Industrial enterprise credit evaluation method and device based on multi-dimensional real-time streaming data

Abstract

The invention provides an industrial enterprise credit evaluation method and device based on multi-dimensional real-time streaming data, and relates to the technical field of big data and credit evaluation. The method comprises the following steps: based on enterprise public registration information, carrying out preliminary classification on small and micro enterprises according to national economy industry classification standards and city-county region division; extracting text features of the enterprise basic information from the enterprise public registration information by using a linear discriminant analysis LDA model, generating word sequence features representing enterprise categories, and grading the enterprises based on a similarity matching result of the word sequence features and a preset industry theme; performing multi-dimensional data association fusion on the enterprise public operation data through a Flink CDC real-time stream processing technology to generate fused data, and calculating an enterprise credit evaluation result based on the fused data; wherein the public operation data comprises at least one of enterprise operational data, bidding and tendering transaction data and e-commerce platform transaction data. The enterprise credit evaluation accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the field of big data and credit evaluation technology, and in particular to a method and device for industrial enterprise credit evaluation based on multi-dimensional real-time streaming data. Background Art

[0002] As an important component of the national economy, small and micro enterprises (SMEs) have long faced the challenges of limited financing channels and difficulty accessing credit resources. Traditional bank credit systems typically rely on static indicators such as corporate financial statements and collateralized assets for credit assessments. However, SMEs' small operating scale, opaque financial information, and weak risk tolerance make it difficult for them to meet traditional credit standards. Despite the gradual advancement of inclusive finance policies in recent years, existing credit evaluation methods still have significant limitations: on the one hand, relying on single-dimensional historical data (such as revenue scale) makes it difficult to dynamically capture a company's actual operating quality and development trends; on the other hand, the lack of real-time integration and analysis capabilities for multi-source heterogeneous data (such as bidding transactions, e-commerce sales, and regional economic indicators) makes it difficult for banks to accurately identify SMEs with development potential, resulting in inefficient credit resource allocation. Therefore, a method to improve the accuracy of corporate credit evaluation is urgently needed. Summary of the Invention

[0003] The embodiments of the present invention provide a method and device for industrial enterprise credit evaluation based on multi-dimensional real-time streaming data to solve the problem of improving the accuracy of enterprise credit evaluation.

[0004] In a first aspect, an embodiment of the present invention provides an industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data, comprising:

[0005] Based on the public registration information of enterprises, a preliminary classification of small and micro enterprises was carried out according to the national economic industry classification standards and the city and county regional divisions to obtain preliminary classification results;

[0006] Using a Linear Discriminant Analysis (LDA) model, extract text features of basic enterprise information from the publicly registered information of the enterprise to generate word sequence features representing the enterprise category. Based on the similarity matching results between the word sequence features and preset industry themes, and in combination with the preliminary classification results, the enterprises are graded. The basic enterprise information includes the enterprise name, registered address, business scope, and public introduction.

[0007] Flink CDC real-time stream processing technology is used to perform multi-dimensional data association and fusion on the enterprise's public operational data to generate fused data, and the enterprise credit evaluation results are calculated based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

[0008] In a possible implementation, the national economic industry classification is based on the national standard "National Economic Industry Classification" (GB / T 4754), and the city and county regional division is based on the administrative division code of the enterprise registration address.

[0009] In one possible implementation, the bidding transaction data includes transaction records of the government procurement network and the centralized procurement network, and the e-commerce platform transaction data includes sales data of Taobao, JD.com and Pinduoduo.

[0010] In a possible implementation, the calculating of the enterprise credit evaluation result based on the fused data includes:

[0011] Determine evaluation data of the enterprise to be evaluated based on the fused data, the evaluation data including: monthly change in product unit price, monthly change in upstream material price, monthly average product price, product sales volume, industry average price, and payment collection period;

[0012] Calculate the comprehensive data index of product strength of the enterprise to be evaluated based on the evaluation data;

[0013] The enterprise credit evaluation result is determined based on the comprehensive product strength data index.

