaluation result of a certain enterprise is excellent qualification, the correlation analysis will check whether its historical performance record contains high-quality Internet of Things project delivery records, such as delivery time: 6 months; satisfaction: 95%, if the record is consistent with the evaluation result, the verification data set is generated, and the correlation information is recorded.

[0127] Specifically, data correlation analysis is achieved by matching key fields, such as comparing Internet of Things certification in the evaluation result with Internet of Things project field in the historical performance record; the verification data set is stored in a structured form, such as enterprise A: excellent qualification; associated performance: Internet of Things project, satisfaction 95%.

[0128] This method ensures the credibility of the evaluation result and provides data support for subsequent audits.

[0129] Step S4, call the external oracle interface through the smart contract to obtain the real-time financial statement update of the bidding enterprise, and judge whether the financial statement update time is before the bidding deadline. If it is before the bidding deadline, it is determined that the financial data is valid. The smart contract interacts with external data sources through the oracle interface to obtain the latest financial statement timestamp of the bidding enterprise. For example, the financial statement update time of a certain enterprise is XX year X month X day, and the bidding deadline is XX year X month X day, the timestamp meets the requirements, indicating that the financial data is valid.

[0130] Step S4-1, obtain the financial statement update timestamp of the bidding enterprise from the external data source through the oracle interface, and store it as the first time data. The oracle serves as a bridge between the blockchain and the external world, and obtains the timestamp information of the financial statement from a trusted data source.

[0131] For example, the oracle may obtain the timestamp from the enterprise financial system or a third-party data platform, such as 20XX-0X-0X, XX:XX:XX, and the first time data is stored in the temporary storage area of the blockchain to ensure the traceability of the data.

[0132] In one possible implementation, the oracle interface adopts a multi-source verification mechanism to obtain timestamps from multiple data sources and determine the final value through a consensus algorithm. For example, if three data sources provide 20XX-0X-0X, XX:XX:XX, 20XX-0X-0X, XX:XX:XX, 20XX-0X-0X, XX:XX:XX, take the majority of the timestamps as the first time data. This method improves the reliability of data acquisition and reduces the potential risk of a single data source.

[0133] Step S4-2, compare the first time data with the preset bid deadline using time comparison logic. If the first time data is earlier than the bid deadline, determine that the financial statement update time is valid, and obtain a validity determination result. The time comparison logic determines the validity of the financial statement by comparing the order of the time stamps. For example, if the first time data is "20XX-0X-0X XX:XX:XX" and the bid deadline is "20XX-0X-0X XX:XX:XX", the time stamp meets the requirements and is determined to be valid.

[0134] Specifically, the time comparison logic can be implemented through simple date comparison, for example, by converting the time stamp to a standard format and comparing the year, month, day, hour, minute, and second one by one.

[0135] The validity determination result is recorded in the form of a Boolean value, for example, valid: yes.

[0136] It should be noted that if the time stamp is later than the bid deadline, the determination result is invalid, and the enterprise may need to resubmit the financial data. This strict time verification ensures the fairness of the evaluation process.

[0137] Step S4-3, obtain the validity determination result through the smart contract interface and generate the financial data validity state. The smart contract updates the financial data state of the enterprise according to the validity determination result.

[0138] For example, if the validity determination result is valid: yes, the smart contract generates the state financial data valid and records it on the chain. The state information is stored in a structured form, for example, enterprise A: financial data valid; time stamp: 2025-0X-0X.

[0139] In one embodiment, the smart contract can also trigger a notification mechanism to send the validity state feedback to the bidding enterprise. For example, if the financial data is valid, the smart contract notifies the enterprise to continue the subsequent evaluation process through the on-chain message system. This automated feedback improves the transparency and efficiency of the evaluation.

[0140] Step S4-4, according to the validity state, call the oracle interface to obtain the complete data set of the financial statement and store it as the first financial data. If the financial data is valid, the oracle interface obtains the complete financial statement data from external data sources, such as the balance sheet and profit statement.

[0141] For example, the first financial data of an enterprise may include total assets: 50 million yuan; net profit: 5 million yuan; and debt ratio: 30%.

[0142] These data are stored in the distributed storage of the blockchain, ensuring that they cannot be tampered with.

[0143] Specifically, the oracle interface fetches data through a secure communication protocol and performs preliminary formatting, such as converting data to JSON format.

[0144] It should be noted that the data acquisition process may involve multi-party verification, such as confirming the integrity of the data through multiple oracle nodes. The first financial data provides comprehensive basic information for subsequent financial health assessment.

[0145] Step S4-5, the integrity and consistency of the first financial data are verified through a data verification algorithm, and the second financial data that passes the verification is obtained. The data verification algorithm checks the format, field integrity and logical consistency of the first financial data. For example, the algorithm may check whether the total assets in the balance sheet are equal to the total liabilities + owner's equity, or whether the net profit in the profit table is consistent with the income and expenditure data.

