Enterprise manufacturing capability quantitative evaluation method and system based on multi-source heterogeneous data fusion, terminal equipment and medium

Abstract

The invention discloses an enterprise manufacturing capability quantitative evaluation method and system based on multi-source heterogeneous data fusion, terminal equipment and a medium, and relates to the technical field of enterprise business quantitative evaluation. The method comprises the steps of obtaining multi-source heterogeneous multi-dimensional operation data of a target enterprise, constructing a multi-dimensional feature set containing at least three of features of resource consumption, a supply chain, human resources, material logic and an asset structure, and realizing quantitative evaluation of enterprise manufacturing capability through feature fusion. According to the method, the defects that an existing evaluation method is high in cost, easy to forge and lack of cross validation are overcome, non-intrusive data collection and cross-dimensional verification logic are adopted, low-cost, automatic and high-accuracy quantification of the real manufacturing capacity of an enterprise is achieved, the information asymmetry risk that a trader pretends to be a manufacturer is effectively avoided, and the real manufacturing capacity of the enterprise is evaluated. The evaluation result can be called by an upper-layer service system, and reliable assistance is provided for related service decisions.

Description

Technical Field

[0001] This invention relates to the field of enterprise business quantitative assessment technology, and in particular to a method, system, terminal equipment and medium for quantitative assessment of enterprise manufacturing capabilities based on multi-source heterogeneous data fusion. Background Technology

[0002] In scenarios such as business collaborations and supply chain finance, it is necessary to accurately assess a company's true manufacturing capabilities and mitigate the associated risks.

[0003] Existing technologies primarily employ three methods for assessing a company's true manufacturing capabilities: first, on-site factory inspections, which are costly, inefficient, prone to rent-seeking, and lack real-time data updates; second, verification based on business registration information, where traders can easily alter their business scope, making it difficult to identify shell companies and verify actual production capacity; and third, single-dimensional data verification, relying solely on single indicators such as electricity consumption, which is easily falsified and lacks cross-verification of cash flow and business flow. None of these technologies can achieve low-cost, accurate, and automated assessment of a company's manufacturing capabilities, leading to a significant information asymmetry problem where traders impersonate manufacturers.

[0004] Therefore, there is an urgent need for a quantitative assessment method for enterprise production capacity that is easy to obtain, efficient, and resistant to fraud, in order to fill the gap in existing technologies. Summary of the Invention

[0005] The technical problem this invention aims to solve is that, in the field of quantitative assessment of enterprise business operations, existing technologies struggle to balance low cost, automation, and assessment accuracy. They are unable to effectively verify a company's true manufacturing capabilities and are easily exploited by traders who falsify data to impersonate manufacturers, leading to information asymmetry risks. Therefore, an effective solution is urgently needed to address these technical problems.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: Firstly, this invention provides a method for quantitatively evaluating enterprise manufacturing capabilities based on the fusion of multi-source heterogeneous data. Obtain multi-dimensional operational data of the target company; Based on the aforementioned multi-dimensional operational data, a multi-dimensional feature set is constructed; wherein, the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets. By fusing the features of each dimension in the multidimensional feature set, a quantitative assessment result of the target enterprise's manufacturing capability is obtained.

[0007] In one implementation, the target enterprise's multi-dimensional operating data includes cash flow data, transaction details, human resource data, input and output data, and enterprise asset data. The acquisition of the target enterprise's multi-dimensional operating data includes: Structured data obtained from at least one of the following sources: authorized bank accounts, tax systems, social security systems, electricity metering systems, or asset management systems of the enterprise; or unstructured data obtained from at least one of the following sources: scanned copies of financial statements or invoices provided by the enterprise.

[0008] In one implementation, the resource consumption characteristics include energy consumption characteristics. Constructing energy consumption characteristics based on the multi-dimensional operational data includes: Identify energy cost expenditures within the cash flow data of the multi-dimensional operational data; Obtain the energy price parameters and transaction time period corresponding to the energy cost expenditure; The energy consumption level of the target enterprise is calculated based on the energy cost expenditure, the energy price parameters, and the transaction time period.

[0009] In one implementation, the supply chain characteristics include supply chain attribute characteristics. The supply chain attribute characteristics are constructed based on the multi-dimensional operational data, including: The transaction details, which include counterparty information and transaction content, are obtained from the multi-dimensional operational data. Semantic analysis is then used to obtain the counterparty business attributes and transaction amount data from the transaction details. The concentration of raw material procurement is calculated using the transaction amount data. Based on the business attributes of the counterparty and the concentration of raw material procurement, the supply chain attributes of the target company are analyzed.

[0010] In one implementation, the human resource features include personnel structure features. Constructing personnel structure features based on the multi-dimensional operational data includes: Based on the human resources data in the multi-dimensional operational data, the distribution of personnel and the average salary level per person are calculated. Based on the personnel distribution, average salary level, and minimum wage standard, the personnel structure characteristics are determined, which reflect the degree of matching between the target enterprise's labor structure and that of a manufacturing enterprise.

[0011] In one implementation, the material logic class feature includes a material conversion consistency feature. The material conversion consistency feature is constructed based on the multi-dimensional operational data, including: Obtain the input and output data from the multi-dimensional operating data; The input and output data are mapped to the industrial BOM knowledge graph to calculate the reachability probability of the production conversion path between input materials and output products, and to evaluate the consistency of material conversion.

[0012] In one implementation, the asset structure features include fixed asset structure features. The fixed asset structure features are constructed based on the multi-dimensional operating data, including: Obtain enterprise asset data from the multi-dimensional operational data; Extract the net value of fixed assets and the total assets from the enterprise's asset data; Based on the net value of fixed assets and the total assets, the fixed asset ratio is calculated, and the fixed asset structure is analyzed.

[0013] In one implementation, the quantitative evaluation result is a manufacturing capability index, and the process of fusing the features of each dimension in the multidimensional feature set to obtain the quantitative evaluation result of the target enterprise's manufacturing capability includes: Based on the industry category of the target enterprise, obtain the weight parameters of each feature dimension that are adapted to the production characteristics of the industry; Based on the weight parameters, the quantitative representation values ​​of the features in each dimension are fused to obtain the manufacturing capability index of the target enterprise; wherein, the fusion includes at least one of logistic regression model fusion or weighted scoring fusion.

[0014] In one implementation, obtaining the weight parameters of each feature dimension adapted to the production characteristics of the target enterprise based on the industry category includes: Obtain a historical labeled sample set of the industry to which the target enterprise belongs. The historical labeled sample set contains several enterprise samples with known real manufacturing capability labels. Each sample corresponds to the feature quantification value of each dimension in the multidimensional feature set. The historical labeled sample set is trained using a supervised machine learning algorithm to obtain the weight parameters of each feature dimension of the industry to which the target enterprise belongs.

[0015] In one implementation, the method further includes: At least two different types of features are selected from the multidimensional feature set, and the consistency of the selected different types of features is verified in at least one of the following dimensions: time dimension, production capacity logic dimension, or material conversion dimension. Based on the results of the consistency verification, the quantitative assessment results of the target company's manufacturing capabilities are supplemented with auxiliary annotations, which are used to characterize the risk of data falsification.

[0016] Secondly, embodiments of the present invention also provide a quantitative evaluation system for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, the system comprising: The multi-dimensional operational data acquisition module is used to acquire multi-dimensional operational data of the target enterprise. A multi-dimensional feature set construction module is used to construct a multi-dimensional feature set based on the multi-dimensional operational data; wherein, the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets. The enterprise manufacturing capability quantitative assessment module is used to fuse the features of each dimension in the multidimensional feature set to obtain the quantitative assessment result of the target enterprise's manufacturing capability.

[0017] Thirdly, embodiments of the present invention also provide a terminal device, the terminal device including a memory, a processor, and an enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion stored in the memory and executable on the processor. When the processor executes the enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion, it implements the steps of the enterprise manufacturing capability quantification evaluation method based on multi-source heterogeneous data fusion as described in any of the above schemes.

[0018] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing an enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion. When the enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion is executed by a processor, it implements the steps of the enterprise manufacturing capability quantification evaluation method based on multi-source heterogeneous data fusion as described in any of the above schemes.