[0014] In a possible implementation, calculating the comprehensive product strength data index of the enterprise to be evaluated based on the evaluation data includes:

[0015] Determine a product strength index based on the monthly average price of the product, the sales volume of the product, the industry average price, and the collection period;

[0016] Determine the industry region coefficient based on the monthly change value of the product unit price and the monthly change value of the upstream material price;

[0017] The product of the industry region coefficient and the product strength index is determined as the comprehensive product strength data index.

[0018] In one possible implementation, determining the product strength index based on the monthly average price of the product, product sales volume, industry average price, and collection period includes:

[0019] Calculate the product's monthly average price and the product's sales volume, and then divide the result by the product's average selling price and the collection period for similar products in the industry;

[0020] The calculation result is recorded as the product strength index.

[0021] In a possible implementation, determining the industry region coefficient according to the monthly change value of the product unit price and the monthly change value of the upstream material price includes:

[0022] Calculate the difference between the monthly change in the product unit price and the monthly change in the upstream material price;

[0023] The difference is recorded as the industry region coefficient.

[0024] In one possible implementation, the product strength comprehensive data index is calculated as follows:

[0025] β = α × (enterprise monthly average price × enterprise monthly sales volume) / (industry average price × collection period)

[0026] Among them, α is the industry regional coefficient, which is obtained by calculating the monthly changes in gross profit margin of single products of similar enterprises.

[0027] In a possible implementation, the calculation formula of the industry region coefficient α is:

[0028] α = Monthly change in product unit price - Monthly change in upstream material price

[0029] Among them, the monthly change in product unit price is dynamically updated based on the historical transaction data of enterprises in the same industry and region.

[0030] In a possible implementation, it also includes: according to the enterprise classification and grading results, the industry and region coefficients are adjusted differently to adapt to the evaluation needs of different industries and regions.

[0031] In a possible implementation, extracting text features of basic enterprise information from the public enterprise registration information using a linear discriminant analysis (LDA) model includes:

[0032] Perform Chinese word segmentation and stop word filtering on the publicly registered information of the enterprise;

[0033] The LDA model is used to extract the text topic distribution corresponding to the basic information of the enterprise from the processed public registration information of the enterprise, and generate word sequence features.

[0034] In one possible implementation, the stop word filtering includes:

[0035] Remove punctuation and the default Chinese stop word list, and merge synonyms and antonyms.

[0036] In a possible implementation, the method further includes:

[0037] When an enterprise credit evaluation instruction and / or an evaluation update timing instruction is received, the multi-dimensional data association and fusion of the enterprise's public operation data through FlinkCDC real-time stream processing technology and subsequent operations are repeatedly performed to re-determine the enterprise credit evaluation result.

[0038] In a second aspect, an embodiment of the present invention provides an industrial enterprise credit evaluation device based on multi-dimensional real-time streaming data, comprising:

[0039] The preliminary classification module is used to conduct preliminary classification of small and micro enterprises based on the publicly registered information of enterprises, according to the national economic industry classification standards and the city and county regional divisions, and obtain preliminary classification results;

[0040] The extraction module is used to extract the text features of the basic information of the enterprise from the public registration information of the enterprise using the LDA model to generate word sequence features representing the enterprise category.

[0041] A grading module, configured to grade enterprises based on the similarity matching results between the word sequence features and the preset industry themes, combined with the preliminary classification results; wherein the basic enterprise information includes the enterprise name, registered address, business scope and public introduction;

[0042] The evaluation module is used to perform multi-dimensional data correlation and fusion on the enterprise's public operational data through Flink CDC real-time stream processing technology to generate fused data, and calculate the enterprise credit evaluation results based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

[0043] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method in the first aspect or any possible implementation of the first aspect.