[0146] If the verification is passed, the second financial data is generated.

[0147] In one possible implementation, the data verification algorithm combines rule verification and statistical analysis. For example, rule verification ensures that the field format meets the standard, and statistical analysis detects outliers in the data, such as abnormally high debt ratios. The second financial data that passes the verification is stored in a structured form, such as Enterprise A: total assets 50 million yuan, net profit 5 million yuan.

[0148] This method ensures the reliability of financial data and provides high-quality input for subsequent assessment.

[0149] Step S4-6, a classification algorithm is used to assess the financial health of the second financial data, and a financial health score is generated. The classification algorithm assesses the financial health of the enterprise based on key indicators of financial data.

[0150] For example, the algorithm may classify enterprises into three levels of health, general, and risk based on indicators such as net profit, debt ratio, and cash flow.

[0151] Assuming that the net profit of a certain enterprise is 5 million yuan, the debt ratio is 30%, and the cash flow is stable, the classification algorithm may generate a financial health score of 0.9, indicating that the enterprise's financial situation is good.

[0152] Specifically, the classification algorithm can be implemented through decision tree or logistic regression. For example, the decision tree makes a hierarchical judgment based on preset thresholds, such as debt ratio less than 50%, and finally generates a score.

[0153] It should be noted that the score can be further refined into multiple dimensions, such as profitability: 0.85; solvency: 0.9; the financial health score directly affects the comprehensive assessment result of the enterprise.

[0154] Step S4-7, determine the final validity of the financial data of the bidding enterprise according to the financial health score and the validity determination result. If the financial health score is higher than the preset threshold, such as 0.8, and the validity determination is yes, it is determined that the financial data is finally valid.

[0155] For example, the financial health score of a certain enterprise is 0.9, and the validity determination is yes, then the final validity is passed.

[0156] The result is recorded in a structured form, for example, enterprise A: financial data passed; score: 0.9.

[0157] In an embodiment, the final validity can also be combined with other factors, such as the industry ranking or credit record of the enterprise. For example, if the enterprise is a leading enterprise in the industry, even if the score is slightly lower than the threshold, it can be determined to be valid by adjusting the weight. This flexible determination method adapts to the needs of different bidding scenarios.

[0158] Step S5, according to the comprehensive qualification vector and the financial data validity result, weighted fusion calculation is carried out to obtain the dynamic verification score of the bidding enterprise. The comprehensive qualification vector contains the technical ability and historical performance information of the enterprise, and the financial data validity result reflects its financial health status; both are fused by weighting to generate a comprehensive score. For example, the comprehensive qualification vector of a certain enterprise highlights 5G certification, and the financial health score is 0.9, and through fusion calculation, a dynamic verification score of 0.87 is generated.

[0159] Step S5-1, obtain the bidding enterprise information and financial data validity, judge the data integrity and consistency through the preset data verification rule, and obtain the verified financial data quality. The data verification rule checks the integrity of the enterprise information and the financial data to ensure the reliability of the data.

[0160] For example, if the registered capital of a certain enterprise is 50 million yuan, the financial report shows that the total assets are 50 million yuan, and all the required fields are complete, the verification is passed, and the financial data quality is high.

[0161] The verification process includes comparing the enterprise name, registered capital and key fields in the financial report to ensure consistency of information.

[0162] Specifically, the data verification rule first checks the field integrity, such as confirming whether the financial report contains assets, liabilities, income and other required items. Secondly, logical verification is carried out to ensure that the data is reasonable, such as the total assets should be equal to the sum of liabilities and owner's equity.

[0163] If it is found that the field is missing or the data is contradictory, the system will mark it as low quality and require the enterprise to supplement the data. The verified financial data quality provides a reliable basis for subsequent evaluation and reduces the evaluation deviation caused by data errors.

[0164] Step S5-2, according to the verified financial data quality, extract the qualification evaluation index, and determine the comprehensive qualification vector by using factor analysis method. Factor analysis method extracts key indicators by analyzing the correlation between financial data and qualification data to form a comprehensive qualification vector. For example, the net profit indicator in the financial data of a certain enterprise is outstanding, and the technical certification level in the qualification data is high. The factor analysis method extracts both as the main factor to generate a comprehensive qualification vector.

[0165] This vector represents the comprehensive ability of the enterprise in a multi-dimensional form, such as technical certification: 0.9; profitability: 0.8.

[0166] In one possible implementation, the factor analysis method first calculates the correlation matrix of each indicator to determine which indicators have a greater impact on the evaluation result.

[0167] For example, technical certification and net profit may have a high correlation and be selected as the main factor. Then, through factor loading analysis, the qualification and financial data are projected into the factor space to generate a comprehensive qualification vector. This method can effectively integrate multi-source data and improve the comprehensiveness and accuracy of the evaluation; the generated comprehensive qualification vector provides a core input for subsequent fusion calculation.