[0019] Beneficial Effects: This invention discloses a method, system, terminal equipment, and medium for quantitative assessment of enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, relating to the field of enterprise business quantitative assessment technology. The method first acquires multi-dimensional operational data of the target enterprise. Then, based on the multi-dimensional operational data, a multi-dimensional feature set is constructed; wherein the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption; the supply chain features reflect the upstream and downstream attributes of the supply chain; the human resource features reflect the human resource structure; the material logic features reflect the material conversion logic; and the asset structure features reflect the proportion of productive assets. Finally, the features of each dimension in the multi-dimensional feature set are fused to obtain the quantitative assessment result of the target enterprise's manufacturing capabilities. This invention achieves non-intrusive assessment through multi-source heterogeneous data fusion and multi-dimensional feature construction, requiring no physical access to production equipment. Its fusion of multi-dimensional core features, combined with cross-class consistency verification to prevent forgery, solves the problem of easy forgery of single-dimensional data, improving the credibility of the assessment. It simultaneously enables automated quantitative assessment, reducing manual review costs and rent-seeking risks, and improving assessment efficiency. It can accurately distinguish between manufacturing and trading companies, resolving information asymmetry issues and providing reliable data for supply chain finance, supplier certification, and other scenarios. Attached Figure Description

[0020] Figure 1 A flowchart illustrating a specific implementation of the enterprise manufacturing capability quantitative assessment method based on multi-source heterogeneous data fusion provided in this embodiment of the invention.

[0021] Figure 2 This is a data flow diagram illustrating the quantitative evaluation method for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion provided in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram illustrating the construction of energy consumption characteristics for a quantitative assessment method of enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, as provided in an embodiment of the present invention.

[0023] Figure 4 This is a schematic diagram illustrating the construction of supply chain attribute features for a quantitative evaluation method of enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, as provided in an embodiment of the present invention.

[0024] Figure 5 This is a schematic diagram of the fixed asset structure construction of the enterprise manufacturing capability quantitative assessment method based on multi-source heterogeneous data fusion provided in an embodiment of the present invention.

[0025] Figure 6 This is a schematic diagram of the principle of the enterprise manufacturing capability quantitative assessment device based on multi-source heterogeneous data fusion provided in the embodiments of the present invention.

[0026] Figure 7 This is a block diagram illustrating the internal structure of the terminal device provided in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0028] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content, operations, or steps, nor does it require execution in the described order. For example, some operations or steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0029] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0030] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. For example, the first control information and the second control information are only used to distinguish different control information and do not limit their order.

[0031] Those skilled in the art will understand that the words "first" and "second" do not limit the quantity or the order of execution, and that the words "first" and "second" do not necessarily imply that they are different.

[0032] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0033] In the technological fields related to industrial production and enterprise collaboration, identifying and quantifying the target enterprise's actual manufacturing capabilities is a crucial prerequisite for ensuring the reliability of related technology applications. With the development of industrial production models, collaboration between enterprises is becoming increasingly frequent; therefore, controlling the actual manufacturing capabilities of partners can support the security and effectiveness of technology applications.

[0034] However, existing technologies used to assess a company's manufacturing capabilities all have technical flaws.

[0035] The first existing technology involves on-site factory inspections, where staff physically visit the target company's production site to verify production equipment, processes, and scale to determine if the company possesses the necessary manufacturing capabilities. However, this method requires significant human and material resources, resulting in high implementation costs. Furthermore, on-site inspections are inefficient and cannot quickly respond to demands for manufacturing capability assessments. They also rely on subjective judgment, increasing the risk of human intervention and potentially distorting the assessment results. Additionally, the data obtained from on-site inspections is static data at a specific point in time, lacking real-time updates and failing to reflect dynamic changes in the company's manufacturing capabilities.

[0036] The second existing technology uses a verification method based on business registration information. This involves indirectly inferring a company's manufacturing capabilities by querying its business registration information, such as its business scope and registered capital. However, from a technical perspective, modifying business registration information is relatively easy; companies can readily change their business scope and other related information, leading to insufficient credibility of this information. Furthermore, business registration information only reflects the company's registration attributes and cannot effectively identify shell companies or verify the company's actual production capacity, making the assessment results based on this technology lack accuracy and reliability.

[0037] The third existing technology uses single-dimensional data verification for assessment, but it only selects one type of production-related data, such as electricity consumption or the quantity of a single material purchased, as the basis for evaluating a company's manufacturing capabilities. Therefore, this method has significant technical flaws: single-dimensional data is easily falsified, for example, by operating equipment with no actual production demand to inflate electricity consumption data, thus passing off false data as real production data. Furthermore, because this technology relies on only a single type of data and lacks cross-verification with multiple types of data, such as cash flow and business flow, it cannot comprehensively reflect the company's true manufacturing capabilities, easily leading to biased assessment results.

[0038] In summary, existing assessment technologies cannot achieve accurate, efficient, and low-cost quantitative assessment of a company's true manufacturing capabilities without intrusion. Specifically, existing technologies either require intrusion into the company's production site or equipment, making implementation difficult; or the information used for assessment has low credibility, limited dimensions, and is easily forged, failing to effectively distinguish between companies with genuine manufacturing capabilities and trading companies with only circulation attributes. This leads to a technical dilemma of information asymmetry in the field of enterprise manufacturing capability assessment. Therefore, there is an urgent need for an automated quantitative assessment technology solution that can circumvent the aforementioned shortcomings of existing technologies, requires no physical access to the company's production equipment, and possesses anti-counterfeiting capabilities. This solution would fill the gap in existing technologies for enterprise manufacturing capability assessment and solve the technical problem that existing technologies cannot accurately, reliably, and efficiently assess a company's true manufacturing capabilities.

[0039] Figure 2 This method demonstrates the entire process of inputting multi-source data, including data from banks, tax bureaus, power companies, and human resources, into a feature extraction and processing layer for processing, and then outputting the results to an application evaluation layer for analysis, ultimately obtaining an MCI index or validation scale. The feature extraction and processing layer includes... Resource consumption characteristics Supply chain characteristics Human resource characteristics Material logic characteristics Handling of asset structure features.

[0040] This embodiment provides a method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, such as... Figure 1 As shown, the specific steps include the following: Step S100: Obtain multi-dimensional operational data of the target company.

[0041] In this embodiment, multi-dimensional operational data refers to a heterogeneous data set from multiple independent data sources that reflects the key aspects of a target company's business activities. Specifically, it can include various data related to the company's manufacturing capabilities, such as fund transfers, transaction cooperation, human resource management, tax declarations, and asset holdings. Multi-dimensional operational data includes core data that directly reflects production and operation, as well as auxiliary data that supports production activities. Through the complementarity of multi-source data, it comprehensively depicts the company's true operating status.

[0042] The main principle for acquiring multi-dimensional operational data is non-intrusive collection, meaning it requires no physical access to the company's production equipment and is obtained solely through authorized legal data interfaces or compliant documents proactively provided by the company. This collection method reduces the difficulty of data acquisition and avoids interference with the company's normal production and operations. Specifically, the data sources cover the main areas of the company's operations. Financial data comes from financial systems such as bank accounts; transaction data comes from business transaction records; human resource data comes from social security and payroll systems; tax data comes from tax declaration systems; and asset data comes from company asset management-related documents.

[0043] The acquired data is complete and authentic. Completeness is demonstrated by its coverage of the core dimensions required for manufacturing capability assessment, while authenticity is ensured through enterprise-authorized data collection and access to compliant data sources. This step provides a heterogeneous data foundation for feature construction, resolving the issues of incomplete and easily falsified information from single data sources, and providing data support for assessing enterprise manufacturing capabilities.

[0044] In one implementation, the target company's multi-dimensional operating data includes cash flow data, transaction details, human resource data, input and output data, and company asset data. The acquisition of the target company's multi-dimensional operating data specifically includes the following steps: Step S110: Obtain structured data from at least one of the following sources: the enterprise's authorized bank account, tax system, social security system, electricity metering system, or asset management system; or obtain unstructured data from at least one of the following sources: the enterprise's financial statements or scanned copies of invoices.