[0044] In an embodiment of the present invention, a preliminary classification of small and micro enterprises is performed based on the national economic industry classification standards and the city and county regional divisions. The text features such as the enterprise name and registered address are extracted using the LDA model to generate word sequences, and the industry themes are matched to achieve grading. This can effectively solve the problems of low efficiency and strong subjectivity caused by the traditional classification method relying on manual experience. At the same time, Flink CDC real-time stream processing technology is used to integrate multi-source heterogeneous data such as enterprise operating data, bidding and tendering, and e-commerce platforms, breaking through the data delay bottleneck of the offline batch processing mode, realizing dynamic multi-dimensional data fusion, and improving the real-time performance and data coverage of credit evaluation. The credit evaluation results finally generated not only reflect the static attributes of the enterprise, but also capture dynamic operating trends, improve the accuracy of enterprise credit evaluation, and thus provide a more comprehensive decision-making basis for bank credit. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flowchart of an implementation method for industrial enterprise credit evaluation based on multi-dimensional real-time streaming data provided by an embodiment of the present invention;

[0046] Figure 2This is a schematic diagram of using the LDA model to extract text features of basic enterprise information from public enterprise registration information;

[0047] Figure 3 It is a structural diagram of an industrial enterprise credit evaluation device based on multi-dimensional real-time streaming data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In the embodiments of this application, we focus on the credit assessment of small and micro-sized enterprises. In actual implementation, the enterprise credit assessment method based on multi-dimensional real-time streaming data provided by this application is applicable not only to small and micro-sized enterprises, but also to the credit assessment of other enterprises, meeting the requirements of real-time and high-precision credit assessment.

[0049] The evaluation of small and micro enterprises mainly needs to be based on data including operational data, intellectual property rights, social security, etc. Due to the small scale of small and micro enterprises and the high probability of operational fluctuations, relying on single-dimensional data (such as revenue scale) for evaluation is one-sided, and the availability risk of the evaluation results is relatively high. The solution provided in the embodiment of the present application, based on tracking the characteristics of small and micro enterprises, performs Flink CDC real-time data calculation on the transaction status of the enterprise's products in the bidding market to obtain a comprehensive product strength data index, and adjusts the evaluation coefficient in the comprehensive product strength data index according to the characteristics of the industry and region, so as to dynamically update the statistics to form a development evaluation of small and micro enterprises.

[0050] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Figure 1 This is a flow chart of the implementation of the industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data provided by the embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0052] S101, based on the public registration information of enterprises, preliminary classification of small and micro enterprises is carried out according to the national economic industry classification standards and city and county regional divisions to obtain preliminary classification results.

[0053] The execution entities of each embodiment of the present application can be servers, processors, microprocessors and other devices with data processing functions. In the actual implementation process, the specific implementation method of the execution entity can be selected according to actual needs. This embodiment does not impose any special restrictions on this, as long as it is a device with data processing functions.

[0054] In one possible implementation, the national economic industry classification is based on the national standard "National Economic Industry Classification" (GB / T 4754), and the city and county regional division is based on the administrative division code of the enterprise registration address.

[0055] S102, using the linear discriminant analysis (LDA) model to extract text features of basic enterprise information from publicly registered enterprise information, generating word sequence features that represent enterprise categories, and grading enterprises based on the similarity matching results between word sequence features and preset industry themes, combined with preliminary classification results; wherein, basic enterprise information includes enterprise name, registered address, business scope and public introduction.

[0056] The preset industry theme is the set enterprise classification label. As shown in the table below, enterprise classification includes primary classification and secondary classification. The primary classification is the provincial and municipal classification standard, and the secondary classification is the national economic industry classification standard.

[0057]

[0058] S103: Perform multi-dimensional data correlation and fusion on the enterprise's public operational data through Flink CDC real-time stream processing technology to generate fused data, and calculate the enterprise credit evaluation results based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

[0059] In one possible implementation, bidding transaction data includes transaction records from the government procurement network and the centralized procurement network, and e-commerce platform transaction data includes sales data from Taobao, JD.com, and Pinduoduo.

