Method and device for identifying industry classification information of enterprise, and program product

Abstract

The invention discloses a method and device for identifying industry classification information of an enterprise and a program product, and relates to the field of big data, and the method comprises the steps: obtaining invoice data of the enterprise from an external data source, and extracting a commodity information set contained in the invoice data; converting the commodity information set of each enterprise into a first word vector set through the trained word vector model, and converting the industry classification information of each enterprise into a second word vector through the word vector model; the second word vector and the first word vector set are calculated in a weighted similarity calculation mode, and the similarity between each enterprise and the industry classification information of each enterprise is obtained; and identifying whether the industry classification information of the N enterprises is correct or not according to the similarity to obtain identification results of the N enterprises. According to the method and the device, the problem that errors exist in industry classification information autonomously registered by an enterprise or industry classification information determined based on personal experience in public credit business of a financial institution in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of big data, and more specifically, to a method, device and program product for identifying industry classification information of an enterprise. Background Art

[0002] The banking industry's corporate credit work includes an important step in determining the industry classification of enterprises. The credit policies and credit calculation logic of different industries may vary greatly. When enterprises register their industry categories in industrial and commercial information, they have greater autonomy, and the registration quality varies, and there are often errors.

[0003] In the existing technology, in the banking industry's corporate credit work, account managers often rely on personal experience to collect various information to accurately judge the industry classification of enterprises. For example, the enterprise proactively provides relevant qualifications, purchase and sales contracts, invoices, patent information, software works, etc., and the account manager judges whether the industry category registered by the enterprise is accurate and appropriate. However, this method relies heavily on the personal experience of the account manager and the materials provided by the customer, resulting in low recognition accuracy and low reliability. At the same time, the reliance on manual methods is time-consuming and inefficient, and it is impossible to process a large number of requests at the same time, resulting in low recognition efficiency.

[0004] With regard to the problem in related technologies that in the corporate credit business of financial institutions, the industry classification information registered independently by enterprises or determined based on personal experience is erroneous, no effective solution has been proposed so far. Summary of the invention

[0005] The main purpose of this application is to provide a method, device and program product for identifying the industry classification information of an enterprise, so as to solve the problem in the related technology that the industry classification information registered by the enterprise independently or determined based on personal experience in the corporate credit business of financial institutions is erroneous.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for identifying the industry classification information of an enterprise is provided, the method comprising: for each of N enterprises, obtaining the invoice data of the enterprise from an external data source, and extracting the commodity information contained in the invoice data to obtain a commodity information set of each enterprise, wherein N is a positive integer; converting the commodity information set of each enterprise into a first word vector set through a trained word vector model, and converting the industry classification information of each enterprise into a second word vector through the word vector model; calculating the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between each enterprise and each enterprise's industry classification information; identifying whether the industry classification information of the N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtaining the identification results of the N enterprises.

[0007] Furthermore, the second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise, including: calculating the distance between each first word vector in the first word vector set and the second word vector in the word vector space to obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed based on the word vector model; determining the weight corresponding to each first word vector based on the number of transactions of the commodity corresponding to each first word vector in the first word vector set; calculating the similarity corresponding to the second word vector based on the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; determining the similarity between the industry classification information of each enterprise and each enterprise based on the similarity between the second word vector corresponding to the industry classification information of each enterprise and the second word vector.

[0008] Furthermore, whether the industry classification information of the N enterprises is correct is identified based on the similarity between the industry classification information of each enterprise, and the identification results of the N enterprises are obtained, including: determining a target similarity that is less than a preset value in the similarity between the industry classification information of each enterprise; determining a target enterprise corresponding to the target similarity, and determining that the industry classification information of the target enterprise is incorrect; generating an error classification report based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise, and obtaining the identification results of the N enterprises.

[0009] Furthermore, after identifying whether the industry classification information of the N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtaining the identification results of the N enterprises, the method also includes: determining the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and M industry classification information, and updating the industry classification information of the target enterprise based on the target industry classification information, wherein M is a positive integer; determining the historical credit behavior of the target enterprise, and evaluating the credit risk of the target enterprise based on the historical credit behavior; and adjusting the credit strategy of the target enterprise based on the credit risk of the target enterprise.

[0010] Furthermore, obtaining the invoice data of the enterprise from the external data source includes: receiving a login request of the enterprise through a front-end web page, and verifying the login request of the enterprise to obtain a verification result; when the verification result indicates that the enterprise has passed the verification, receiving the authorization information of the enterprise, wherein the authorization information is credential information for obtaining the invoice data of the enterprise from the external data source; generating a data acquisition request based on the HTTP protocol and the authorization information, and sending the data acquisition request to the external data source; receiving return data from the external data source in response to the data acquisition request, and determining the invoice data of the enterprise within a preset time period based on the return data.

[0011] Furthermore, the product information contained in the invoice data is extracted to obtain a product information set for each enterprise, including: filtering the invoice data to obtain sales invoice data; determining the product names contained in the sales invoice data to obtain a product name set; sorting each product name in the product name set according to product attributes to obtain each sorted product name, and determining a product information set for each enterprise based on each sorted product name, wherein the product attributes include at least one of the following: the transaction frequency of the product corresponding to the product name, the transaction amount of the product corresponding to the product name, and the product name.

[0012] Furthermore, before converting the product information set of each enterprise into a first word vector set through the trained word vector model, and converting the industry classification information of each enterprise into a second word vector through the word vector model, the method also includes: acquiring Chinese corpus through a preset channel to obtain a corpus set; performing preprocessing operations on the text information in the corpus set to obtain a word set, wherein the preprocessing operations at least include: font conversion and word segmentation processing; inputting the word set into an untrained word vector model; analyzing the semantic similarity between words in the word set through the untrained word vector model, constructing a model parameter set, and constructing the trained word vector model based on the model parameter set.

[0013] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for identifying the industry classification information of an enterprise is provided, and the device includes: an extraction unit, which is used to obtain the invoice data of each enterprise in N enterprises from an external data source, and extract the commodity information contained in the invoice data to obtain a commodity information set of each enterprise, wherein N is a positive integer; a conversion unit, which is used to convert the commodity information set of each enterprise into a first word vector set through a trained word vector model, and convert the industry classification information of each enterprise into a second word vector through the word vector model; a calculation unit, which is used to calculate the second word vector and each first word vector in the first word vector set by weighted calculation of similarity to obtain the similarity between the industry classification information of each enterprise and each enterprise; an identification unit, which is used to identify whether the industry classification information of the N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtain the identification results of the N enterprises.

[0014] Furthermore, the calculation unit includes: a first calculation subunit, used to calculate the distance between each first word vector in the first word vector set and the second word vector in the word vector space, and obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed based on the word vector model; a first determination subunit, used to determine the weight corresponding to each first word vector based on the number of transactions of the commodity corresponding to each first word vector in the first word vector set; a second calculation subunit, used to calculate the similarity corresponding to the second word vector based on the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; a second determination subunit, used to determine the similarity between each enterprise and the industry classification information of each enterprise based on the second word vector corresponding to the industry classification information of each enterprise and the similarity corresponding to the second word vector.