[0045] In this embodiment, the target company's multi-dimensional operational data can include five categories of operational data, each corresponding to one of five feature categories. Specifically, cash flow data supports the construction of resource consumption features, transaction details support the construction of supply chain features, human resource data supports the construction of human resource features, input and output data support the construction of material logic features, and enterprise asset data supports the construction of asset structure features. Data acquisition must adhere to the principles of legality, compliance, and enterprise authorization to ensure the reliability of data sources and the compliance of data usage.

[0046] Structured data is acquired from various professional systems authorized by the enterprise. These systems feature structured data storage, with standardized data formats that are easy to process directly. For example, corporate bank account transaction data obtained from a bank includes structured fields such as counterparty, transaction amount, transaction time, and summary; VAT invoice data obtained from the tax system includes standard fields such as invoice code, goods name, amount, and invoice date; social security payment lists obtained from the social security system include structured information such as the number of insured persons, contribution base, and payment amount; electricity payment records obtained from the electricity metering system include data such as payment amount, payment cycle, and electricity consumption; and fixed asset data obtained from the asset management system includes information such as asset name, original value, net value, and purchase date.

[0047] Unstructured data mainly comes from documents such as scanned copies of financial statements and invoices provided by companies. This data does not have a standard structured format and requires preprocessing before it can be used. For example, scanned copies of paper financial statements provided by companies need to use Optical Character Recognition (OCR) technology to extract the text information from the image, and then perform structured processing to convert it into calculable values ​​and fields. Scanned copies of invoices also need OCR recognition to extract key information such as invoice numbers, product names, and amounts, and then standardize the format.

[0048] By collecting data from multiple sources, data heterogeneity was ensured, covering multiple dimensions such as finance, taxation, energy, human resources, and assets, providing a data foundation for subsequent multi-feature construction. Simultaneously, by directly using structured data and processing unstructured data via OCR, both efficiency and accuracy in data processing were balanced, avoiding the problem of unstructured data being difficult to utilize. Secondly, the clear correspondence between data classification and feature construction ensured data relevance, preventing redundant data from interfering with the evaluation process and improving the efficiency and accuracy of subsequent feature construction. Furthermore, the legality of data sources and authorization requirements guaranteed the compliance of data use and adherence to relevant data security regulations.

[0049] In addition, the cash flow data in the multi-dimensional operational data can be accessed not only by bank corporate transaction records, but also by the financial accounting data of the enterprise's ERP system, or the corporate account transaction data of a third-party payment platform. The data processing logic is consistent with that of bank corporate transaction records.

[0050] Step S200: Based on the multi-dimensional operational data, construct a multi-dimensional feature set; wherein the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets.

[0051] In this embodiment, the multidimensional feature set refers to a set of features extracted and constructed based on multidimensional operational data that can characterize the core dimensions of an enterprise's manufacturing capabilities, covering the constituent elements of manufacturing capabilities. Among these, resource consumption features reflect the consumption levels of various production resources during the enterprise's production process, specifically energy consumption, material consumption, and other related features; supply chain features reflect the enterprise's transaction relationships with upstream and downstream partners and its industry attributes, specifically supply chain attributes, raw material procurement concentration, and other related features; human resource features reflect the enterprise's personnel size, structure, salary levels, and other human resource conditions, specifically personnel structure, blue-collar index, and other related features; material logic features reflect the rationality of the material transformation between raw material inputs and product outputs, specifically material transformation consistency and other related features; and asset structure features reflect the proportion of productive assets in the enterprise's asset composition, specifically fixed asset structure, heavy asset ratio, and other related features.

[0052] When constructing a multidimensional feature set, it is required to include at least three of the five types of features mentioned above. A single type of feature cannot fully cover the core dimensions of manufacturing capabilities, and a combination of multiple features can achieve complementary advantages. For example, resource consumption features reflect the physical input of production, supply chain features reflect the foundation of industrial collaboration, and human resource features reflect the human support for production. A combination of these three types of features can describe manufacturing capabilities from the three dimensions of input, collaboration, and human resources.

[0053] Transforming heterogeneous operational data into features with industrial semantics enables the transformation from raw data to evaluation indicators. Each type of feature corresponds to a core dimension of manufacturing capability. The process of constructing the feature set is a process of extracting the core elements of an enterprise's manufacturing capability, providing direct data support for feature fusion and quantitative evaluation, and avoiding evaluation biases caused by the disorder of raw data.

[0054] In one implementation, the resource consumption characteristics include energy consumption characteristics. Constructing energy consumption characteristics based on the multi-dimensional operational data specifically includes the following steps: Step S211: Identify energy cost expenditures in the cash flow data of the multi-dimensional operating data; Step S212: Obtain the energy price parameters and transaction time period corresponding to the energy cost expenditure; Step S213: Calculate the energy consumption level of the target enterprise based on the energy cost expenditure, the energy price parameters, and the transaction time period.

[0055] In this embodiment, energy consumption characteristics are an example of resource consumption characteristics. Energy consumption characteristics can be obtained through at least one of the following methods: deduced from energy expenditures based on bank statements, deduced from payment records on third-party payment platforms, or directly obtained from data from an authorized electricity metering system. The specific construction process of energy consumption characteristics is as follows: Figure 3 As shown. Figure 3 The paper demonstrates the reverse logic from cash flow to physical energy consumption.

[0056] In the construction process, the first step is to identify energy cost expenditures within the cash flow data of the multi-dimensional operational data. Specifically, in implementation, such as... Figure 3 As shown, the system iterates through transaction records in the cash flow data, such as reading corporate bank statements, and filters target expenditures using two methods: keyword matching and counterparty identification. Keyword matching involves searching for records whose transaction summaries contain related terms such as "electricity bill," "electricity bill payment," and "energy fee." Counterparty identification involves filtering expenditure records where the counterparty's name is "power supply bureau," "power company," or "energy supply company," indicating an entity engaged in energy supply business. By combining these two methods, the system identifies the company's expenditures on energy consumption.

[0057] Subsequently, the energy price parameters and transaction time period corresponding to the energy expenditure are obtained. The energy price parameter refers to the industrial electricity price in the target company's region. Since industrial electricity typically follows a peak-valley time-of-use pricing policy, and some regions also adjust prices based on seasonal differences, the specific electricity price standard within the corresponding transaction time period is obtained. The transaction time period refers to the billing cycle corresponding to each energy expenditure, which can be monthly, quarterly, or other cycles, extracted from transaction records or determined based on industry-standard billing cycles. The energy price parameter can be obtained from the official electricity price standards published by the local power authority to ensure the accuracy of the electricity price data.

[0058] Based on the above data, the energy consumption level is then calculated. Specifically, the formula for calculating the total physical power consumption is as follows:

[0059] in, This represents the estimated total physical power consumption, expressed in kilowatt-hours (kWh). Indicates the first The amount of energy expenses is in yuan. Indicates the first The transaction time corresponding to each expenditure The industrial time-of-use electricity price is expressed in yuan per kilowatt-hour. This represents the line loss correction factor, used to correct for losses during power transmission. Its value typically ranges from 0.95 to 1.05 and can be determined based on the power transmission conditions and industry averages in the target company's region. Finally, all identified energy cost expenditure records are summed.

[0060] After obtaining the energy consumption level, the energy consumption per unit of output is calculated to further characterize the company's energy consumption efficiency. The formula is as follows:

[0061] in, The normalized scoring function for resource consumption features is used to represent energy consumption per unit of output value, in kilowatt-hours per yuan. Total sales refer to the total operating revenue of the target enterprise within the corresponding statistical time window, in yuan.

[0062] Energy consumption per unit of output can be used to determine the degree of matching between a company's energy consumption and output. If the value is less than 0.1 times the industry threshold, it is considered abnormal, indicating that the company may not be a genuine manufacturing company.

[0063] In addition to electricity bills, the energy types corresponding to energy expenditures can also be determined using payment records for water, industrial gas, and steam bills. The calculation logic is similar to the reverse calculation method for electricity bills, requiring only adjustments to energy price parameters and line loss correction coefficients.