[0060] In this embodiment, small and micro enterprises are preliminarily classified based on the national economic industry classification standards and the city and county regional divisions. The LDA model is used to extract text features such as enterprise names and registered addresses to generate word sequences, and industry themes are matched to achieve grading. This can effectively solve the low efficiency and strong subjectivity of traditional classification methods that rely on manual experience. At the same time, Flink CDC real-time stream processing technology is used to integrate multi-source heterogeneous data such as enterprise operating data, bidding and tendering, and e-commerce platforms, breaking through the data delay bottleneck of offline batch processing mode, realizing dynamic multi-dimensional data fusion, and improving the real-time performance and data coverage of credit evaluation. The final credit evaluation results not only reflect the static attributes of the enterprise, but also capture dynamic operating trends, improve the accuracy of enterprise credit evaluation, and thus provide a more comprehensive decision-making basis for bank credit.

[0061] In one possible implementation, the enterprise credit evaluation result is calculated based on the fused data, including:

[0062] Determine the evaluation data of the company to be evaluated based on the integrated data. The evaluation data includes: monthly change in product unit price, monthly change in upstream material prices, monthly average product price, product sales volume, industry average price, and collection period;

[0063] Calculate the comprehensive data index of product strength of the enterprise to be evaluated based on the evaluation data;

[0064] The enterprise credit evaluation result is determined based on the comprehensive product strength data index.

[0065] The purpose of determining the evaluation data of the company to be evaluated is to analyze the company's product competitiveness. Product competitiveness includes the type of main products (including physical goods, services, software, etc. sold), product functions, and comparisons with similar products.

[0066] In this example, by extracting core evaluation indicators such as product unit price changes, upstream material prices, and sales volume from the fused data and calculating a comprehensive product strength data index based on these indicators, a company's market competitiveness and operational stability can be quantified. Compared to traditional evaluation based on a single financial indicator, this method combines multi-dimensional dynamic data such as price fluctuations, supply chain costs, and sales efficiency to accurately reflect a company's true operational capabilities, avoid credit misjudgments caused by one-sided data, and thus improve banks' controllability of credit risks for small and micro enterprises.

[0067] In one possible implementation, the comprehensive product strength index of the enterprise to be evaluated is calculated based on the evaluation data, including:

[0068] Determine the product strength index based on the product's monthly average price, product sales volume, industry average price, and collection period;

[0069] Determine the industry region coefficient based on the monthly change in product unit price and the monthly change in upstream material prices;

[0070] The product of the industry regional coefficient and the product strength index is determined as the comprehensive product strength data index.

[0071] Among them, the monthly average product price and product sales volume reflect the company's own operating performance, the industry average price and collection period reflect the industry benchmark, and the monthly change in product unit price and the monthly change in upstream material price reflect regional market fluctuations. By combining the company's own operating performance, industry benchmarks and regional market fluctuations, the company's product competitiveness can be clearly determined.

[0072] In this embodiment, the design of multiplying the product strength index with the industry regional coefficient to generate a comprehensive data index addresses the problem of traditional models ignoring regional economic differences and industry characteristics. By dynamically linking a company's own operating performance with industry benchmarks and incorporating regional market volatility adjustment coefficients, the regional adaptability and industry differentiation of the evaluation results are significantly enhanced, allowing credit evaluations to reflect both individual company capabilities and the external market environment.

[0073] In one possible implementation, the product strength index is determined based on the product's monthly average price, product sales volume, industry average price, and collection period, including:

[0074] Calculate the product's monthly average price and product sales volume, then divide it by the product of the average selling price of similar products in the industry and the collection period;

[0075] The calculation result is recorded as the product strength index.

[0076] In this embodiment, the product strength index calculation method can intuitively quantify the comprehensive performance of a company's products in terms of price competitiveness, sales volume, and capital turnover efficiency. By comparing company data with the industry average, the impact of industry fluctuations on the evaluation is eliminated. The introduction of the collection period indicator also strengthens the assessment of capital chain health, avoids the hidden risks caused by excessively long payment periods, and makes the evaluation results more valuable for industry reference.

[0077] In one possible implementation, the industry region coefficient is determined based on the monthly change in product unit price and the monthly change in upstream material prices, including:

[0078] Calculate the difference between the monthly change in product unit price and the monthly change in upstream material price;

[0079] The difference is recorded as the industry region coefficient.