[0015] Furthermore, the identification unit includes: a third determination subunit, used to determine a target similarity that is less than a preset value in the similarity between each enterprise and the industry classification information of each enterprise; a fourth determination subunit, used to determine the target enterprise corresponding to the target similarity, and determine that the industry classification information of the target enterprise is incorrect; a first generation subunit, used to generate an error classification report based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise, to obtain the identification results of the N enterprises.

[0016] Furthermore, the device also includes: a first determination unit, used to identify whether the industry classification information of the N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and after obtaining the identification results of the N enterprises, determine the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and the M industry classification information, and update the industry classification information of the target enterprise based on the target industry classification information, wherein M is a positive integer; a second determination unit, used to determine the historical credit behavior of the target enterprise, and evaluate the credit risk of the target enterprise based on the historical credit behavior; an adjustment unit, used to adjust the credit strategy of the target enterprise based on the credit risk of the target enterprise.

[0017] Furthermore, the extraction unit includes: a verification subunit, used to receive a login request from the enterprise through a front-end web page, and verify the login request from the enterprise to obtain a verification result; a receiving subunit, used to receive authorization information from the enterprise when the verification result indicates that the enterprise has passed the verification, wherein the authorization information is credential information for obtaining the invoice data of the enterprise from the external data source; a second generating subunit, used to generate a data acquisition request based on the HTTP protocol and the authorization information, and send the data acquisition request to the external data source; a fifth determining subunit, used to receive return data from the external data source in response to the data acquisition request, and determine the invoice data of the enterprise within a preset time period based on the return data.

[0018] Furthermore, the extraction unit includes: a screening subunit, used to screen the invoice data to obtain sales invoice data; a sixth determination subunit, used to determine the product names contained in the sales invoice data to obtain a product name set; a sorting subunit, used to sort each product name in the product name set according to product attributes to obtain each sorted product name, and determine a product information set for each enterprise based on each sorted product name, wherein the product attributes include at least one of the following: the transaction frequency of the product corresponding to the product name, the transaction amount of the product corresponding to the product name, and the product name.

[0019] Furthermore, the device also includes: an acquisition unit, which is used to acquire Chinese corpus through a preset channel to obtain a corpus set before converting the product information set of each enterprise into a first word vector set through a trained word vector model and converting the industry classification information of each enterprise into a second word vector through the word vector model; a processing unit, which is used to perform preprocessing operations on the text information in the corpus to obtain a word set, wherein the preprocessing operations at least include: font conversion and word segmentation processing; an input unit, which is used to input the word set into an untrained word vector model; a construction unit, which is used to analyze the semantic similarity between words in the word set through the untrained word vector model, construct a model parameter set, and construct the trained word vector model based on the model parameter set.

[0020] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer program product is provided, including a computer program, which, when executed by a processor, implements the method for identifying the industry classification information of an enterprise described in any of the above items, and which, when executed by a processor, implements the steps of the method for identifying the industry classification information of an enterprise described in each embodiment of the present application.

[0021] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes stored computer instructions, wherein when the computer instructions are executed by a processor, a method for identifying the industry classification information of any of the above-mentioned enterprises is implemented.

[0022] In order to achieve the above-mentioned purpose, according to one aspect of the present application, an electronic device is provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement the method for identifying the industry classification information of the enterprise described in any one of the above items.

[0023] In an embodiment of the present application, for each of N enterprises, the invoice data of the enterprise is obtained from an external data source, and the commodity information contained in the invoice data is extracted to obtain a commodity information set of each enterprise, wherein N is a positive integer; the commodity information set of each enterprise is converted into a first word vector set through a trained word vector model, and the industry classification information of each enterprise is converted into a second word vector through the word vector model; the second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between the industry classification information of each enterprise and each enterprise; based on the similarity between the industry classification information of each enterprise and each enterprise, it is identified whether the industry classification information of the N enterprises is correct, and the identification results of the N enterprises are obtained, thereby solving the technical problem that in the corporate credit business of financial institutions, the industry classification information registered by enterprises themselves or determined based on personal experience is erroneous.

[0024] By obtaining the invoice data of each enterprise from external data sources and extracting key product information, it is possible to construct a product information set that reflects the actual operating status of the enterprise. Through the trained word vector model, the product information and the registered industry classification of the enterprise are converted into word vectors, and the weighted similarity calculation is introduced. It is possible to comprehensively consider a variety of product information and accurately measure the degree of fit between the first word vector set and the second word vector, thereby quantitatively evaluating the accuracy of each enterprise's industry classification, reducing dependence on manual judgment, and significantly improving the accuracy and efficiency of financial institutions in identifying enterprise industry classification information. It achieves the effect of automatically identifying and marking enterprises with potentially incorrect industry classifications, greatly optimizing the credit decision-making process of financial institutions and enhancing the risk control capabilities of financial institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0026] Figure 1 It is a hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for identifying industry classification information of an enterprise provided in accordance with the first embodiment of the present application;

[0027] Figure 2 is a flowchart of an optional method for identifying industry classification information of an enterprise provided in Embodiment 1 of the present application;

[0028] Figure 3 It is a schematic diagram of the overall process of optional identification of industry classification information of an enterprise provided according to the first embodiment of the present application;

[0029] Figure 4It is a timing diagram of collecting invoice data by an optional financial institution business system provided in Embodiment 1 of the present application;

[0030] Figure 5 is a structural diagram of an optional neural network provided according to Embodiment 1 of the present application;

[0031] Figure 6 is a schematic diagram of an optional word vector space provided according to Embodiment 1 of the present application;

[0032] Figure 7 is a schematic diagram of a device for identifying industry classification information of an enterprise provided in Embodiment 2 of the present application;

[0033] Figure 8 This is a schematic diagram of an electronic device for identifying the industry classification information of an enterprise provided in Example 5 of the present application. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] It should be noted that the methods and devices determined by the processing methods, devices, storage media and electronic devices of the present application documents can be used in the field of financial technology to improve the recognition accuracy and efficiency in the process of identifying misclassified enterprise industry classification information, and can also be used in any field other than the field of financial technology. The application fields of the methods and devices of the processing methods, devices, storage media and electronic devices of the present application documents are not limited.

[0036] It should be noted that the collected information (including but not limited to user device information, user personal information, collected data, used data, generated data, processed data, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, collected information, used information, generated information, processed information, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data are in compliance with the relevant laws, regulations and standards of relevant countries and regions, necessary confidentiality measures are taken, and public order and good customs are not violated, and corresponding operation entrances are provided for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation entrances for users to choose to agree or refuse the results of automated decision-making; if the user chooses to refuse, the expert decision-making process will be entered.