[0064] This step enables non-intrusive energy consumption data acquisition, eliminating the need for physical access to the enterprise's electricity metering or production equipment. This reduces the difficulty and cost of data acquisition while avoiding interference with the enterprise's normal production and operation. Furthermore, the introduction of time-of-use pricing and line loss correction coefficients improves the accuracy of energy consumption level calculations, avoiding estimation biases caused by a single electricity price standard. Simultaneously, the calculation of energy consumption per unit of output reflects the enterprise's total energy consumption, demonstrating energy utilization efficiency and providing a comprehensive basis for manufacturing capacity assessment. Moreover, the construction of this feature transforms financial data (cash flow) into industrial semantic data (energy consumption), realizing the transformation of non-industrial data sources into a representation of industrial manufacturing capacity.

[0065] In one implementation, the supply chain characteristics include supply chain attribute characteristics. Constructing supply chain attribute characteristics based on the multi-dimensional operational data specifically includes the following steps: Step S221: Obtain transaction details containing counterparty information and transaction content from the multi-dimensional business data, and obtain counterparty business attributes and transaction amount data from the transaction details through semantic analysis; Step S222: Calculate the raw material procurement concentration using the transaction amount data; Step S223: Analyze the supply chain attributes of the target company based on the business attributes of the counterparty and the concentration of raw material procurement.

[0066] In this embodiment, by analyzing the attributes of a company's trading partners and the distribution of transaction amounts, the company's role and industry attributes in the supply chain are identified, thus constructing supply chain attribute characteristics. The specific identification logic is as follows: Figure 4 As shown, Figure 4 It demonstrates the steps and processes of semantic analysis, library matching, and weighted scoring.

[0067] The process begins by acquiring transaction details containing counterparty information and transaction content, followed by semantic analysis. Semantic analysis employs Natural Language Processing (NLP) technology to segment and identify entities in the counterparty's name. Specifically, the counterparty's name is first segmented into several lexical units; for example, "XX Chemical Materials Co., Ltd." is segmented into "XX", "Chemical", "Materials", and "Co., Ltd." Then, entity identification is performed, extracting core terms related to industry attributes, such as "Chemical", "Metal", "Precision", and "Materials," which are related to raw material supply, and "Trade", "Import & Export", and "Logistics," which are related to circulation links. Through semantic analysis, the business attributes of the counterparty are extracted from its name to determine whether it is a raw material supplier, trader, or other type of enterprise. Simultaneously, transaction amount data is extracted to provide a numerical basis for calculating procurement concentration.

[0068] After word segmentation, the names of trading partners are matched for attributes to construct a supply chain attribute library containing positive and negative feature libraries. The positive feature library includes words related to raw material supply and production processing, such as "chemicals," "metals," "precision," "materials," and "molds." The negative feature library includes words related to commodity circulation and services, such as "trade," "import and export," "logistics," and "consulting." If a trading partner name matches a positive feature library, it is marked as 1; if it matches a negative feature library, it is marked as 0; otherwise, it is marked as 0.5. This supply chain attribute library is based on historically verified factory samples and is automatically updated using high-frequency tail words extracted by the TF-IDF algorithm, requiring no manual maintenance and ensuring the timeliness and accuracy of business attribute identification.

[0069] Subsequently, the formula for calculating the concentration of raw material procurement is defined:

[0070] in, The raw material procurement concentration is represented by a value between 0 and 1. The sum of raw material transaction amounts represents the total amount of all transactions where the counterparty is a raw material supplier, in yuan. The sum of total outflow transaction amounts represents the total amount of all outflow transactions of the target company within the statistical time window, in yuan. The raw material procurement concentration reflects the proportion of a company's cash outflow used for raw material procurement. The higher the proportion, the more pronounced the company's production orientation.

[0071] Subsequently, based on a comprehensive analysis of the counterparty's business attributes and the concentration of raw material procurement, the company's supply chain attributes are determined. For example, if the counterparty has a high proportion of raw material suppliers and a high concentration of raw material procurement... A higher value, such as greater than 0.5, indicates that the company's supply chain is mainly based on raw material procurement, and its industry attribute is biased towards manufacturing. If the trading partners are mostly trading companies and logistics companies, and the concentration of raw material procurement is low, such as less than 0.2, it indicates that the company's supply chain is mainly based on commodity circulation, and its industry attribute is biased towards trade.

[0072] To meet the scaling requirements for matching supply chain attributes with characteristic integration, the concentration of raw material procurement is... After normalization, we get The value ranges from 0 to 1, and the closer it is to 1, the more obvious the production orientation of the enterprise's supply chain.

[0073] By employing NLP technology to perform semantic analysis on counterparty names, this approach overcomes the limitations of traditional methods that rely solely on transaction amounts. It accurately identifies the industry attributes of counterparties, solving the problem of traders impersonating manufacturing companies by fabricating large transaction amounts. Furthermore, the calculation of raw material procurement concentration quantifies the proportion of raw material purchases by enterprises, providing objective numerical data for supply chain attribute assessment. Simultaneously, the dynamically updated supply chain attribute database ensures the adaptability and scalability of the technical solution, enabling it to address constantly changing market entity names and business types. Moreover, the construction of this feature characterizes the enterprise's manufacturing capability foundation from the perspective of supply chain collaboration, providing data support for comprehensive evaluation and avoiding the one-sidedness of judging from a single dimension.

[0074] In one implementation, the human resources features include personnel structure features. Constructing personnel structure features based on the multi-dimensional operational data specifically includes the following steps: Step S231: Based on the human resources data in the multi-dimensional operational data, calculate the distribution of personnel and the average salary level per person; Step S232: Based on the personnel number distribution, average salary level and minimum wage standard, determine the personnel structure characteristics, which reflect the degree of matching between the target enterprise's labor structure and the manufacturing enterprise.

[0075] In this embodiment, since there are significant differences in the labor structure between manufacturing enterprises and trading and R&D enterprises, the labor structure of an enterprise can be determined based on its human resource data by analyzing the distribution of personnel and salary levels, thereby constructing the personnel structure characteristics.

[0076] The process begins by calculating the distribution of employees and average salary levels based on human resources data. Employee distribution refers to the distribution of employees across different salary ranges, such as the number and percentage of employees in the low-salary range (close to the local minimum wage), the medium-salary range, and the high-salary range. Average salary levels... The calculation formula is:

[0077] The total salary expenditure refers to the total funds spent by enterprises on employee salaries within the statistical time window, expressed in yuan. This represents the total number of employees, i.e., the average number of employees employed by the company during that period.

[0078] Subsequently, the formulas for determining personnel structure characteristics include the formula for determining the average salary range and the formula for determining the proportion of low-salary personnel, which are as follows:

[0079]

[0080] in, This indicates the statutory minimum wage standard in the location of the target company, expressed in yuan / month. The calculated average salary for all employees is based on a monthly salary less than 1.5 times the base salary. The number of employees refers to those whose salaries are less than 1.5 times the local minimum wage. This represents the total number of employees.

[0081] Manufacturing companies, with a higher proportion of blue-collar workers, typically have overall wages ranging from 1.2 to 1.8 times the local minimum wage. This range aligns with the current wage levels for blue-collar workers and distinguishes them from trading companies (which usually have higher average wages) and R&D companies (which have significantly higher average wages). The labor structure of manufacturing companies is typically pyramidal, with a high proportion of low-wage frontline production workers—a key characteristic that differentiates them from other types of enterprises. In contrast, the wage distribution in trading and R&D companies is usually more flat, with low-wage employees making up well below 60%.

[0082] By combining the above two formulas, if both conditions are met, the company is classified as having a manufacturing workforce structure. If neither condition is met, or only one condition is met, the company is classified as having a non-manufacturing workforce structure, such as a trading or R&D workforce. This characteristic reflects the degree of matching between the target company's workforce structure and that of a manufacturing company. The higher the degree of matching, the more solid the company's workforce base for manufacturing capabilities.

[0083] By considering the degree to which the two judgment criteria are met, a normalized score for personnel structural characteristics is constructed. If both the average salary range and the proportion of low-wage workers are met, then... Assign values ​​between 0.8 and 1.0; if only one condition is met, It can be calculated as 0.4-0.7 using linear interpolation; if none of these conditions are met, It can be assigned a value between 0.0 and 0.3.