[0080] In this embodiment, by calculating the difference between changes in product unit prices and changes in upstream material prices as the industry regional coefficient, it can dynamically reflect the company's cost-shifting ability and regional market supply and demand characteristics. For example, when raw material prices rise, if the company can simultaneously increase its selling price (i.e., the difference approaches zero), it indicates that it has strong market bargaining power. This coefficient incorporates regional cost pressures into the credit evaluation system, helping banks identify small and micro enterprises that can still maintain profitability in specific market environments, and improving the refinement of credit decision-making.

[0081] In one possible implementation, the product strength comprehensive data index is calculated as follows:

[0082] β = α × (enterprise monthly average price × enterprise monthly sales volume) / (industry average price × collection period)

[0083] Among them, α is the industry regional coefficient, which is obtained by calculating the monthly changes in gross profit margin of single products of similar enterprises.

[0084] In one possible implementation, the calculation formula of the industry region coefficient α is:

[0085] α = Monthly change in product unit price - Monthly change in upstream material price

[0086] Among them, the monthly change in product unit price is dynamically updated based on the historical transaction data of enterprises in the same industry and region.

[0087] In one possible implementation method, it also includes: according to the results of enterprise classification and grading, the industry and regional coefficients are adjusted differently to adapt to the evaluation needs of different industries and regions.

[0088] For example, the operating conditions of manufacturing products will be significantly affected by fluctuations in raw material prices, while the operating conditions of retail products will fluctuate with the change of sales seasons.

[0089] For the manufacturing industry: Due to significant fluctuations in raw material prices (such as steel prices), the calculation formula for α is adjusted to:

[0090] α 制造 =(monthly change in product unit price - monthly change in upstream material price) × 1.2

[0091] Among them, the adjustment coefficient is 1.2, which is used to amplify the impact of raw material costs on regional competitiveness and reflect the manufacturing industry's sensitivity to the supply chain.

[0092] Retail industry: In view of the strong seasonality of sales (such as holiday promotions), the adjusted formula is:

[0093] α 零售 =(monthly change in product unit price - monthly change in upstream material price) × 0.8

[0094] + Sales volume volatility × 0.2

[0095] Among them, the introduction of a sales volume volatility weight of 0.2 reduces the absolute dominance of price factors, which is more in line with the characteristics of the retail industry that relies on sales scale.

[0096] In this embodiment, the industry and regional coefficients are adjusted differently according to the results of enterprise classification and grading, which enables the credit evaluation model to be deeply adapted to the characteristics of different industries and regions, and significantly improves the accuracy and practicality of the evaluation results.

[0097] In one possible implementation, the LDA model is used to extract text features of basic enterprise information from public enterprise registration information, including:

[0098] Perform Chinese word segmentation and stop word filtering on public enterprise registration information;

[0099] The LDA model is used to extract the text topic distribution corresponding to the basic information of the enterprise from the processed public registration information of the enterprise, and generate word sequence features.

[0100] Figure 2 This is a flowchart of extracting text features of basic enterprise information from public enterprise registration information using the LDA model. Figure 2As shown in the figure, word sequences are extracted from documents recording the public registration information of enterprises, and the text topic distribution of different word sequences is inferred. The word sequence features are output, and the efficiency of text feature reading is improved through Chinese word segmentation. Stop word filtering reduces the interference of text noise on feature extraction.

[0101] In this embodiment, word sequence features are automatically generated from corporate public information through Chinese word segmentation, stop word filtering and LDA topic extraction, which solves the problems of low efficiency and poor consistency of traditional manual text feature annotation.

[0102] In one possible implementation, stop word filtering includes:

[0103] Remove punctuation and the default Chinese stop word list, and merge synonyms and antonyms.

[0104] For example, after filtering out punctuation and redundant words ("registered address" and "city") in "Registered Address: District B, City A," the key regional term "District B" is retained to ensure accurate regional division. This step optimizes the representational capabilities of text features, making subsequent topic matching more accurate for real-world business scenarios.