[0037] Example 1

[0038] According to an embodiment of the present application, a method embodiment for identifying the industry classification information of an enterprise is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0039] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal (or mobile device) for implementing a method for identifying industry classification information of an enterprise is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (102a, 102a, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (US) port (which may be included as one of the ports of the US bus), a network interface, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0040] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as program instructions / data storage devices corresponding to the method for identifying the industry classification information of an enterprise in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, the above-mentioned method for identifying the industry classification information of an enterprise is realized. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0043] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0044] Under the above operating environment, this application provides Figure 2 A method for identifying the industry classification information of the enterprise shown. Figure 2 This is a flowchart of a method for identifying industry classification information of an enterprise according to Example 1 of the present application.

[0045] Step S101, for each of the N enterprises, obtain the enterprise's invoice data from an external data source, and extract the commodity information contained in the invoice data to obtain a commodity information set for each enterprise, where N is a positive integer.

[0046] In this first embodiment, in order to improve the accuracy of the industry classification of enterprises, it is necessary to obtain the invoice data authorized by each enterprise from an external data source, namely the tax system. The detailed commodity information in the invoice data is an important basis for judging the actual business industry of the enterprise. Exemplarily, the execution subject of the method for identifying the industry classification information of the enterprise in this embodiment 1 can be an identification system, and the identification system can guide the enterprise to complete the tax system authorization process through the HTTP protocol to ensure legal and compliant access to its tax records.

[0047] Then, the acquired invoice data is parsed, focusing on extracting the product names in the invoice data, identifying specific products and service items, and forming a product information set for each enterprise. The product information set not only includes the product name or service item name, but also key information such as transaction quantity and transaction amount, thus providing rich information for subsequent industry classification.

[0048] Through the above steps, a real portrait can be constructed for each enterprise based on the actual sales product information, so that errors in the enterprise industry classification can be identified based on the real portrait, thereby achieving the effect of improving the accuracy of enterprise classification.

[0049] Step S102: convert the product information set of each enterprise into a first word vector set through the trained word vector model, and convert the industry classification information of each enterprise into a second word vector through the word vector model.

[0050] In the first embodiment of the present invention, the trained word vector model is used to map the product name into the pre-trained word vector space so that each product is represented as a vector of fixed length, that is, the product information set extracted from the invoice data of N companies is converted into the first word vector set. In this conversion process, these vectors capture the similarities and differences of the products at the semantic level. The word vector conversion of the product information set not only considers the direct semantic representation of the product name, but also utilizes the relationship between words in the vector space, thereby providing a quantitative and structured data basis for subsequent industry classification analysis.

[0051] At the same time, by mapping the text description of the industry classification into the same word vector space, the industry classification information of each enterprise is converted into the second word vector corresponding to each enterprise. The second word vector represents the semantic representation of the industry classification registered by each enterprise or preset by each enterprise, which is convenient for semantic comparison and similarity calculation with the commodity information in the first word vector set.

[0052] By converting the company's product information and industry classification into vector representations of the word vector model respectively, it is possible to quantitatively compare the differences between the company's actual business operations and the registered industry in the multi-dimensional vector space, providing an innovative method based on data science and artificial intelligence technology to identify industry classification errors.

[0053] Step S103, calculating the similarity between the second word vector and each first word vector in the first word vector set by weighted similarity calculation, and obtaining the similarity between each enterprise and the industry classification information of each enterprise.

[0054] In the first embodiment of the present invention, the degree of match between the industry classification information (second word vector) of each enterprise and the word vector representation of the enterprise's product information set (first word vector set) can be evaluated by weighted similarity calculation method. Specifically, by calculating the similarity between the second word vector and each first word vector in the first word vector set, it is possible to quantitatively evaluate the degree of fit between the enterprise's industry classification and the actual business operations, and perform weighted calculation on the similarity corresponding to each first word vector to improve the accuracy of the calculation results. The high or low similarity value can help determine whether the industry classification is accurate. A low value may mean that there is an error in the industry classification and further review and adjustment is required.

[0055] Step S104, identifying whether the industry classification information of N enterprises is correct based on the similarity between each enterprise and the industry classification information of each enterprise, and obtaining the identification results of the N enterprises.

[0056] In the first embodiment of the present invention, for each enterprise (i.e., each of the N enterprises), by calculating the multi-dimensional semantic similarity between its commodity information set (first word vector set) and industry classification information (second word vector), the accuracy of the enterprise industry classification can be identified. For the semantic similarity corresponding to each enterprise, its similarity score can be compared with the preset industry classification similarity threshold to obtain the identification result of each enterprise. For example, if the similarity score of an enterprise is higher than the set similarity threshold, it means that its industry classification is highly consistent with the actual business commodity information in semantics, and thus the classification information is considered to be correct. On the contrary, if the similarity score is lower than the threshold, the system will mark its industry classification as suspicious or wrong, requiring further manual review or error classification correction.

[0057] Based on the industry classification identification results of N enterprises, the similarity score of each enterprise and the judgment result of whether the classification is correct or not can be listed to evaluate the accuracy of industry classification in the enterprise's credit risk, thereby optimizing the financial institution's credit policy and management strategy for the enterprise, significantly improving the efficiency and accuracy of credit business work.

[0058] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, the second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise, including: calculating the distance between each first word vector and the second word vector in the first word vector set in the word vector space to obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed based on the word vector model; determining the weight corresponding to each first word vector based on the number of transactions of the commodity corresponding to each first word vector in the first word vector set; calculating the similarity corresponding to the second word vector based on the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; determining the similarity between the industry classification information of each enterprise and each enterprise based on the similarity between the second word vector corresponding to the industry classification information of each enterprise and the second word vector.

[0059] In the first embodiment, in order to improve the accuracy of evaluating the enterprise industry classification information, the similarity and weight between the first word vector set and the second word vector set may be calculated.

[0060] In an optional embodiment, first, based on a word vector model trained by natural language processing technology such as Word2Vec, a multidimensional word vector space is constructed, in which each word (including product name and industry classification description) is represented as a vector, and these vectors are arranged in the space according to semantic relationships. The distance between the second word vector in the word vector space and each first word vector in the first word vector set is calculated (for example, cosine similarity, Euclidean distance, etc.), and the smaller the distance, the higher the similarity.

[0061] Then, the number of transactions of the commodity corresponding to each first word vector in the corresponding enterprise operation process is calculated. The number of transactions reflects the contribution of the commodity to the enterprise's business activities. According to the number of transactions of the commodity, a weight is assigned to each first word vector. Among them, since the commodity transactions with more transactions can better represent the main business of the enterprise, the commodity transactions with more transactions will receive a higher weight.

[0062] Secondly, based on the similarity scores calculated between each first word vector and the second word vector, and the weight corresponding to each first word vector, a weighted average calculation is performed to obtain the similarity corresponding to the second word vector. Specifically, the similarity score of each first word vector is multiplied by its corresponding weight, and then the weighted similarity scores of all first word vectors are added and divided by the total weight of the first word vector set to obtain a value reflecting the overall semantic similarity between the second word vector and the first word vector set, that is, the similarity corresponding to the second word vector mentioned above.