[0168] Step S5-3, if at least one indicator in the comprehensive qualification vector exceeds the preset threshold, then the qualification evaluation index and the financial data quality are fused by a linear weighting method to obtain a preliminary fusion score.

[0169] The linear weighting method weights and sums the qualification and financial indicators according to the preset weights.

[0170] For example, if the technical certification indicator value in the comprehensive qualification vector is 0.9, which exceeds the threshold of 0.8, and the financial data quality is high, then through weight distribution, the technical certification is 0.6 and the financial data quality is 0.4, and the preliminary fusion score is calculated as 0.85.

[0171] Specifically, the linear weighting method calculates the weighted sum by assigning weights to each indicator.

[0172] For example, the weight of technical certification is 0.6 and the weight of financial data quality is 0.4, and the fusion formula is: score = 0.6 x technical certification value + 0.4 x financial data quality value.

[0173] It should be noted that the weight distribution can be dynamically adjusted according to the priority of the bidding project. For example, in an Internet of Things project with high technical requirements, the weight of technical certification may be increased to 0.7.

[0174] This flexible weight setting improves the adaptability of the evaluation and ensures that the score reflects the project requirements.

[0175] Step S5-4, for the preliminary fusion score, a preset fusion weight distribution is adopted to perform a weighted fusion calculation to obtain an adjusted fusion score. The fusion weight distribution is further optimized based on the business scenario to ensure that the score more accurately reflects the comprehensive ability of the enterprise. Illustratively, the preliminary fusion score 0.85 is adjusted by the weight technology certification 0.7 and the financial data quality 0.3 to generate an adjusted fusion score 0.87. The adjusted score is more in line with the technical priority requirements of the Internet of Things device procurement project.

[0176] In an embodiment, the fusion weight distribution is determined by analyzing successful cases of historical bidding projects. For example, if historical data shows that technical ability has a greater impact on the bid-winning rate, the weight of technical certification is increased. The adjusted fusion score is generated by weighted calculation, for example: score = 0.7 x technical certification value + 0.3 x financial data quality value. This method ensures the relevance of the score while retaining the auxiliary role of financial data. The adjusted fusion score provides a more accurate input for subsequent credit rating judgments.

[0177] Step S5-5, according to the adjusted fusion score, a logistic regression model is applied to judge the credit rating of the enterprise and output a dynamic verification score. The logistic regression model predicts the credit rating of the enterprise, such as high, medium, and low, by analyzing the fusion score.

[0178] Illustratively, if the adjusted fusion score is 0.87, the logistic regression model predicts a high credit rating and outputs a dynamic verification score of 0.87, which comprehensively reflects the qualifications and financial ability of the enterprise.

[0179] Specifically, the logistic regression model learns the mapping relationship between the fusion score and the credit rating by training historical data.

[0180] For example, the model may determine that a score of 0.8 or above corresponds to a high credit rating based on a large amount of historical bidding data. It should be noted that the model can also combine other features, such as the industry experience or registered capital of the enterprise, to further improve the accuracy of the prediction. The dynamic verification score, as the core output of the evaluation, provides a direct basis for the subsequent approval process.

[0181] Step S5-6, if the dynamic verification score is lower than the preset threshold, the fusion weight distribution is adjusted through a score calculation model to recalculate the dynamic verification score. The score calculation model optimizes the weights through iteration to maximize the stability of the score. Illustratively, if the dynamic verification score of an enterprise is 0.6, which is lower than the threshold of 0.8, the weight is adjusted, for example, the weight of financial data quality is increased from 0.3 to 0.4, and the score is recalculated to 0.82.

[0182] In one possible implementation, the score calculation model optimizes the weights through gradient descent, seeking a combination of weights that maximizes the score. If the recalculated score is still below the threshold, the system may require the enterprise to supplement data, such as updating financial statements or providing additional qualifications. This approach ensures the fairness of the evaluation while providing the enterprise with an opportunity for improvement. The recalculated dynamic verification score improves the reliability of the evaluation results.

[0183] Step S57, obtain the recalculated dynamic verification score, output the evaluation results, and generate the comprehensive evaluation score of the bidding enterprise. The comprehensive evaluation score integrates the results of qualifications, performance, and financial data, and outputs in a structured form. For example, the comprehensive evaluation score of an enterprise is 0.82, recorded as Enterprise A: Comprehensive Evaluation Score 0.82; Technical Capability: 0.9; Financial Health: 0.8. This score provides a comprehensive evaluation basis for the tendering party.

[0184] Specifically, the output of the comprehensive evaluation score includes the score and evaluation details, such as the specific values of the technical capability, financial health, and other dimensions. The output results are stored through the blockchain to ensure transparency and non-tamperability. It should be noted that the evaluation results can also generate a visual report, such as a column chart showing the scores of each dimension, to facilitate quick review by the tendering party. This structured output improves the practicality and operability of the evaluation.