[0105] For example, "limited" and "limited company" are synonyms, so merge "limited" and "limited company".

[0106] In this embodiment, merging synonyms and near-synonyms in stop word filtering can reduce the interference of text noise on feature extraction and enhance semantic consistency.

[0107] In the specific implementation process, the code for Chinese word segmentation and stop word filtering of enterprise public registration information is as follows:

[0108]

[0109]

[0110] In a possible implementation, the method further includes:

[0111] Upon receiving an enterprise credit evaluation instruction and / or an evaluation update timing instruction, the system repeatedly performs multi-dimensional data association and fusion of the enterprise's public operational data using Flink CDC real-time stream processing technology, and subsequent operations to re-determine the enterprise credit evaluation result.

[0112] In the actual implementation process, corporate credit evaluation instructions are sent by users themselves and automatically generated by the system based on fluctuations in corporate operating data or external events.

[0113] For example: When a bank's risk control personnel discover that a company is involved in a major lawsuit or equity change, they can manually send an evaluation instruction through the management platform to immediately initiate the credit re-evaluation process.

[0114] For example, the system monitors fluctuation thresholds in a company's operating data in real time and pre-sets the following trigger conditions: If a business indicator shows an abnormality: if a company's monthly e-commerce sales drop by more than 30% month-over-month, its accounts receivable delinquency rate exceeds 15% for two consecutive weeks, or its bidding success rate falls below 50% of the industry average, the system automatically generates a credit evaluation instruction to initiate a credit reassessment process. This design enables banks to promptly respond to fluctuations in the operations of small and micro enterprises, adjust credit policies before risks are exposed, and significantly enhance the risk prevention and control capabilities of inclusive finance services.

[0115] External event association: If the industrial and commercial information shows that the enterprise is included in the list of abnormal operations, or the judicial platform updates the litigation record, the system automatically generates an enterprise credit evaluation instruction, triggering an update of the credit evaluation results.

[0116] In addition, the system supports flexible configuration of update cycles for scheduled instructions. Optionally, evaluation update schedule instructions can be set according to monthly, quarterly, and annual update cycles.

[0117] Specifically, monthly updates: full data fusion and credit score calculation are automatically performed on the 1st of each month, covering corporate operating data (such as sales, inventory turnover) and external transaction data (such as bidding transaction volume).

[0118] Quarterly update: At the beginning of each quarter, the dynamic adjustment results of the industry regional coefficient α (such as regional policy changes, raw material price index) are integrated, and the comprehensive product strength index is recalculated.

[0119] Annual update: Based on corporate annual reports and tax data, we retrospectively calibrate historical credit scores and optimize model parameters.

[0120] In this embodiment, the credit evaluation process is repeatedly executed through a timing or event triggering mechanism to ensure that the evaluation results are dynamically adjusted according to the business status of the enterprise.

[0121] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0122] The following are device embodiments of the present invention. For details not fully described therein, reference may be made to the corresponding method embodiments described above.

[0123] Figure 3 The following is a schematic diagram showing the structure of an industrial enterprise credit evaluation device based on multi-dimensional real-time streaming data provided by an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, which are described in detail as follows:

[0124] like Figure 3As shown, the industrial enterprise credit evaluation device based on multi-dimensional real-time streaming data includes:

[0125] A preliminary classification module 301 is used to perform preliminary classification of small and micro enterprises based on publicly registered enterprise information, in accordance with national economic industry classification standards and city and county regional divisions, and obtain preliminary classification results;

[0126] The extraction module 302 is used to extract the text features of the basic information of the enterprise from the public registration information of the enterprise using the linear discriminant analysis (LDA) model to generate word sequence features representing the enterprise category.