[0063] Finally, the similarity between the industry classification information of each enterprise (represented by the second word vector) and the enterprise product information set (represented by the weighted first word vector set) is determined based on the similarity corresponding to each second word vector. The similarity corresponding to each second word vector directly reflects the consistency between the industry classification information of the enterprise and its actual business operations.

[0064] By calculating the semantic similarity between word vectors, we can quantify the consistency between an enterprise’s industry classification information and its actual business operations. By assigning word vector weights based on the number of transactions, we can quantify the importance of an enterprise’s business operations, thereby achieving an accuracy assessment of the enterprise’s industry classification information and providing financial institutions with more efficient and accurate credit risk management.

[0065] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, whether the industry classification information of N enterprises is correct is identified based on the similarity between the industry classification information of each enterprise, and the identification results of N enterprises are obtained, including: determining a target similarity that is less than a preset value in the similarity between the industry classification information of each enterprise; determining a target enterprise corresponding to the target similarity, and determining that the industry classification information of the target enterprise is incorrect; generating an error classification report based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise, and obtaining the identification results of N enterprises.

[0066] In the first embodiment, in order to accurately identify enterprises with incorrect industry classification, an incorrect classification report can be generated to directly reflect the enterprises with inaccurate industry classification, thereby providing a basis for subsequent credit strategy decisions and risk management.

[0067] Specifically, a preset value can be set in advance as a criterion for judging the correctness of industry classification, and the value is determined based on a comprehensive consideration of semantic similarity and industry classification accuracy. The preset value reflects the minimum requirement for industry classification information and the actual business commodity information of the enterprise to be judged to be similar.

[0068] For each enterprise, the similarity between its industry classification information (second word vector) and the actual operating commodity information set (first word vector set) is calculated to identify all similarities with similarity scores lower than the preset value. These scores are marked as target similarities, which means that the enterprise industry classification may be significantly different from the actual operating situation.

[0069] For each target similarity, the second word vector corresponding to the target similarity is determined, and then the target enterprise corresponding to the second word vector corresponding to the target similarity is determined, and the industry classification information of the target enterprise is marked as suspicious classification or wrong classification. After marking, it can be reconfirmed through manual review to further confirm the accuracy of the industry classification.

[0070] All target enterprises whose industry classification information is marked as suspicious or misclassified and their related information are summarized (including enterprise name, registered industry category, actual business commodity information (first word vector set), and similarity score with industry classification information, etc.), and an error classification report is generated, which lists in detail the industry classification errors of each target enterprise and the degree of matching with the actual business commodity information, that is, the identification results of the N enterprises mentioned above.

[0071] Through this series of steps, target enterprises with suspicious or erroneous industry classification information can be accurately screened out from N enterprises, and an error classification report can be generated, enabling financial institutions to automatically identify and correct industry classification errors, thereby improving the efficiency and accuracy of credit work. At the same time, it also fully considers the complexity of the actual operation of enterprises and the dynamic changes of industry classification information, thereby improving the accuracy of financial institutions in identifying the industry classification information of enterprises.

[0072] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, after identifying whether the industry classification information of N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtaining the identification results of N enterprises, the above method also includes: determining the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and M industry classification information, and updating the industry classification information of the target enterprise based on the target industry classification information, where M is a positive integer; determining the historical credit behavior of the target enterprise, and evaluating the credit risk of the target enterprise based on the historical credit behavior; and adjusting the credit strategy of the target enterprise based on the credit risk of the target enterprise.

[0073] In the first embodiment, for target enterprises that are identified as having inaccurate industry classification information, it is necessary to re-evaluate the industry classification and adjust the credit strategy.

[0074] Specifically, a number of (i.e., the M above, where M is a positive integer) industry classification information can be pre-set as a reference benchmark, covering all possible industry categories, for semantic similarity comparison with the actual business commodity information of the target enterprise. For each target enterprise, based on the first word vector set corresponding to the commodity information set of the target enterprise and the second word vector set of the M industry classification information, the similarity between all the first word vectors in the first word vector set corresponding to the commodity information set of the target enterprise and the M second word vector sets corresponding to the M industry classification information in the second word vector set is calculated.

[0075] Among the M industry classification information, the industry classification with the highest similarity to the first word vector set of the target enterprise is selected as the target industry classification information. The target industry classification information is closer to the actual operating conditions of the target enterprise. Then, based on the target industry classification information, the industry classification information of the target enterprise in the bank system is updated to ensure that the credit decision is based on the most accurate industry information of the enterprise. For the target enterprise, the financial institution will retrieve its historical credit records, including loan frequency, repayment status, default history, etc., based on the updated industry classification information of the target enterprise, in order to fully understand the credit behavior of the enterprise. Based on the historical credit behavior of the target enterprise and combined with the newly determined industry classification information, the credit risk level of the target enterprise is recalculated using a risk assessment model (such as a credit scoring model, a default prediction model, etc.). According to the re-evaluated credit risk level, the credit strategy of the target enterprise is automatically adjusted, including but not limited to credit limit, loan interest rate, guarantee requirements, etc., to ensure that the credit policy matches the risk status of the enterprise. The updated credit strategy will be applied to the subsequent credit activities of the target enterprise to optimize the allocation of credit resources, reduce credit risks, and improve the management efficiency of the credit business of financial institutions.

[0076] Through the above steps, it is possible to effectively identify target enterprises that are incorrectly classified in the industry classification information of multiple enterprises, and to propose more accurate target industry classification information for the target enterprises. At the same time, by evaluating and adjusting credit strategies based on the latest industry classification information and historical credit behavior, it is possible to provide financial institutions with a dynamic and accurate credit risk management mechanism, which not only improves the efficiency of the credit work of financial institutions, but also enhances the control of financial institutions over the accuracy of industry classification information.

[0077] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, obtaining the invoice data of the enterprise from an external data source includes: receiving a login request from the enterprise through a front-end web page, and verifying the login request of the enterprise to obtain a verification result; when the verification result indicates that the enterprise has passed the verification, receiving the authorization information of the enterprise, wherein the authorization information is credential information for obtaining the invoice data of the enterprise from an external data source; generating a data acquisition request based on the HTTP protocol and the authorization information, and sending the data acquisition request to the external data source; receiving return data from the external data source in response to the data acquisition request, and determining the invoice data of the enterprise within a preset time period based on the return data.

[0078] In the first embodiment, the front-end webpage of the financial institution business system serves as the entrance for the enterprise to interact with the banking system and is responsible for receiving login requests from the enterprise. The enterprise usually needs to provide login credentials (such as user name and password) and log in to the financial institution according to the login credentials. After receiving the login request, the financial institution business system will verify the login credentials of the enterprise to confirm whether the login request comes from a registered corporate account. The verification process may include matching with the user information stored in a preset storage space (e.g., a database) to ensure the accuracy and security of the account.