[0084] Determining human resource structure through salary structure analysis relies on objectively verifiable and difficult-to-falsify data, overcoming the limitations of solely relying on employee numbers. Furthermore, clear, quantitative criteria enhance the objectivity and operability of human resource structure assessment, avoiding biases from subjective experience. In addition, accurately distinguishing between manufacturing and non-manufacturing human resource structures provides a human resource dimension to manufacturing capability assessment, as labor force structure is a core element of manufacturing capability, and manufacturing companies must possess a corresponding frontline production workforce. Moreover, the logic behind this feature's construction aligns with the actual production and operation realities of manufacturing enterprises, making the technical solution practical and feasible.

[0085] In one implementation, the material logical class feature includes a material conversion consistency feature. Constructing the material conversion consistency feature based on the multi-dimensional operational data specifically includes the following steps: Step S241: Obtain the input and output data from the multi-dimensional operating data; Step S242: Map the input data and output data to the industrial BOM knowledge graph, calculate the reachability probability of the production conversion path between input materials and output products, and evaluate the consistency of material conversion.

[0086] In this embodiment, the material conversion consistency feature is an example of a material logic feature. Based on the material conversion laws of industrial production, the feasibility of conversion between an enterprise's incoming materials and outgoing products is verified to determine whether the enterprise possesses genuine production and processing capabilities, thus constructing the material conversion consistency feature. Incoming and outgoing data include, but are not limited to, VAT invoice data, text data from inbound and outbound documents processed by optical character recognition (OCR), and order records from the enterprise's ERP system. All data must be legally obtained with the enterprise's authorization to ensure data authenticity and compliance.

[0087] The process begins by acquiring input and output data. Input data primarily comes from VAT input invoice details, containing a set of names for purchased raw materials, components, and other materials (Input). Output data primarily comes from VAT output invoice details, containing a set of names for sold finished products, semi-finished products, and other products (Output). This data records detailed information about the company's material inputs and product outputs, used to verify material conversion.

[0088] Subsequently, the input and output data are mapped to an industrial BOM (Bill of Materials) knowledge graph. The industrial BOM knowledge graph is a material transformation relationship network built upon massive amounts of industrial production data, containing knowledge about the transformation paths and process requirements from various raw materials and components to finished products. This knowledge graph is automatically generated from massive amounts of historical real-world factory data through association rule mining, possessing incremental learning capabilities. It continuously updates the material mapping relationships with the input of new samples, ensuring the accuracy and timeliness of transformation path judgment.

[0089] The assessment of material conversion consistency is achieved by calculating the path reachability probability, using the following formula:

[0090] in, This represents the probability of material conversion consistency, with a value between 0 and 1. The number of successful path matching from Input to Output refers to the number of material types in the input material set (Input) that can find an effective conversion path through the industrial BOM knowledge graph and ultimately form a certain product in the output product set (Output). The total number of Output products refers to the total number of product types contained in the output product set (Output).

[0091] For example, if the input materials are "plastic granules" and "hardware accessories", and the output products are "plastic shells" and "assemblies", and through an industrial BOM knowledge graph query, "plastic granules" can be converted into "plastic shells", and "hardware accessories" can be combined with other materials to be converted into "assemblies", then the path matching success count is 2. If the total number of output products is 2, then the following is calculated: This indicates extremely high consistency in material conversion; if the incoming material is "steel" and the outgoing product is "software services," there is no conversion path from steel to software services in the industrial BOM knowledge graph, resulting in a successful path match count of 0. The calculated... This indicates that the material conversion is completely inconsistent. Normally, if... If the value is below a set threshold, such as 0.2, it is determined that there is a mismatch between purchase and sales, and the company is suspected to be a shell trading company without real manufacturing capabilities.

[0092] probability of path reachability Normalization is performed to obtain . It reflects the rationality of material transformation and provides standardized input for feature fusion.

[0093] By constructing an objective standard for material transformation using an industrial BOM knowledge graph, the biases of subjective judgment are avoided, making the assessment of material transformation consistency more scientific. Furthermore, the quantitative calculation of path reachability probability enables accurate assessment of material transformation consistency, rather than qualitative judgment. Simultaneously, the incremental learning capability of the knowledge graph ensures the adaptability of the technical solution, covering the transformation relationships of constantly emerging new materials and products. Moreover, this feature verifies the enterprise's manufacturing capabilities from the perspective of material transformation logic, solving the problem of traders impersonating manufacturing enterprises by fabricating purchase and sales invoices, as traders' purchase and sales invoices typically lack genuine material transformation paths and are difficult to verify using this feature. In addition, the construction of this feature transforms tax data into a representation of industrial manufacturing capabilities, further enriching the fusion dimensions of multi-source heterogeneous data.

[0094] In one implementation, the asset structure features include fixed asset structure features. Constructing fixed asset structure features based on the multi-dimensional operating data specifically includes the following steps: Step S251: Obtain enterprise asset data from the multi-dimensional operating data; Step S252: Extract the net value of fixed assets and the total value of assets from the enterprise asset data; Step S253: Based on the net value of fixed assets and the total assets, calculate the fixed asset ratio and analyze the fixed asset structure.

[0095] In this embodiment, the fixed asset structure feature is an implementation feature of the asset structure class feature. Figure 5 It demonstrates the steps and processes of asset data access, percentage calculation, benchmark comparison, and feature output.

[0096] In the feature construction process, enterprise asset data forms the foundation for this feature, primarily derived from the enterprise's annual or quarterly balance sheets and fixed asset lists, and may also include data from filed finance lease contracts. This data meticulously records information such as the enterprise's total assets, fixed asset composition, and value, ensuring the authenticity and accuracy of the data used in feature construction. Net Fixed Asset Value This refers to the balance of fixed assets after deducting accumulated depreciation from their original value. It reflects the current actual value of fixed assets. The net value of productive fixed assets (such as machinery, equipment, and factory buildings) is the core focus and is separated from the total fixed assets. Total Assets It refers to the total value of all assets owned or controlled by an enterprise, including various types of assets such as current assets, fixed assets, and intangible assets, reflecting the size of the enterprise's assets.

[0097] The formula for calculating the fixed asset ratio is:

[0098] in, The percentage of fixed assets is expressed as a percentage. Net fixed assets, in yuan. This represents the total assets, expressed in yuan.

[0099] To convert the fixed asset ratio into a normalized score usable for feature fusion, the Sigmoid function is used, with the following formula:

[0100] in, The normalized score represents the structural characteristics of fixed assets, with values ​​ranging from 0 to 1. To adjust the parameters used to control the steepness of the score curve, the parameters can be determined according to industry characteristics. The industry benchmark value refers to the average fixed asset ratio of the target company's industry, which is obtained by statistically analyzing the asset data of a large number of real manufacturing companies in the industry.

[0101] For example, the net fixed assets of a manufacturing company Total assets: 30 million yuan If the value is 100 million yuan, then the calculation is as follows: If the industry benchmark is 25%, If the value is 0.1, then the calculation is as follows: A score of approximately 0.62 is relatively high, indicating that its fixed asset structure conforms to the characteristics of a manufacturing enterprise. If the net fixed asset value of a trading company... Total assets: 5 million yuan For 100 million yuan, It is far below the industry benchmark. A lower score indicates that its asset structure is biased towards trade, and the final output is the asset structure feature score.

[0102] The fixed asset ratio quantifies the asset-heavy nature of enterprises, providing an objective asset dimension for distinguishing between manufacturing and trading companies. Furthermore, the use of a sigmoid function for normalization ensures that the fixed asset structure feature can be integrated with other features on the same scale, guaranteeing the rationality of feature fusion. Simultaneously, the introduction of industry benchmarks makes the assessment of this feature more aligned with industry realities, avoiding misjudgments caused by differences in asset structures across different industries. Moreover, fixed asset data typically possesses high authenticity and stability, making it difficult to falsify; this feature further strengthens the assessment scheme's resistance to fraud. In addition, this feature reflects the enterprise's manufacturing capability foundation at the asset composition level. Productive fixed assets are the material guarantee for manufacturing enterprises to carry out production activities, and their proportion directly affects the strength of the enterprise's manufacturing capability.