[0127] The grading module 303 is used to grade enterprises based on the similarity matching results between word sequence features and preset industry themes, combined with the preliminary classification results; wherein the basic information of the enterprise includes the enterprise name, registered address, business scope and public introduction;

[0128] Evaluation module 304 is used to perform multi-dimensional data correlation and fusion on the enterprise's public operational data through Flink CDC real-time stream processing technology to generate fused data, and calculate the enterprise credit evaluation results based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

[0129] In one possible implementation, the evaluation module 304 is specifically configured to:

[0130] Determine evaluation data of the enterprise to be evaluated based on the fused data, the evaluation data including: monthly change in product unit price, monthly change in upstream material price, monthly average product price, product sales volume, industry average price, and payment collection period;

[0131] Calculate the comprehensive data index of product strength of the enterprise to be evaluated based on the evaluation data;

[0132] The enterprise credit evaluation result is determined based on the comprehensive product strength data index.

[0133] In one possible implementation, the evaluation module 304 is specifically configured to:

[0134] Determine a product strength index based on the monthly average price of the product, the sales volume of the product, the industry average price, and the collection period;

[0135] Determine the industry region coefficient based on the monthly change value of the product unit price and the monthly change value of the upstream material price;

[0136] The product of the industry region coefficient and the product strength index is determined as the comprehensive product strength data index.

[0137] In one possible implementation, the evaluation module 304 is specifically configured to:

[0138] Calculate the product's monthly average price and the product's sales volume, and then divide the result by the product's average selling price and the collection period for similar products in the industry;

[0139] The calculation result is recorded as the product strength index.

[0140] In one possible implementation, the evaluation module 304 is specifically configured to:

[0141] Calculate the difference between the monthly change in the product unit price and the monthly change in the upstream material price;

[0142] The difference is recorded as the industry region coefficient.

[0143] In a possible implementation, the evaluation module 304 is further configured to perform differentiated parameter adjustments on the industry and region coefficients based on the enterprise classification and grading results to adapt to evaluation requirements of different industries and regions.

[0144] In a possible implementation, the extraction module 302 is specifically configured to perform Chinese word segmentation and stop word filtering on the enterprise public registration information;

[0145] The LDA model is used to extract the text topic distribution corresponding to the basic information of the enterprise from the processed public registration information of the enterprise, and generate word sequence features.

[0146] In one possible implementation, the evaluation module 304 is further configured to, upon receiving an enterprise credit evaluation instruction and / or an evaluation update timing instruction, repeatedly execute the multi-dimensional data association and fusion of the enterprise's public operational data using Flink CDC real-time stream processing technology and subsequent operations, and re-determine the enterprise credit evaluation result.

[0147] In this example, by extracting core evaluation indicators such as product unit price changes, upstream material prices, and sales volume from the fused data and calculating a comprehensive product strength data index based on these indicators, a company's market competitiveness and operational stability can be quantified. Compared to traditional evaluation based on a single financial indicator, this method combines multi-dimensional dynamic data such as price fluctuations, supply chain costs, and sales efficiency to accurately reflect a company's true operational capabilities, avoid credit misjudgments caused by one-sided data, and thus improve banks' controllability of credit risks for small and micro enterprises.

[0148] For the sake of convenience and brevity, the division of the above functional modules / units is only used as an example. In actual applications, the above functions can be assigned to different functional modules / units as needed. The above modules / units can be implemented in the form of hardware, software, or a combination of hardware and software.

[0149] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods in the above-mentioned method embodiments.

[0150] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the methods in the above-mentioned method embodiments.

[0151] The computer program includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. Computer-readable media may include any entity or device capable of carrying computer program code, recording media, USB flash drives, mobile hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunications signals, and software distribution media.