[0185] Step S6, if the dynamic verification score is higher than the preset threshold, execute the automatic approval logic through the smart contract, store the verification process records on the chain using a hash function, and obtain the non-tamperable audit log. The dynamic verification score reflects the comprehensive competitiveness of the enterprise, and if it is higher than the threshold, such as 0.8, the automatic approval logic of the smart contract is triggered.

[0186] For example, the dynamic verification score of an enterprise is 0.82, and the smart contract automatically approves its bidding qualification and records the verification process.

[0187] Step S6-1, if the verification score is higher than the preset threshold, obtain the data to be approved through the trigger logic of the smart contract, execute the automatic approval operation, and generate the approval result. The smart contract extracts the enterprise data to be approved, such as the qualification vector and financial data, based on the score threshold, and generates the approval result. For example, the score of an enterprise is 0.82, and the smart contract generates the result Enterprise A: Approved.

[0188] In one embodiment, the smart contract is implemented through a conditional trigger mechanism, such as approval if the score > 0.8. The trigger logic checks whether the score meets the condition, and if it does, it extracts the relevant data and generates the approval result. The approval result is stored on the chain to ensure transparency and traceability. This automated approval process improves efficiency and reduces human intervention.

[0189] Step S6-2, the approval result and the verification record are encrypted using a hash function to generate encrypted data. The hash function converts the approval result and the verification record into a fixed-length hash value, ensuring the uniqueness and security of the data. For example, the approval result of enterprise A: approval and the verification record generate a hash value of abc123.

[0190] Specifically, the hash function uses the SHA-256 algorithm, and the input data includes the approval result, the verification score, and the timestamp, etc. The generated hash value is fixed in length, for example, 256 bits, suitable for on-chain storage.

[0191] It should be noted that the one-way nature of the hash function ensures that the data cannot be tampered with, providing a reliable guarantee for subsequent audits. The encrypted data prepares a secure input for on-chain storage.

[0192] Step S6-3, the encrypted data is written to the on-chain storage through the blockchain network to generate a storage address. The encrypted data is written to the blockchain through a distributed ledger, generating a unique storage address, such as 0x1234; the writing process is completed through a consensus mechanism, ensuring that all nodes agree on the data.

[0193] In one possible implementation, the writing process uses a proof-of-work or proof-of-stake mechanism to ensure data consistency. The storage address is recorded in hexadecimal form for easy retrieval in the future. For example, after the encrypted data of a certain enterprise is written, the address 0x1234 is generated and recorded in the log of the smart contract.

[0194] This method ensures the distributed storage and traceability of data.

[0195] Step S6-4, according to the storage address, the on-chain data is obtained, the data integrity is verified, and the verification status is obtained. The verification process compares the hash value of the on-chain data with the original encrypted data to determine whether the data is consistent. For example, if the data hash value in the storage address 0x1234 matches the original hash value abc123, the verification status is complete.

[0196] Specifically, the verification process is implemented through the query interface of the blockchain node, which extracts the data in the storage address and calculates its hash value. If the hash values are consistent, it indicates that the data has not been tampered with. It should be noted that the verification can also be combined with a timestamp check to ensure the time validity of the data. The verification status provides a reliable basis for the generation of subsequent audit logs.

[0197] Step S6-5, if the verification status is complete, an audit log is generated through a smart contract to record the approval and storage information. The audit log is stored in a structured form, containing information such as approval result, timestamp, storage address, etc. For example, the audit log records enterprise A: approval; time: 20XX-0X-0X; storage address: 0x1234.

[0198] In one embodiment, the audit log is generated through the log function of the smart contract and stored in the public ledger of the blockchain. The log content supports multi-party query, for example, the tenderer can retrieve the approval record through the storage address. This method improves the transparency of the evaluation process and facilitates supervision and review.

[0199] Step S6-6, according to the audit log content, the decentralized network distributes the log data, and generates a distribution confirmation. The log data is distributed to multiple nodes through the blockchain network to ensure wide dissemination of data. Illustratively, after the distribution is completed, a confirmation message is generated to confirm that the distribution is complete: 10 nodes.

[0200] Specifically, the distribution process is implemented through a peer-to-peer network, and the log data is transmitted to each node in an encrypted form. It should be noted that the distribution confirmation is verified by node signature to ensure that all nodes receive consistent data. The distribution confirmation enhances the credibility and accessibility of the audit log and provides a basis for subsequent consensus feedback.

[0201] Step S6-7, through the distribution confirmation, obtain the consensus feedback of the network nodes, and determine the final on-chain record state. The consensus feedback is verified by a majority of nodes to confirm the consistency of the log data. Illustratively, if 8 out of 10 nodes confirm that the data is consistent, the record state is confirmed.