[0079] After the enterprise login verification is passed, the authorization information provided by the enterprise is further received. This information is the credential for the enterprise to obtain invoice data from an external data source (such as a tax system). For example, it contains an authorization code or a specific API key, etc., which allows the business system of a financial institution to access and obtain the invoice data of the enterprise in a legal manner. Based on the received authorization information, the HTTP protocol is used to construct a data acquisition request, which is a set of standardized network communication instructions for requesting data from an external data source. After the construction is completed, the data acquisition request is sent to the external data source, that is, the tax system, through the network to request the invoice information within the scope of the enterprise's authorization.

[0080] After receiving the data acquisition request, the external data source will verify the validity of the request based on the authorization information, and return the invoice data within the enterprise's authorization scope as response data to the financial institution business system. The invoice data received by the financial institution business system may contain information within a longer time range, so it is necessary to filter and determine the data based on a preset time period (such as the most recent quarter or the most recent year). The selection of this time period is intended to ensure that the information obtained is the latest and can reflect the recent operating conditions of the enterprise. It can be flexibly adjusted according to the actual production situation, and is not specifically limited in this embodiment 1.

[0081] Through this series of steps, the business system of financial institutions can safely and compliantly obtain the latest and accurate invoice data of enterprises, providing a solid data foundation for the subsequent identification of industry classification errors and enhancing the accuracy and reliability of credit decisions. By making full use of information and network technologies, the automatic acquisition of enterprise data is realized, which greatly improves the efficiency and accuracy of bank credit work. At the same time, through strict verification and authorization processes, the security and privacy protection of enterprise data are guaranteed, meeting data compliance requirements.

[0082] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, the commodity information contained in the invoice data is extracted to obtain a commodity information set of each enterprise, including: screening the invoice data to obtain sales invoice data; determining the commodity names contained in the sales invoice data to obtain a commodity name set; sorting each commodity name in the commodity name set according to commodity attributes to obtain each sorted commodity name, and determining the commodity information set of each enterprise based on each sorted commodity name, wherein the commodity attributes include at least one of the following: the transaction frequency of the commodity corresponding to the commodity name, the transaction amount of the commodity corresponding to the commodity name, and the commodity name.

[0083] In the first embodiment, in order to extract and construct the enterprise's commodity information set from the invoice data, first, the received invoice data can be filtered to retain only the sales invoice data, that is, the invoice information issued when the enterprise sells the goods. This screening process excludes the input invoice data obtained by the enterprise when purchasing goods, ensuring that the subsequent analysis focuses on the main business and sales behavior of the enterprise.

[0084] From the filtered sales invoice data, the commodity names recorded on each invoice are extracted to form a commodity name set. The commodity name set includes the names of all commodities (or service items) sold by the enterprise, which is used for subsequent analysis of the enterprise's business scope.

[0085] For each product name in the product name set, in order to determine the main business and industry classification information of the enterprise, it is necessary to count and evaluate the transaction frequency and transaction amount of the enterprise's sales products. Among them, the transaction frequency can reflect the importance of the product in the enterprise's sales, and the transaction amount is directly related to the contribution of the product to the enterprise's revenue.

[0086] The product names in the product name set are sorted according to the transaction frequency and transaction amount of the products. The sorting results can intuitively display the core sales products of the enterprise and the sales products that have the greatest impact on the financial status of the enterprise. The sorted product names are further integrated to construct a product information set for each enterprise. The generated product information set not only contains the names of all the products sold by the enterprise, but also marks the transaction frequency, transaction amount, and sorting position of each product, providing a detailed product information reference for subsequent industry classification error identification.

[0087] Through this series of steps, the company's commodity information set can be accurately extracted from the original invoice data, which not only fully reflects the types of goods sold by the company, but also quantifies the importance of each commodity to the company's operations, providing strong data support for subsequent analysis and decision-making, and ensuring the automation and efficiency of the construction from invoice data to commodity information sets. At the same time, through the statistics and evaluation of commodity attributes, the data value of the commodity information set is improved, and the accuracy of the recognition results of the company's industry classification information is further improved.

[0088] Optionally, in the method for identifying the industry classification information of an enterprise provided in Example 1 of the present application, before converting the product information set of each enterprise into a first word vector set through a trained word vector model, and converting the industry classification information of each enterprise into a second word vector through a word vector model, the above method also includes: obtaining Chinese corpus through a preset channel to obtain a corpus set; performing preprocessing operations on the text information in the corpus set to obtain a word set, wherein the preprocessing operations include at least: font conversion and word segmentation processing; inputting the word set into an untrained word vector model; analyzing the semantic similarity between words in the word set through an untrained word vector model, constructing a model parameter set, and constructing a trained word vector model based on the model parameter set.

[0089] In the first embodiment of the present invention, in order to construct and train a word vector model to analyze Chinese semantics, a large amount of Chinese text data can be collected through preset channels (such as Internet public resources, professional databases, Chinese encyclopedia entries, multiple public corpora, etc., to ensure the diversity and authority of the corpus) to form a corpus.

[0090] Preprocess the text information in the corpus, mainly including font conversion and word segmentation. Font conversion ensures the consistency of simplified and traditional Chinese, avoiding analysis errors caused by font differences. Word segmentation breaks the text into independent words to prepare for subsequent word vector model training.

[0091] After preprocessing, the resulting word set is input into an untrained word vector model. This model uses natural language processing techniques such as Word2Vec to analyze the semantic relationship between words. During the training process, the word vector model analyzes the semantic similarity between words in the word set. This process is based on neural network technology. Through a large number of iterative learning, the model parameters are gradually optimized to reflect the true semantic relationship between words. After sufficient training, the word vector model will produce a set of optimal model parameters. These parameters integrate the semantic similarity information between words and form the basis of the trained word vector model.

[0092] Finally, based on the obtained model parameter set, a trained word vector model is constructed. This word vector model can convert words into multi-dimensional vectors and represent words in vector space, providing a semantic analysis basis for subsequent enterprise industry classification error identification.

[0093] Through the above steps, semantic information can be effectively extracted from Chinese corpus, and a word vector model can be constructed through training, which can perform deep semantic analysis on words, providing a solid algorithmic foundation for subsequent industry classification error identification. The entire process combines data acquisition, preprocessing, model training and parameter optimization, aiming to establish an efficient and accurate Chinese semantic understanding model, thereby improving the efficiency and accuracy of financial institutions in identifying the industry classification information of enterprises.

[0094] Optionally, in this embodiment 1, Figure 3 It is a schematic diagram of the overall process of identifying the industry classification information of an enterprise. Figure 3 It includes three main stages: data collection, model training and model operation. Figure 3 The process begins with the corporate customer initiating an authorization request through the front-end web page, allowing the bank system to access its invoice data in the tax department. This authorization is completed through a web interface deployed on various terminals or mobile H5 using the HTTP protocol, ensuring the standardization and security of data requests.