[0152] In the above embodiments, the descriptions of each embodiment have their own focus. For parts not described or recorded in detail in one embodiment, please refer to the relevant descriptions of other embodiments. Unless otherwise specified or there is a logical conflict, the terms and / or descriptions between different embodiments are consistent and can be referenced to each other. The technical features of different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0153] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A credit evaluation method for industrial enterprises based on multi-dimensional real-time streaming data, characterized in that: include: Based on the public registration information of enterprises, a preliminary classification of small and micro enterprises was carried out according to the national economic industry classification standards and the city and county regional divisions to obtain preliminary classification results; Using a linear discriminant analysis (LDA) model to extract text features of basic enterprise information from the publicly registered information of the enterprise, a word sequence feature representing the enterprise category is generated, and based on the similarity matching results between the word sequence feature and the preset industry theme, the enterprise is graded in combination with the preliminary classification results; wherein the basic enterprise information includes the enterprise name, registered address, business scope, and public introduction; Flink CDC real-time stream processing technology is used to perform multi-dimensional data association and fusion on the enterprise's public operational data to generate fused data, and the enterprise credit evaluation results are calculated based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

2. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 1 is characterized in that: The calculating of the enterprise credit evaluation result based on the fused data includes: Determine evaluation data of the enterprise to be evaluated based on the fused data, the evaluation data including: monthly change in product unit price, monthly change in upstream material price, monthly average product price, product sales volume, industry average price, and payment collection period; Calculate the comprehensive data index of product strength of the enterprise to be evaluated based on the evaluation data; The enterprise credit evaluation result is determined based on the comprehensive product strength data index.

3. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 2 is characterized in that: The calculation of the comprehensive product strength data index of the enterprise to be evaluated based on the evaluation data includes: Determine a product strength index based on the monthly average price of the product, the sales volume of the product, the industry average price, and the collection period; Determine the industry region coefficient based on the monthly change value of the product unit price and the monthly change value of the upstream material price; The product of the industry region coefficient and the product strength index is determined as the comprehensive product strength data index.

4. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 3 is characterized in that: The product strength index is determined based on the monthly average price of the product, product sales volume, industry average price and collection period, including: Calculate the product's monthly average price and the product's sales volume, and then divide the result by the product's average selling price and the collection period for similar products in the industry; The calculation result is recorded as the product strength index.

5. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 3 is characterized in that: Determining the industry region coefficient according to the monthly change value of the product unit price and the monthly change value of the upstream material price includes: Calculate the difference between the monthly change in the product unit price and the monthly change in the upstream material price; The difference is recorded as the industry region coefficient.

6. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 1 is characterized in that: The method of extracting text features of basic enterprise information from the public enterprise registration information using the linear discriminant analysis (LDA) model includes: Perform Chinese word segmentation and stop word filtering on the publicly registered information of the enterprise; The LDA model is used to extract the text topic distribution corresponding to the basic information of the enterprise from the processed public registration information of the enterprise, and generate word sequence features.

7. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 6 is characterized in that: The stop word filtering includes: Remove punctuation and the default Chinese stop word list, and merge synonyms and antonyms.

8. The industrial enterprise credit evaluation method based on multi-dimensional real-time streaming data according to claim 1 is characterized in that: Also includes: When an enterprise credit evaluation instruction and / or an evaluation update timing instruction is received, the multi-dimensional data association and fusion of the enterprise's public operation data using Flink CDC real-time stream processing technology and subsequent operations are repeatedly performed to re-determine the enterprise credit evaluation result.

9. An industrial enterprise credit evaluation device based on multi-dimensional real-time streaming data, characterized in that: include: The preliminary classification module is used to conduct preliminary classification of small and micro enterprises based on the publicly registered information of enterprises, according to the national economic industry classification standards and the city and county regional divisions, and obtain preliminary classification results; The extraction module is used to extract the text features of the basic information of the enterprise from the public registration information of the enterprise using the LDA model to generate word sequence features representing the enterprise category. A grading module, configured to grade enterprises based on the similarity matching results between the word sequence features and the preset industry themes, combined with the preliminary classification results; wherein the basic enterprise information includes the enterprise name, registered address, business scope and public introduction; The evaluation module is used to perform multi-dimensional data correlation and fusion on the enterprise's public operational data through Flink CDC real-time stream processing technology to generate fused data, and calculate the enterprise credit evaluation results based on the fused data; wherein the public operational data includes at least one of the enterprise's operating data, bidding transaction data, and e-commerce platform transaction data.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images