[0202] In one possible implementation, the consensus feedback uses a Byzantine fault-tolerant algorithm to tolerate partial node failures or malicious behavior. The final record state is stored in a structured form, for example, Enterprise A: Record state confirmed; Node number: 8.

[0203] This method ensures the reliability and stability of the on-chain record.

[0204] Step S7, according to the tamper-proof audit log, retrieve the relevant verification path from the blockchain network, judge the integrity of the path, and if it is complete, determine the final output of the review result.

[0205] The audit log is stored in the blockchain and contains complete records of approval and verification, providing a reliable basis for review. Illustratively, the audit log of a certain enterprise contains the approval time: 20XX-0X-0X; Hash value: abc123; Storage address: 0x1234.

[0206] Step S7-1, obtain the audit log from the blockchain network using a distributed node query protocol, extract a log dataset containing timestamps and transaction hashes, and obtain the original audit log; the distributed query protocol ensures the integrity and consistency of the log data through multi-node collaboration. For example, the query protocol extracts logs from 10 nodes to generate a dataset: Enterprise A: approved; time: 20XX-0X-0X; hash: abc123.

[0207] Specifically, the query protocol verifies the authenticity of the log data by verifying the digital signature of the nodes. The extracted log dataset contains fields such as timestamps, transaction hashes, and approval results, and is stored in a structured format. It should be noted that the query process may involve multiple rounds of communication to ensure that all nodes return consistent data. The original audit log provides complete basic data for subsequent path analysis.

[0208] Step S7-2, for the original audit log, parse the log structure, extract the node sequence and hash chain in the verification path, and generate a path dataset. The log structure usually contains multiple records, each corresponding to a verification step. For example, after log parsing, the path nodes A-node B-node C are extracted, and the hash chain is abc123-def456-ghi789.

[0209] In one embodiment, the parsing process extracts key fields such as timestamps, hash values, and node identifiers using regular expressions.

[0210] The generated path dataset is stored in a list format, such as path: [node A, abc123]-[node B, def456]; this structured parsing improves data processing efficiency and provides clear input for path integrity verification.

[0211] Step S7-3, if the node sequence in the path dataset is consistent with the preset blockchain topology, use the Merkle tree verification algorithm to calculate the hash value of each node in the path, judge the path integrity, and obtain the integrity verification result. The Merkle tree algorithm verifies the integrity of the path by calculating the hash value layer by layer.

[0212] For example, if the hash chain of the path nodes A-node B-node C matches the root hash stored on the chain, the verification result is complete.

[0213] Specifically, the Merkle tree algorithm first pairs the hash values in the path two by two, calculates the upper layer hash, and generates the root hash.

[0214] Then, compare the root hash with the root hash stored on the chain. If they are consistent, the path is complete.

[0215] It should be noted that the algorithm can also detect missing or tampering in the path, for example, if the node hash value does not match, the verification result is "incomplete". The integrity verification result ensures the credibility of the log data.

[0216] Step S7-4, according to the integrity verification result, if the integrity verification is passed, the transaction timestamp and associated metadata are extracted from the path data set to generate the review data set. The review data set integrates the key information in the path, such as timestamp, hash value and approval result. Exemplarily, the review data set records enterprise A: time 20XX-0X-0X; hash: abc123; approval.

[0217] In one possible implementation, the review data set is stored in a structured format, such as JSON format, to facilitate subsequent analysis and comparison. The extraction process ensures that all fields are complete, such as the timestamp format "YYYY-MM-DD HH:MM:SS"; the review data set provides a direct input for the generation of the final review result.

[0218] Step S7-5, the SHA-256 algorithm is used to calculate the hash value of the review data set to generate the review hash value and determine the uniqueness of the review result.

[0219] The SHA-256 algorithm converts the review data set into a fixed length hash value, ensuring the uniqueness and irreversibility of the result. Exemplarily, the review data set generates a hash value xyz789.

[0220] Specifically, the SHA-256 algorithm inputs all fields of the review data set, such as timestamp, approval result and path information, to generate a 256-bit hash value.

[0221] It should be noted that the uniqueness of the hash value ensures the verifiability of the review result, and any data tampering will cause the hash value to change. The review hash value provides a reliable identification for the final review.

[0222] Step S7-6, through the network communication protocol, the review hash value is compared with the preset audit standard to judge whether the review result meets the standard, and the final review result is output.

[0223] The audit standard includes the requirements of hash value consistency, timestamp validity, etc. Exemplarily, if the review hash value xyz789 is consistent with the hash value stored on the chain, and the timestamp is within the valid range, the review is passed.

[0224] In one embodiment, the network communication protocol transmits hash values ​​through an encrypted channel to ensure the security of the comparison process. The comparison process checks whether the hash values ​​match and whether the timestamps meet the bid deadline requirements. The final review result is output in a structured form, for example, Company A: Review Passed; Hash: xyz789.

[0225] This approach ensures the transparency and credibility of the review process, providing a final conclusion for the tender evaluation.