[0095] After receiving the customer's authorization, the financial institution's business system interacts with the tax system, and the tax system pushes the authorized enterprise's invoice data to the financial institution's business system. The invoice data is stored in a structured form, contains rich transaction information, and is the basis for subsequent analysis. The financial institution's business system pre-processes the received invoice data, including screening sales invoices, eliminating useless information, and summarizing and sorting by product name, forming a clear and structured set of product information for each enterprise, providing high-quality input for subsequent model training.

[0096] A large amount of Chinese public corpus is obtained from authoritative channels, and the collected corpus is segmented and decomposed into independent words. These words are then input into an untrained word vector model for training. The model analyzes the co-occurrence frequency and semantic relationship between words, learns the representation of words in a multidimensional vector space, and finally constructs a word vector model that can reflect Chinese semantics. After the training is completed, the parameters of the word vector model are saved to obtain a set of model parameters. These parameters contain the rules for mapping words to vector space and are the key basis for the operation of the model. The trained word vector model is loaded, and the product information set of each enterprise is batch processed. The product names in the product information set are converted into corresponding word vectors to prepare for subsequent analysis and comparison. The similarity between the word vectors of the enterprise industry classification information and the product name is calculated, and weighted according to the number of product names and the proportion of total sales to generate the final industry classification recognition result. This result reflects the degree of consistency between the actual business situation of the enterprise and the registered industry classification, and provides a quantitative basis for identifying industry classification errors.

[0097] Optionally, in this embodiment 1, Figure 4 It is a time sequence diagram of the collection of invoice data by the business system of a financial institution, showing the data flow and authorization interaction process among the enterprise, the business system of the financial institution and the tax system. Figure 4 The starting point is that the enterprise accesses the business system of the financial institution, usually through a front-end H5 web interface. During the interaction between the enterprise and the business system, the customer will be guided to an H5 web page, which provides a secure interface that allows the customer to directly initiate an authorization request to the tax system. The purpose of the authorization request is to allow the financial institution to legally access the customer's invoice data in the tax department. When the tax system confirms the identity of the enterprise and completes the authorization process, it will return a message of successful authorization to the business system of the financial institution. After receiving this message, the business system of the financial institution will promptly notify the enterprise that its authorization has been accepted and prepare for the next step of data collection. Once the enterprise is authorized successfully, the tax system will actively push the relevant invoice data to the business system of the financial institution. Finally, the business system of the financial institution is responsible for receiving and processing the invoice data from the tax system. After receiving the data, the system will perform a series of preprocessing, including screening, integration, formatting and other operations to meet the needs of subsequent data analysis and model training.

[0098] Optionally, in this embodiment 1, Figure 5 It is a structural diagram of a neural network, which is divided into an input layer, a hidden layer, and an output layer. The input layer contains input nodes x1, x2, x3, ..., x k , each input node is connected to all nodes in the hidden layer. The hidden layer contains hidden nodes h1, h2, h3, ..., h nEach hidden node receives input from the input layer and passes it through the weight w ki The output layer contains output nodes y1, y2, y3, ..., y m Each output node receives input from the hidden layer and transmits it through a weight w ′ ij Weighted. Figure 5 It shows the connections and data flow between layers in a neural network, as well as the role of weights in data transmission.

[0099] Optionally, in this embodiment 1, Figure 6 It is a schematic diagram of the word vector space, which contains a three-dimensional coordinate system, which contains multiple entities and their relationships. Entities in different directions include, but are not limited to: chips, computers, optical fibers, memory, communications, steel, aluminum alloys, coatings, treatment agents, chemicals, mining and ores, connecting rods, automobiles, parts, seeds, fertilizers, and transmitters. This distribution reflects the characteristics of words in different semantic dimensions, making words with similar semantics close in space, and words with different semantics far apart.

[0100] Figure 6 The relative position relationship between words in the word vector space intuitively reflects their semantic similarity in the word vector space. For example, words such as computer, chip, and memory are adjacent on the Z axis, indicating that they have a close semantic relationship in the field of information technology. This distribution based on the word vector space provides a powerful analysis tool for subsequent industry classification error identification, which can help the system understand the potential semantics of the product names in the enterprise invoice data and evaluate the industry matching degree accordingly.

[0101] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0102] In summary, the method for identifying the industry classification information of an enterprise provided in an embodiment of the present application obtains the invoice data of each of N enterprises from an external data source, and extracts the commodity information contained in the invoice data to obtain a commodity information set of each enterprise, wherein N is a positive integer; the commodity information set of each enterprise is converted into a first word vector set through a trained word vector model, and the industry classification information of each enterprise is converted into a second word vector through a word vector model; the second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between the industry classification information of each enterprise and each enterprise; based on the similarity between the industry classification information of each enterprise and each enterprise, it is identified whether the industry classification information of N enterprises is correct, and the identification results of N enterprises are obtained, which solves the problem in the related technology that the industry classification information registered by the enterprise itself or the industry classification information determined based on personal experience in the public credit business of financial institutions is incorrect.

[0103] By obtaining the invoice data of each enterprise from external data sources and extracting key product information, it is possible to construct a product information set that reflects the actual operating status of the enterprise. Through the trained word vector model, the product information and the registered industry classification of the enterprise are converted into word vectors, and the weighted similarity calculation is introduced. It is possible to comprehensively consider a variety of product information and accurately measure the degree of fit between the first word vector set and the second word vector, thereby quantitatively evaluating the accuracy of each enterprise's industry classification, reducing dependence on manual judgment, and significantly improving the accuracy and efficiency of financial institutions in identifying enterprise industry classification information. It achieves the effect of automatically identifying and marking enterprises with potentially incorrect industry classifications, greatly optimizing the credit decision-making process of financial institutions and enhancing the risk control capabilities of financial institutions.

[0104] Example 2

[0105] The embodiment of the present application also provides a device for identifying the industry classification information of an enterprise. It should be noted that the device for identifying the industry classification information of an enterprise in the embodiment of the present application can be used to execute the method for identifying the industry classification information of an enterprise provided in the embodiment of the present application. The device for identifying the industry classification information of an enterprise provided in the embodiment of the present application is introduced below.

[0106] According to an embodiment of the present application, a device for implementing the above-mentioned method for identifying the industry classification information of an enterprise is also provided. Figure 7 As shown, the device comprises:

[0107] Specifically, the extraction unit 701 is used to obtain the invoice data of each of the N enterprises from an external data source, and extract the commodity information contained in the invoice data to obtain a commodity information set of each enterprise, where N is a positive integer.

[0108] The conversion unit 702 is used to convert the commodity information set of each enterprise into a first word vector set through the trained word vector model, and convert the industry classification information of each enterprise into a second word vector through the word vector model.

[0109] The calculation unit 703 is used to calculate the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise.

[0110] The identification unit 704 is used to identify whether the industry classification information of N enterprises is correct based on the similarity between each enterprise and the industry classification information of each enterprise, and obtain the identification results of the N enterprises.