[0103] Step S300: The features of each dimension in the multidimensional feature set are fused to obtain a quantitative evaluation result of the target enterprise's manufacturing capability.

[0104] In this embodiment, feature fusion refers to the process of comprehensively calculating each feature in a multi-dimensional feature set based on the importance of each feature dimension and corresponding weight parameters. By integrating multi-dimensional information through reasonable mathematical methods, the limitations of a single feature are eliminated. The quantitative evaluation result refers to a numerical value or label obtained through feature fusion that objectively reflects the strength of the target company's manufacturing capabilities. Specifically, it can be the Manufacturing Capability Index (MCI), a binary classification label for genuine manufacturers / traders, a company profile report, etc.

[0105] The key to feature fusion lies in the rationality of the weight parameters. The weights should be adapted to the production characteristics of the target company's industry, because different industries have different degrees of dependence on each feature dimension in their manufacturing processes. For example, heavy industry is highly dependent on resource consumption and fixed assets, while light industry may focus more on human resource structure and material conversion efficiency. By fusing features with industry-adapted weights, the evaluation results are more in line with the actual situation of the industry.

[0106] By transforming multi-dimensional feature information into a single, comparable quantitative result, an objective and accurate assessment of a company's manufacturing capabilities is achieved. Compared to qualitative assessments, quantitative results are more objective and practical, facilitating automated processing and decision-making by upper-level business systems. It also solves the problems of ambiguous and difficult-to-quantify comparisons in existing technologies, providing a clear basis for classifying and ranking a company's manufacturing capabilities.

[0107] In one implementation, the quantitative evaluation result is a manufacturing capability index. The process of fusing the features of each dimension in the multidimensional feature set to obtain the quantitative evaluation result of the target enterprise's manufacturing capability specifically includes the following steps: Step S310: Based on the industry category of the target enterprise, obtain the weight parameters of each feature dimension that are adapted to the production characteristics of the industry; Step S320: Based on the weight parameters, the quantitative representation values ​​of the features of each dimension are fused to obtain the manufacturing capability index of the target enterprise; wherein, the fusion includes at least one of logistic regression model fusion or weighted scoring fusion.

[0108] In this embodiment, by combining industry-adapted weight parameters, the quantitative representation values ​​of each dimension feature are comprehensively calculated to obtain a quantitative result that can fully reflect the enterprise's manufacturing capabilities, thereby realizing a quantitative assessment of the enterprise's manufacturing capabilities.

[0109] First, obtain weight parameters that are appropriate for the production characteristics of different industries. Different industries have significantly different production and operational characteristics, and their dependence on various feature dimensions also varies. Therefore, weight parameters must be industry-specific. For example, heavy industries such as steel and chemicals have a high dependence on energy consumption and fixed assets in their production processes; therefore, the weights of resource consumption and asset structure features should be relatively large. Light industries such as textiles and electronics assembly place greater emphasis on human resource structure and material conversion efficiency; therefore, the weights of human resource and material logic features should be appropriately increased. Supply chain-intensive industries such as automotive parts manufacturing require even higher weights for supply chain attribute features.

[0110] The determination of weight parameters must be based on industry characteristics and historical data to ensure their rationality and relevance. These parameters can be obtained through industry research reports, expert judgment, or automatic optimization through training on historical labeled sample sets. The goal is to ensure that the weight parameters accurately reflect the industry's production characteristics and their requirements for each manufacturing capability dimension.

[0111] Subsequently, the quantitative representation values ​​of each dimension's features are comprehensively calculated and integrated based on the weighting parameters, resulting in the Manufacturing Capability Index (MCI). The quantitative representation values ​​of each dimension's features refer to the normalized values ​​of the features constructed through the aforementioned steps, including... , , , , The values ​​are all between 0 and 1, ensuring the feasibility of fusion computing.

[0112] The fusion methods mainly include two types: logistic regression model fusion and weighted scoring fusion. Enterprises can choose one of them or use them in combination according to their actual assessment needs. Logistic regression model fusion uses the quantitative representation values ​​of each feature as input variables and the weight parameters as model coefficients to calculate the manufacturing capability index through the logistic regression algorithm.

[0113] Weighted scoring fusion is another fusion method, and the formula is:

[0114] in, to Let be the weight parameters for each feature dimension, and satisfy . The MCI value ranges from 0 to 1, and the overall score is obtained directly by weighted summation.

[0115] Furthermore, the fusion method can be expanded according to technological development and actual needs. For example, other machine learning algorithms such as Random Forest, Support Vector Machine (SVM) or Neural Networks can be used. As long as the fusion is based on the aforementioned feature dimensions and industry-adaptive weights, it falls within the protection scope of this solution.

[0116] Based on the range of Manufacturing Capability Index (MCI) values, the assessment results are divided into multiple manufacturing capability level intervals: MCI ≥ 0.8 is Level 1, representing strong manufacturing capability; 0.6 ≤ MCI < 0.8 is Level 2, representing relatively strong manufacturing capability; 0.4 ≤ MCI < 0.6 is Level 3, representing medium manufacturing capability; and MCI < 0.4 is Level 4, representing weak manufacturing capability or non-manufacturing capability. The thresholds for the level intervals can be dynamically adjusted according to the production characteristics of the target industry to adapt to the assessment needs of different industries.

[0117] Simultaneously, the manufacturing capability index (MCI) is generated along with contribution vectors and anomaly cause vectors corresponding to each feature dimension. The contribution vectors characterize the degree of influence of each feature dimension on the MCI, for example... The proportion of abnormal causes is recorded in the vector of abnormal causes, which records the combination of features that failed the consistency verification and the deviation logic, such as "high energy consumption but low concentration of raw material procurement". This enables the interpretable output of the assessment results, which is convenient for the upper-level business system to conduct risk analysis and manual review.

[0118] By introducing industry-adaptive weight parameters, the evaluation results are made more aligned with the production characteristics of different industries, avoiding evaluation biases caused by uniform weights and improving the accuracy of the evaluation. Furthermore, multiple fusion methods are provided, balancing the complexity and simplicity of the evaluation. Enterprises can flexibly choose according to data conditions and evaluation needs, enhancing the practicality of the technical solution. Simultaneously, the output of the Manufacturing Capability Index (MCI) achieves standardized quantification of enterprise manufacturing capabilities, facilitating horizontal comparisons between different enterprises and vertical comparisons of the same enterprise at different times. Moreover, feature fusion integrates multi-dimensional heterogeneous features into a single quantitative indicator, realizing a closed loop from multi-source data to evaluation results, aligning with the automated and precise evaluation goals of this invention. In addition, the scalability of the fusion methods ensures that the technical solution can adapt to the development and iteration of machine learning algorithms, extending the technology's lifecycle.

[0119] In one implementation, obtaining the weight parameters of each feature dimension that are adapted to the production characteristics of the target enterprise based on the industry category to which the enterprise belongs specifically includes the following steps: Step S311: Obtain the historical labeled sample set of the industry to which the target enterprise belongs. The historical labeled sample set contains several enterprise samples with known real manufacturing capability labels. Each sample corresponds to the feature quantification value of each dimension in the multidimensional feature set. Step S312: Use a supervised machine learning algorithm to train the historical labeled sample set to obtain the weight parameters of each feature dimension of the industry to which the target enterprise belongs.

[0120] In this embodiment, automatic optimization of weight parameters ensures the rationality of feature fusion and the accuracy of evaluation. Through supervised machine learning algorithms, the weights of each feature dimension appropriate to the target company's industry are automatically determined using historical labeled sample sets, avoiding the subjectivity and limitations of manually setting weights.

[0121] First, obtain the historical labeled sample set of the target company's industry. This historical labeled sample set contains multiple samples of companies with known actual manufacturing capabilities. Each sample corresponds to the quantified values ​​of features across all dimensions in a complete multi-dimensional feature set. , , , , The sample set should include corresponding manufacturing capability labels. Manufacturing capability labels are qualitative or quantitative markers of a company's actual manufacturing capabilities. Qualitative labels include "genuine manufacturing company" and "fake manufacturing company (trader)," while quantitative labels include the company's actual manufacturing capability rating or market recognition score. The sample set should be large enough to cover companies of different sizes and production types within the industry, ensuring representativeness and diversity, thereby improving the accuracy of the weight parameters training.