[0226] In one possible implementation, the review results can be automatically notified to relevant parties, such as the tendering party and bidding companies, via smart contracts.

[0227] The notification includes the review result, hash value, and timestamp, stored on-chain for easy verification by multiple parties. For example, after a company's review is approved, the smart contract generates a notification to Company A: Review Approved; Time: 20XX-0X-0X, which is sent via the on-chain messaging system.

[0228] This automated notification improves the efficiency and transparency of the evaluation process.

[0229] Specifically, the output of the review results can also be combined with visualization tools, such as generating review reports that display hash values, timestamps, and approval details; the reports are presented in the form of charts, such as timelines that show the node sequence of the verification path, making it easier for the tendering party to quickly understand the review process.

[0230] It should be noted that the on-chain storage of the review results ensures their immutability, and any party can retrieve the original data for verification via the storage address. This method significantly improves the credibility and operability of the assessment.

[0231] In one embodiment, the review results can also be correlated with historical bidding data to generate a statistical analysis report. For example, analyzing a company's review pass rate in multiple bidding processes can assess its stability and competitiveness. The report might show Company A: review pass rate 90%; average score 0.85, providing decision-making reference for the bidding party. This correlation analysis further enhances the comprehensiveness of the evaluation and optimizes the bidding process.

[0232] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An engineering bidding platform management method, characterized in that, The method comprises the following steps: obtaining a demand description text from a bidding project chain through a blockchain node, extracting a keyword sequence by using natural language processing, and obtaining a bidding demand feature vector; vector conversion processing is performed on the qualification certificate data submitted by the bidding enterprise according to the bidding demand feature vector, the similarity between the converted vector and the demand feature vector is judged, and if the similarity exceeds a preset threshold, it is determined that the qualification is preliminarily matched; performance index data is extracted from the historical performance records on the bidding enterprise chain, and a comprehensive qualification vector is obtained by using a vector embedding method to fuse the performance index data and the preliminary qualification matching result; The method for extracting performance index data from the historical performance records on the bidding enterprise chain, and obtaining a comprehensive qualification vector by using a vector embedding method to fuse the performance index data and the preliminary qualification matching result, comprises: extracting performance index data from the historical performance records on the bidding enterprise chain, and obtaining a structured performance index data set by using a preset index extraction rule; by using a weighted average method, a weighted performance index set is generated; from the preliminary qualification matching result, the enterprise qualification data is obtained, the weighted performance index set and the enterprise qualification data are fused by using a vector embedding method, and an initial comprehensive qualification vector is generated; if the dimension of the initial comprehensive qualification vector exceeds a preset threshold, principal component analysis is used for dimension reduction processing to obtain an optimized comprehensive qualification vector; according to the optimized comprehensive qualification vector, the similarity between vectors is calculated, and the qualification matching degree between enterprises is determined by using a cosine similarity algorithm; by using the qualification matching degree between enterprises, the qualification grade of the enterprise is judged by combining a preset classification rule, and a final qualification evaluation result is generated; according to the final qualification evaluation result, a data correlation analysis method is used to associate the evaluation result with the historical performance records on the chain to generate a verification data set of the comprehensive qualification vector; by calling an external oracle interface through a smart contract, real-time financial statement updates of the bidding enterprise are obtained, and it is judged whether the financial statement update time is before the bidding deadline, and if it is before the bidding deadline, the financial data validity is determined; according to the comprehensive qualification vector and the financial data validity result, weighted fusion calculation is performed to obtain a dynamic verification score of the bidding enterprise; The method for obtaining a dynamic verification score of a bidding enterprise by performing weighted fusion calculation according to the comprehensive qualification vector and the financial data validity result comprises: obtaining the bidding enterprise information and the financial data validity, judging the data integrity and consistency by using a preset data verification rule, and obtaining the verified financial data quality; according to the verified financial data quality, the qualification evaluation index is extracted, and the comprehensive qualification vector is determined by using a factor analysis method; if at least one index in the comprehensive qualification vector exceeds a preset threshold, the qualification evaluation index and the financial data quality are fused by using a linear weighting method to obtain a preliminary fusion score; for the preliminary fusion score, a preset fusion weight distribution is used to perform weighted fusion calculation to obtain an adjusted fusion score; According to the adjusted fusion score, a logistic regression model is applied to judge the enterprise credit level and output a dynamic verification score; If the dynamic verification score is lower than a preset threshold, a score calculation model is used to adjust the fusion weight distribution, and the dynamic verification score is recalculated; The recalculated dynamic verification score is obtained through the evaluation result output to generate a comprehensive evaluation score of the bidding enterprise; If the dynamic verification score is higher than the preset threshold, an automatic approval logic is executed through a smart contract, a hash function is used to store the verification process record on the chain to obtain an unforgeable audit log; According to the unforgeable audit log, the relevant verification path is retrieved from the blockchain network, the path integrity is judged, and if the path is complete, the final output of the review result is determined; The step of retrieving the relevant verification path from the blockchain network according to the unforgeable audit log, judging the path integrity, and determining the final output of the review result if the path is complete, comprises: Obtain the audit log from the blockchain network, extract the log data set containing the timestamp and transaction hash using a distributed node query protocol, and obtain the original audit log; For the original audit log, parse the log structure, extract the node sequence and hash chain in the verification path, and generate a path data set; If the node sequence in the path data set is consistent with the preset blockchain topology structure, use the Merkle tree verification algorithm to calculate the hash value of each node in the path, judge the path integrity, and obtain the integrity verification result; According to the integrity verification result, if the integrity verification is passed, extract the transaction timestamp and associated metadata from the path data set to generate a review data set; Use the SHA-256 algorithm to calculate the hash of the review data set to generate a review hash value and determine the uniqueness of the review result; Compare the review hash value with the preset audit standard through a network communication protocol to determine whether the review result meets the standard and output the final review result.