[0111] The device for identifying the industry classification information of an enterprise provided in an embodiment of the present application obtains the invoice data of each enterprise in N enterprises from an external data source through an extraction unit 701, and extracts the commodity information contained in the invoice data to obtain a commodity information set of each enterprise, wherein N is a positive integer; the conversion unit 702 converts the commodity information set of each enterprise into a first word vector set through a trained word vector model, and converts the industry classification information of each enterprise into a second word vector through the word vector model; the calculation unit 703 calculates the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between the industry classification information of each enterprise and each enterprise; the identification unit 704 identifies whether the industry classification information of N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtains the identification results of N enterprises, which solves the problem in the related art that the industry classification information registered by the enterprise itself or the industry classification information determined based on personal experience in the corporate credit business of financial institutions is incorrect.

[0112] By obtaining the invoice data of each enterprise from external data sources and extracting key product information, it is possible to construct a product information set that reflects the actual operating status of the enterprise. Through the trained word vector model, the product information and the registered industry classification of the enterprise are converted into word vectors, and the weighted similarity calculation is introduced. It is possible to comprehensively consider a variety of product information and accurately measure the degree of fit between the first word vector set and the second word vector, thereby quantitatively evaluating the accuracy of each enterprise's industry classification, reducing dependence on manual judgment, and significantly improving the accuracy and efficiency of financial institutions in identifying enterprise industry classification information. It achieves the effect of automatically identifying and marking enterprises with potentially incorrect industry classifications, greatly optimizing the credit decision-making process of financial institutions and enhancing the risk control capabilities of financial institutions.

[0113] Optionally, in the device for identifying industry classification information of an enterprise provided in Example 2 of the present application, the above-mentioned calculation unit 703 includes: a first calculation subunit, used to calculate the distance between each first word vector and the second word vector in the first word vector set in the word vector space, and obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed based on the word vector model; a first determination subunit, used to determine the weight corresponding to each first word vector based on the number of transactions of the commodity corresponding to each first word vector in the first word vector set; a second calculation subunit, used to calculate the similarity corresponding to the second word vector based on the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; a second determination subunit, used to determine the similarity between the industry classification information of each enterprise and each enterprise based on the similarity between the second word vector corresponding to the industry classification information of each enterprise and the second word vector.

[0114] Optionally, in the device for identifying the industry classification information of an enterprise provided in Example 2 of the present application, the above-mentioned identification unit 704 includes: a third determination subunit, used to determine a target similarity that is less than a preset value in the similarity between each enterprise and the industry classification information of each enterprise; a fourth determination subunit, used to determine the target enterprise corresponding to the target similarity, and determine that the industry classification information of the target enterprise is incorrect; a first generation subunit, used to generate an error classification report based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise, to obtain identification results of N enterprises.

[0115] Optionally, in the device for identifying the industry classification information of an enterprise provided in Example 2 of the present application, the device further includes: a first determination unit, used to identify whether the industry classification information of N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and after obtaining the identification results of the N enterprises, determine the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and the M industry classification information, and update the industry classification information of the target enterprise based on the target industry classification information, wherein M is a positive integer; a second determination unit, used to determine the historical credit behavior of the target enterprise, and evaluate the credit risk of the target enterprise based on the historical credit behavior; an adjustment unit, used to adjust the credit strategy of the target enterprise based on the credit risk of the target enterprise.

[0116] Optionally, in the device for identifying the industry classification information of an enterprise provided in Example 2 of the present application, the above-mentioned extraction unit 701 includes: a verification subunit, which is used to receive the enterprise's login request through the front-end web page, and verify the enterprise's login request to obtain a verification result; a receiving subunit, which is used to receive the enterprise's authorization information when the verification result indicates that the enterprise has passed the verification, wherein the authorization information is credential information for obtaining the enterprise's invoice data from an external data source; a second generation subunit, which is used to generate a data acquisition request based on the HTTP protocol and the authorization information, and send the data acquisition request to the external data source; a fifth determination subunit, which is used to receive return data from the external data source in response to the data acquisition request, and determine the enterprise's invoice data within a preset time period based on the return data.

[0117] Optionally, in the device for identifying the industry classification information of an enterprise provided in Example 2 of the present application, the above-mentioned extraction unit 701 includes: a screening subunit, used to screen the invoice data to obtain sales invoice data; a sixth determination subunit, used to determine the product names contained in the sales invoice data to obtain a product name set; a sorting subunit, used to sort each product name in the product name set according to product attributes to obtain each sorted product name, and determine the product information set of each enterprise based on each sorted product name, wherein the product attributes include at least one of the following: the transaction frequency of the product corresponding to the product name, the transaction amount of the product corresponding to the product name, and the product name.

[0118] Optionally, in the device for identifying the industry classification information of an enterprise provided in Example 2 of the present application, the above-mentioned device also includes: an acquisition unit, which is used to obtain Chinese corpus through a preset channel to obtain a corpus set before converting the product information set of each enterprise into a first word vector set through a trained word vector model and converting the industry classification information of each enterprise into a second word vector through the word vector model; a processing unit, which is used to perform preprocessing operations on the text information in the corpus set to obtain a word set, wherein the preprocessing operations include at least: font conversion and word segmentation processing; an input unit, which is used to input the word set into an untrained word vector model; a construction unit, which is used to analyze the semantic similarity between words in the word set through an untrained word vector model, construct a model parameter set, and construct a trained word vector model based on the model parameter set.

[0119] It should be noted that the above extraction unit 701, conversion unit 702, calculation unit 703 and identification unit 704 correspond to steps S201 to S204 in Example 1, and the two modules and the corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102 enterprise industry classification information identification, ..., 102n), and the above modules can also be run in the computer terminal 10 provided in Example 1 as part of the device.

[0120] Example 3

[0121] An embodiment of the present application may provide an electronic device, Figure 8 is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 8 As shown, the electronic device may include: one or more ( Figure 8 (only one is shown) processor 802, memory 804, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0122] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application, and the processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, realizing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0123] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: for each of the N enterprises, obtain the enterprise's invoice data from an external data source, and extract the commodity information contained in the invoice data to obtain a commodity information set for each enterprise, where N is a positive integer; convert the commodity information set of each enterprise into a first word vector set through a trained word vector model, and convert the industry classification information of each enterprise into a second word vector through the word vector model; calculate the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between the industry classification information of each enterprise and each enterprise; identify whether the industry classification information of the N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtain the identification results of the N enterprises.

[0124] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: calculate the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise, including: calculating the distance between each first word vector and the second word vector in the first word vector set in the word vector space to obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed based on the word vector model; determining the weight corresponding to each first word vector based on the number of transactions of the commodity corresponding to each first word vector in the first word vector set; calculating the similarity corresponding to the second word vector based on the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; determining the similarity between the industry classification information of each enterprise and each enterprise based on the similarity between the second word vector corresponding to the industry classification information of each enterprise and the second word vector.

[0125] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: identify whether the industry classification information of N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtain the identification results of N enterprises, including: determining a target similarity that is less than a preset value in the similarity between the industry classification information of each enterprise and each enterprise; determining the target enterprise corresponding to the target similarity, and determining that the industry classification information of the target enterprise is incorrect; generating an error classification report based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise, and obtaining the identification results of N enterprises.