[0122] The training process utilizes supervised machine learning algorithms to optimize weight parameters. Commonly used algorithms include Gradient Boosting, Random Forest, and Logistic Regression. Companies can choose the appropriate algorithm based on the characteristics of their sample set and evaluation needs. The core logic of the training process is to use the quantified feature values ​​of each sample as input and the manufacturing capability label as output. Through iterative optimization, the weight parameters of each feature dimension are adjusted so that the model's predicted output, i.e., the manufacturing capability index calculated based on the weights, is as close as possible to the true label of the sample.

[0123] Taking the gradient boosting algorithm as an example, the training process first initializes a simple model, such as a constant model, and calculates the initial prediction error. Then, multiple weak learners, such as decision trees, are iteratively built. Each weak learner optimizes the prediction error based on the error of the previous model by adjusting the feature weights. Finally, the prediction results of all weak learners are weighted and combined to obtain the final model, where the feature coefficients are the optimized weight parameters.

[0124] The weight parameters obtained through this training process accurately reflect the contribution of each feature dimension to the assessment of manufacturing capabilities of enterprises within the industry. Features with a greater impact on manufacturing capability prediction have larger weight parameters, while features with a smaller impact have smaller weight parameters. For example, in training on a heavy industry sample set, the weight parameters of resource consumption features and asset structure features are automatically adjusted to a higher level, while in training on a light industry sample set, the weight parameters of human resource features and material logic features will be relatively higher.

[0125] By automatically optimizing weights using supervised machine learning algorithms, the subjective bias of manually setting weights is avoided, making the weight parameters more objective and scientific. Furthermore, training based on historical labeled sample sets from the industry ensures a high degree of fit between the weight parameters and industry production characteristics, improving the accuracy of feature fusion and evaluation results. Simultaneously, the diversity of algorithm choices enhances the flexibility and adaptability of the technical solution, enabling it to meet the training needs of different industries and sample sets. Moreover, it makes automatic updating of weight parameters possible; when industry production characteristics change or a large number of new samples are added, weight parameters adapted to the new situation can be obtained through retraining, ensuring the timeliness of the technical solution. In addition, this step further enhances the automation level of the entire evaluation method, forming a complete automated process from data collection and feature construction to weight determination and feature fusion, significantly reducing the cost of manual intervention.

[0126] In one implementation, the method further includes the following steps: Step S410: Select at least two different types of features from the multidimensional feature set, and verify the consistency of the selected different types of features in at least one of the time dimension, production capacity logic dimension, or material conversion dimension. Step S420: Based on the results of the consistency verification, the quantitative assessment results of the target enterprise's manufacturing capabilities are supplemented with auxiliary annotations, which are used to characterize the risk of data falsification.

[0127] In this embodiment, cross-feature type consistency verification improves the reliability of evaluation results and prevents multi-dimensional feature collusion fraud. By utilizing the strong coupling logical relationship between different types of features, the authenticity and consistency of feature data are verified, thereby identifying abnormal feature combinations and providing risk labeling for quantitative evaluation results.

[0128] First, selecting at least two different types of features for consistency verification is based on the inherent logical connections between various dimensions of manufacturing capabilities. Genuine manufacturing companies inevitably exhibit reasonable matching relationships between features such as resource consumption, human resources, supply chain, material logic, and asset structure, while fraudulent companies find it difficult to simultaneously forge multiple types of features and maintain logical consistency. The selected feature types can be arbitrarily combined from the five feature categories. Common combinations include resource consumption and human resources, supply chain and material logic, and asset structure and resource consumption, ensuring coverage of the core logical connections within manufacturing capabilities.

[0129] Consistency verification dimensions include at least one of the following: time dimension, capacity logic dimension, and material conversion dimension. Time dimension verification refers to the consistency of change trends of different types of characteristics over time. For example, the monthly change trend of energy consumption should match the monthly change trend of product output, and the time of raw material procurement should precede the time of product sales. Capacity logic dimension verification refers to the matching of capacity scale reflected by different types of characteristics. For example, the scale of production personnel reflected by human resource structure should be compatible with the capacity scale reflected by energy consumption, and the production scale reflected by the proportion of fixed assets should match the scale of raw materials procured from the supply chain. Material conversion dimension verification refers to the consistency of the material conversion logic of material-related characteristics with other characteristics. For example, the types and quantities of incoming materials should match the scale of energy consumption, and the types of outgoing products should be compatible with the production scope of fixed assets.

[0130] Subsequently, auxiliary annotation is performed based on the consistency verification results. Specifically, the degree of anomaly of the feature combination is determined according to the verification results, and corresponding fraud risk labels are generated. The specific judgment logic is as follows: if different types of features fully comply with the preset consistency rules in the selected verification dimensions, it is judged as "no fraud risk" and the auxiliary label is low risk; if the verification results have slight deviations but do not exceed the reasonable fluctuation range, such as a slight lag in the time dimension or a slight mismatch in production capacity, it is judged as "low fraud risk" and the auxiliary label is medium risk, with the specific dimension of the deviation noted. If the verification results have significant deviations and the core logical relationship does not hold, such as high energy consumption but no corresponding raw material procurement, or a manufacturing-oriented human resource structure but a very low proportion of fixed assets, it is judged as "high fraud risk" and the auxiliary label is high risk, with a detailed list of abnormal feature combinations and the logical points of mismatch.

[0131] The pre-defined consistency rules are reasonable ranges or logical relationships derived from statistical analysis of characteristic data from real manufacturing enterprises within the industry. For example, the consistency rule for the capacity logic dimension could be "the deviation between the theoretical capacity corresponding to the production personnel scale and the actual capacity corresponding to energy consumption shall not exceed ±20%", while the consistency rule for the material conversion dimension could be "the path reachability probability of input materials being converted into output products shall not be less than 0.5". These rules can be dynamically adjusted according to industry characteristics and sample data to ensure their rationality and adaptability.

[0132] By employing consistency verification, this method becomes adversarial, effectively countering fraudulent activities by traders impersonating manufacturers. For example, when traders inflate electricity consumption figures by claiming "air conditioner is running inactive," the verification of input and output invoices will fail due to a lack of genuine raw material procurement. The wages are too low, and the matching degree between wages and production capacity verification is abnormal because the labor structure does not conform to the characteristics of manufacturing enterprises. It is difficult to verify the fraud in a single dimension through a combination of multiple schemes, which greatly increases the cost of fraud.

[0133] By employing cross-feature type logical verification, a strongly coupled verification relationship of multi-dimensional features is constructed, solving the problem of single-dimensional features being easily forged. Since the cost of simultaneously forging multiple types of features while maintaining logical consistency is extremely high, this significantly improves the anti-forgery capability of the assessment scheme. Secondly, the auxiliary labeling of forgery risks provides a reliable reference for the quantitative assessment results. Upper-level business systems or manual review can combine the manufacturing capability index and forgery risk labeling for comprehensive decision-making, avoiding potential misjudgments that might result from relying solely on quantitative results. Simultaneously, the multi-verification dimensions and flexible feature combination selection ensure the comprehensiveness and relevance of the verification, covering different types of forgery scenarios. Furthermore, the dynamic adjustment capability of preset consistency rules allows the technical solution to adapt to changes in manufacturing capability characteristics across different industries and time periods, exhibiting strong scalability and adaptability. The addition of this step completes the entire assessment process, encompassing feature construction, consistency verification, feature fusion, and risk labeling, further enhancing the reliability and practicality of the assessment results and providing a more comprehensive decision-making basis for upper-level business scenarios.

[0134] The manufacturing capability index and auxiliary labeling results can be applied to at least one of the following business scenarios: supplier authenticity verification and ranking on B2B e-commerce platforms, credit line approval in supply chain finance, supplier risk screening in cross-border trade, and qualification review of enterprises entering industrial parks. The assessment results serve only as a technical risk reference and do not directly constitute a legal determination of an enterprise's operating qualifications.