2. The engineering bidding platform management method of claim 1, wherein: The step of obtaining the demand description text from the blockchain node through the blockchain node, extracting the keyword sequence using natural language processing, and obtaining the bidding demand feature vector, comprises: Obtain the chain-stored bidding project demand description text from the blockchain node, and obtain the original text data using a distributed query protocol; Use the word segmentation technology in natural language processing to perform word segmentation processing on the original text data to obtain a word sequence; If the word sequence contains a preset stop word, filter the stop word to obtain a filtered keyword sequence; Convert the keyword sequence to a word vector sequence through a pre-trained word embedding model to generate an initial feature vector; According to the initial feature vector, use principal component analysis algorithm to reduce the dimension of the vector to obtain an optimized bidding demand feature vector; If the similarity between the optimized feature vector and the preset business demand template exceeds a preset threshold, it is determined that the feature vector meets the business demand, and a structured demand description is generated; Through the structured demand description, a clustering algorithm is used to classify the bidding project to obtain a classification result.

3. The engineering bidding platform management method of claim 1, wherein: The step of judging the similarity of the converted vector and the demand feature vector after the vector conversion processing of the qualification certificate data submitted by the bidding enterprise according to the bidding demand feature vector, and determining the preliminary matching of the qualification if the similarity exceeds the preset threshold value, comprises: Obtaining bidding demand data and extracting a feature vector therefrom, generating a demand feature vector by using a keyword extraction and semantic analysis method; Structuring the qualification certificate data submitted by the bidding enterprise, extracting key fields by using a text parsing technology, and generating qualification data; Converting the qualification data into a qualification feature vector by using a pre-trained word vector model; Calculating the cosine similarity of the qualification feature vector and the demand feature vector to obtain a similarity value; If the similarity value exceeds the preset threshold value, determining the preliminary matching of the qualification and generating a matching result; For the matching result, using a clustering analysis method to group the bidding enterprises to obtain a grouping result; According to the grouping result, extracting the qualification feature vector of each group of bidding enterprises to generate a ranking list.

4. The engineering bidding platform management method of claim 1, wherein: The step of judging whether the financial statement update time is before the bidding deadline by calling an external oracle interface through a smart contract to obtain real-time financial statements of the bidding enterprise, and determining the validity of the financial data if it is before the bidding deadline, comprises: Obtaining the financial statement update timestamp of the bidding enterprise from an external data source through an oracle interface and storing it as first time data; Comparing the first time data with the preset bidding deadline by using time comparison logic, and if the first time data is earlier than the bidding deadline, determining that the financial statement update time is valid to obtain a validity determination result; Obtaining the validity determination result through a smart contract interface to generate a financial data validity state; According to the validity state, calling the oracle interface to obtain the complete data set of the financial statement and storing it as first financial data; Verifying the integrity and consistency of the first financial data by using a data verification algorithm to obtain second financial data that passes the verification; Using a classification algorithm to perform a financial health assessment on the second financial data to generate a financial health score; According to the financial health score and the validity determination result, obtaining the final validity of the financial data of the bidding enterprise.

5. The engineering bidding platform management method of claim 1, wherein: The step of passing the smart contract to execute automatic approval logic if the dynamic verification score is higher than the preset threshold value, and using a hash function to store the verification process record on the chain to obtain an unforgeable audit log, comprises: If the verification score is higher than the preset threshold value, obtaining the data to be approved by triggering logic of the smart contract, performing an automatic approval operation to obtain an approval result; Using a hash function to encrypt the approval result and the verification record to generate encrypted data; Writing the encrypted data into on-chain storage through a blockchain network to generate a storage address; Obtaining the on-chain data according to the storage address, verifying the data integrity, and obtaining a verification state; If the verification state is complete, generating an audit log through a smart contract to record the approval and storage information; According to the content of the audit log, distributing the log data by using a decentralized network to generate a distribution confirmation; Obtaining the consensus feedback of the network nodes through the distribution confirmation to obtain the final on-chain record state.

6. An engineering tendering platform management system characterized by, Comprise: The demand feature extraction module is configured to acquire demand description text from the tender project chain by a blockchain node, extract a keyword sequence by natural language processing, and obtain a tender demand feature vector. The qualification matching module is configured to perform vector conversion processing on qualification certificate data submitted by a bidding enterprise according to the tender demand feature vector, judge the similarity between the converted vector and the demand feature vector, and determine preliminary qualification matching if the similarity exceeds a preset threshold. The performance fusion module is configured to extract performance index data from historical performance records on a bidding enterprise chain, fuse the performance index data and the preliminary qualification matching result by a vector embedding method, and obtain a comprehensive qualification vector. The performance index data is extracted from the historical performance records on the bidding enterprise chain, the performance index data and the preliminary qualification matching result are fused by the vector embedding method, and the comprehensive qualification vector is obtained, including: The performance index data is extracted from the historical performance records on the bidding enterprise chain, the performance index data and the preliminary qualification matching result are fused by the vector embedding method, and the comprehensive qualification vector is obtained, including: The performance index data is extracted from the historical performance records on the bidding enterprise chain, the performance index data and the preliminary qualification matching result are fused by the vector embedding method, and the comprehensive qualification vector is obtained, including: If the dimension of the initial comprehensive qualification vector exceeds a preset threshold, the principal component analysis method is used for dimension reduction processing to obtain an optimized comprehensive qualification vector. The similarity between vectors is calculated according to the optimized comprehensive qualification vector, and the qualification matching degree between enterprises is determined by using the cosine similarity algorithm. The qualification matching degree between enterprises is used to determine the qualification grade of the enterprise according to a preset classification rule, and a final qualification evaluation result is generated. According to the final qualification evaluation result, a data correlation analysis method is used to associate the evaluation result with the historical performance records on the chain to generate a verification data set of the comprehensive qualification vector. The financial verification module is configured to obtain real-time financial statement updates of a bidding enterprise by calling an external oracle interface through a smart contract, judge whether the financial statement update time is before the bidding deadline, and determine the validity of the financial data if the bidding deadline is before the bidding deadline. The scoring calculation module is configured to perform weighted fusion calculation according to the comprehensive qualification vector and the financial data validity result to obtain a dynamic verification score of the bidding enterprise. The scoring calculation module is configured to perform weighted fusion calculation according to the comprehensive qualification vector and the financial data validity result to obtain a dynamic verification score of the bidding enterprise, including: The bidding enterprise information and the financial data validity are obtained, the data integrity and consistency are judged by a preset data verification rule, and the verified financial data quality is obtained. According to the verified financial data quality, the qualification evaluation index is extracted, and the comprehensive qualification vector is determined by using the factor analysis method. If at least one index in the comprehensive qualification vector exceeds a preset threshold, the qualification evaluation index and the financial data quality are fused by a linear weighting method to obtain a preliminary fusion score. ​ For the preliminary fusion score, a preset fusion weight distribution is adopted to perform a weighted fusion calculation to obtain an adjusted fusion score; According to the adjusted fusion score, a logistic regression model is applied to judge the enterprise credit level and output a dynamic verification score; If the dynamic verification score is lower than a preset threshold, the fusion weight distribution is adjusted through a score calculation model to recalculate the dynamic verification score; The recalculated dynamic verification score is obtained through an evaluation result output to generate a comprehensive evaluation score of the bidding enterprise; If the dynamic verification score is higher than the preset threshold, a smart contract execution module is used to automatically approve the logic, a hash function is used to store the verification process record on the chain to obtain an unforgeable audit log; An audit verification module is used to retrieve the relevant verification path from the blockchain network according to the unforgeable audit log, judge the path integrity, and if the path is complete, determine the final output of the review result; The step of retrieving the relevant verification path from the blockchain network according to the unforgeable audit log, judging the path integrity, and if the path is complete, determining the final output of the review result, includes: Obtain the audit log from the blockchain network, use a distributed node query protocol to extract a log data set containing a timestamp and a transaction hash to obtain an original audit log; For the original audit log, parse the log structure, extract the node sequence and hash chain in the verification path, and generate a path data set; If the node sequence in the path data set is consistent with the preset blockchain topology structure, use the Merkle tree verification algorithm to calculate the hash value of each node in the path, judge the path integrity, and obtain the integrity verification result; According to the integrity verification result, if the integrity verification is passed, extract the transaction timestamp and associated metadata from the path data set to generate a review data set; Use the SHA-256 algorithm to perform hash calculation on the review data set to generate a review hash value and determine the uniqueness of the review result; Through a network communication protocol, compare the review hash value with the preset audit standard to judge whether the review result meets the standard and output the final review result.

Images