[0135] In summary, the method proposed in this embodiment utilizes non-invasive, difficult-to-forge multidimensional data to construct an automated model that accurately and cost-effectively quantifies and assesses a company's true manufacturing capabilities, addressing the information asymmetry problem of traders masquerading as manufacturers. This method can automatically infer a company's industrial load level, raw material conversion consistency, and labor structure characteristics without requiring physical access to production equipment, thereby reducing manual verification costs and improving the technical accuracy of identifying manufacturing companies.

[0136] This technical solution can be deployed on local servers of banks or industrial park management entities for private deployment, or it can be deployed in the cloud as Software as a Service (SaaS), accessible to upper-layer business systems via API interfaces. In addition to manufacturing capability indices, classification labels, and enterprise profile reports, the evaluation results can also generate blockchain digital certificates (NFTs) for trusted storage and transfer of enterprise manufacturing capabilities.

[0137] During implementation, this technical solution can adopt a federated learning or multi-party secure computation (MPC) architecture. Data does not leave the enterprise's local area or the authorized institution's local area. Only model parameters, intermediate feature values ​​or gradient information are transmitted in encrypted form to achieve cross-institutional data collaborative verification. Under the premise of complying with relevant laws and regulations, data privacy and assessment accuracy are guaranteed.

[0138] like Figure 6 As shown in the figure, this embodiment of the invention provides a quantitative evaluation system for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion. The system includes: a multi-dimensional business data acquisition module 10, a multi-dimensional feature set construction module 20, and an enterprise manufacturing capability quantitative evaluation module 30.

[0139] Specifically, the multi-dimensional operational data acquisition module 10 is used to acquire multi-dimensional operational data of the target enterprise; the multi-dimensional feature set construction module 20 is used to construct a multi-dimensional feature set based on the multi-dimensional operational data; wherein, the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets; the enterprise manufacturing capability quantitative assessment module 30 is used to fuse the features of each dimension in the multi-dimensional feature set to obtain a quantitative assessment result of the enterprise manufacturing capability of the target enterprise.

[0140] Based on the above embodiments, the present invention also provides a terminal device, the principle block diagram of which can be as follows: Figure 7As shown, the terminal device includes a processor, memory, network interface, display screen, and temperature sensor connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a quantitative evaluation method for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion. The display screen can be an LCD screen or an e-ink screen. The temperature sensor is pre-installed inside the terminal device to detect the operating temperature of internal components.

[0141] Those skilled in the art will understand that Figure 7 The block diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal device to which the present invention is applied. A specific terminal device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0142] In one embodiment, a terminal device is provided, including a memory and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by one or more processors, the one or more programs including instructions for performing operations as described in the embodiments of the methods above.

[0143] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0144] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0145] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, characterized in that, The method includes: Obtain multi-dimensional operational data of the target company; Based on the aforementioned multi-dimensional operational data, a multi-dimensional feature set is constructed; wherein, the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets. By fusing the features of each dimension in the multidimensional feature set, a quantitative assessment result of the target enterprise's manufacturing capability is obtained.

2. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The target company's multi-dimensional operational data includes cash flow data, transaction details, human resource data, input and output data, and company asset data. Obtaining the target company's multi-dimensional operational data includes: Structured data obtained from at least one of the following sources: authorized bank accounts, tax systems, social security systems, electricity metering systems, or asset management systems of the enterprise; or unstructured data obtained from at least one of the following sources: scanned copies of financial statements or invoices provided by the enterprise.

3. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The resource consumption characteristics include energy consumption characteristics. Energy consumption characteristics are constructed based on the multi-dimensional operational data, including: Identify energy cost expenditures within the cash flow data of the multi-dimensional operational data; Obtain the energy price parameters and transaction time period corresponding to the energy cost expenditure; The energy consumption level of the target enterprise is calculated based on the energy cost expenditure, the energy price parameters, and the transaction time period.

4. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The supply chain characteristics include supply chain attribute characteristics, which are constructed based on the multi-dimensional operational data, including: The transaction details, which include counterparty information and transaction content, are obtained from the multi-dimensional operational data. Semantic analysis is then used to obtain the counterparty business attributes and transaction amount data from the transaction details. The concentration of raw material procurement is calculated using the transaction amount data. Based on the business attributes of the counterparty and the concentration of raw material procurement, the supply chain attributes of the target company are analyzed.

5. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The human resources features include personnel structure features, which are constructed based on the multi-dimensional operational data, including: Based on the human resources data in the multi-dimensional operational data, the distribution of personnel and the average salary level per person are calculated. Based on the personnel distribution, average salary level, and minimum wage standard, the personnel structure characteristics are determined, which reflect the degree of matching between the target enterprise's labor structure and that of a manufacturing enterprise.

6. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The material logic features include material conversion consistency features. These features are constructed based on the multi-dimensional operational data and include: Obtain the input and output data from the multi-dimensional operating data; The input and output data are mapped to the industrial BOM knowledge graph to calculate the reachability probability of the production conversion path between input materials and output products, and to evaluate the consistency of material conversion.

7. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The asset structure features include fixed asset structure features, which are constructed based on the multi-dimensional operating data, including: Obtain enterprise asset data from the multi-dimensional operational data; Extract the net value of fixed assets and the total assets from the enterprise's asset data; Based on the net value of fixed assets and the total assets, the fixed asset ratio is calculated, and the fixed asset structure is analyzed.

8. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The quantitative assessment result is a manufacturing capability index. The process of fusing the features across the multidimensional feature set to obtain the quantitative assessment result of the target enterprise's manufacturing capability includes: Based on the industry category of the target enterprise, obtain the weight parameters of each feature dimension that are adapted to the production characteristics of the industry; Based on the weight parameters, the quantitative representation values ​​of the features in each dimension are fused to obtain the manufacturing capability index of the target enterprise; wherein, the fusion includes at least one of logistic regression model fusion or weighted scoring fusion.

9. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 8, characterized in that, The step of obtaining weight parameters for each feature dimension that matches the production characteristics of the target enterprise based on its industry category includes: Obtain a historical labeled sample set of the industry to which the target enterprise belongs. The historical labeled sample set contains several enterprise samples with known real manufacturing capability labels. Each sample corresponds to the feature quantification value of each dimension in the multidimensional feature set. The historical labeled sample set is trained using a supervised machine learning algorithm to obtain the weight parameters of each feature dimension of the industry to which the target enterprise belongs.

10. The method for quantitatively evaluating enterprise manufacturing capabilities based on multi-source heterogeneous data fusion according to claim 1, characterized in that, The method further includes: At least two different types of features are selected from the multidimensional feature set, and the consistency of the selected different types of features is verified in at least one of the following dimensions: time dimension, production capacity logic dimension, or material conversion dimension. Based on the results of the consistency verification, the quantitative assessment results of the target company's manufacturing capabilities are supplemented with auxiliary annotations, which are used to characterize the risk of data falsification.

11. A quantitative evaluation system for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion, characterized in that, The system includes: The multi-dimensional operational data acquisition module is used to acquire multi-dimensional operational data of the target enterprise. A multi-dimensional feature set construction module is used to construct a multi-dimensional feature set based on the multi-dimensional operational data; wherein, the multi-dimensional feature set includes at least three of the following: resource consumption features, supply chain features, human resource features, material logic features, and asset structure features. The resource consumption features reflect the level of production resource consumption, the supply chain features reflect the upstream and downstream attributes of the supply chain, the human resource features reflect the human resource structure, the material logic features reflect the material conversion logic, and the asset structure features reflect the proportion of productive assets. The enterprise manufacturing capability quantitative assessment module is used to fuse the features of each dimension in the multidimensional feature set to obtain the quantitative assessment result of the target enterprise's manufacturing capability.

12. A terminal device, characterized in that, The terminal device includes a memory, a processor, and an enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion stored in the memory and executable on the processor. When the processor executes the enterprise manufacturing capability quantification evaluation program based on multi-source heterogeneous data fusion, it implements the steps of the enterprise manufacturing capability quantification evaluation method based on multi-source heterogeneous data fusion as described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a quantitative evaluation program for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion. When the program is executed by a processor, it implements the steps of the quantitative evaluation method for enterprise manufacturing capabilities based on multi-source heterogeneous data fusion as described in any one of claims 1-10.

Images