[0126] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: after identifying whether the industry classification information of N enterprises is correct based on the similarity between the industry classification information of each enterprise and each enterprise, and obtaining the identification results of N enterprises, the above method also includes: determining the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and M industry classification information, and updating the industry classification information of the target enterprise based on the target industry classification information, where M is a positive integer; determining the historical credit behavior of the target enterprise, and evaluating the credit risk of the target enterprise based on the historical credit behavior; adjusting the credit strategy of the target enterprise based on the credit risk of the target enterprise.

[0127] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: obtain the invoice data of the enterprise from the external data source, including: receiving the enterprise's login request through the front-end web page, and verifying the enterprise's login request to obtain a verification result; when the verification result indicates that the enterprise has passed the verification, receiving the enterprise's authorization information, wherein the authorization information is the credential information for obtaining the enterprise's invoice data from the external data source; generating a data acquisition request based on the HTTP protocol and the authorization information, and sending the data acquisition request to the external data source; receiving the return data of the external data source in response to the data acquisition request, and determining the enterprise's invoice data within a preset time period based on the return data.

[0128] The processor can call the information and application stored in the memory through the transmission device to perform the following steps: extract the product information contained in the invoice data to obtain a product information set for each enterprise, including: screening the invoice data to obtain the sales invoice data; determining the product name contained in the sales invoice data to obtain a product name set; sorting each product name in the product name set according to the product attributes to obtain each sorted product name, and determining the product information set for each enterprise based on each sorted product name, wherein the product attributes include at least one of the following: the transaction frequency of the product corresponding to the product name, the transaction amount of the product corresponding to the product name, and the product name.

[0129] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: before converting the product information set of each enterprise into a first word vector set through the trained word vector model, and converting the industry classification information of each enterprise into a second word vector through the word vector model, the above method also includes: obtaining Chinese corpus through a preset channel to obtain a corpus set; performing preprocessing operations on the text information in the corpus set to obtain a word set, wherein the preprocessing operations at least include: font conversion and word segmentation processing; inputting the word set into an untrained word vector model; analyzing the semantic similarity between words in the word set through the untrained word vector model, constructing a model parameter set, and constructing a trained word vector model based on the model parameter set.

[0130] By adopting the embodiment of the present application, a method for identifying the industry classification information of an enterprise is provided. For each of N enterprises, the invoice data of the enterprise is obtained from an external data source, and the commodity information contained in the invoice data is extracted to obtain a commodity information set of each enterprise, wherein N is a positive integer; the commodity information set of each enterprise is converted into a first word vector set through a trained word vector model, and the industry classification information of each enterprise is converted into a second word vector through a word vector model; the second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between the industry classification information of each enterprise and each enterprise; based on the similarity between the industry classification information of each enterprise and each enterprise, whether the industry classification information of N enterprises is correct is identified to obtain the identification results of N enterprises, thereby solving the technical problem that the industry classification information registered by the enterprise itself or the industry classification information determined based on personal experience in the public credit business of financial institutions is wrong.

[0131] By obtaining the invoice data of each enterprise from external data sources and extracting key product information, it is possible to construct a product information set that reflects the actual operating status of the enterprise. Through the trained word vector model, the product information and the registered industry classification of the enterprise are converted into word vectors, and the weighted similarity calculation is introduced. It is possible to comprehensively consider a variety of product information and accurately measure the degree of fit between the first word vector set and the second word vector, thereby quantitatively evaluating the accuracy of each enterprise's industry classification, reducing dependence on manual judgment, and significantly improving the accuracy and efficiency of financial institutions in identifying enterprise industry classification information. It achieves the effect of automatically identifying and marking enterprises with potentially incorrect industry classifications, greatly optimizing the credit decision-making process of financial institutions and enhancing the risk control capabilities of financial institutions.

[0132] It can be understood by those skilled in the art that Figure 8The structure shown is for illustration only, and the electronic device may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, or other terminal devices. Figure 8 The structure of the electronic device is not limited. Figure 8 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 8 Different configurations are shown.

[0133] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0134] Example 4

[0135] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the method for identifying the industry classification information of an enterprise provided in the first embodiment.

[0136] Optionally, in this embodiment, the above storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0137] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of the method for identifying the industry classification information of an enterprise.

[0138] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0139] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0144] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for identifying industry classification information of an enterprise, characterized in that: include: For each of the N enterprises, invoice data of the enterprise is obtained from an external data source, and commodity information contained in the invoice data is extracted to obtain a commodity information set of each enterprise, where N is a positive integer; The product information set of each enterprise is converted into a first word vector set through the trained word vector model, and the industry classification information of each enterprise is converted into a second word vector through the word vector model; Calculate the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise; Whether the industry classification information of the N enterprises is correct is identified based on the similarity between each enterprise and the industry classification information of each enterprise, and the identification results of the N enterprises are obtained.

2. The method according to claim 1, characterized in that: The second word vector and each first word vector in the first word vector set are calculated by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise, including: Calculating the distance between each first word vector in the first word vector set and the second word vector in the word vector space to obtain the similarity between the second word vector and each first word vector in the first word vector set, wherein the word vector space is constructed according to the word vector model; Determining a weight corresponding to each first word vector according to the number of transactions of the commodity corresponding to each first word vector in the first word vector set; Calculate the similarity corresponding to the second word vector according to the similarity between the second word vector and each first word vector in the first word vector set and the weight corresponding to each first word vector; The similarity between each enterprise and the industry classification information of each enterprise is determined based on the second word vector corresponding to the industry classification information of each enterprise and the similarity corresponding to the second word vector.

3. The method according to claim 1, characterized in that Identifying whether the industry classification information of the N enterprises is correct based on the similarity between each enterprise and the industry classification information of each enterprise, and obtaining the identification results of the N enterprises, including: Determine a target similarity less than a preset value among the similarities between each enterprise and the industry classification information of each enterprise; Determining a target enterprise corresponding to the target similarity, and determining that the industry classification information of the target enterprise is incorrect; An error classification report is generated based on the industry classification information of the target enterprise and the target similarity corresponding to the target enterprise to obtain the identification results of the N enterprises.

4. The method according to claim 3, characterized in that After identifying whether the industry classification information of the N enterprises is correct based on the similarity between each enterprise and the industry classification information of each enterprise, and obtaining the identification results of the N enterprises, the method further includes: Determine the target industry classification information of the target enterprise by calculating the similarity between the target enterprise and M industry classification information, and update the industry classification information of the target enterprise according to the target industry classification information, wherein M is a positive integer; Determine the historical credit behavior of the target enterprise, and assess the credit risk of the target enterprise based on the historical credit behavior; The credit strategy of the target enterprise is adjusted according to the credit risk of the target enterprise.

5. The method according to claim 1, characterized in that Obtaining the invoice data of the enterprise from an external data source, including: Receiving a login request from the enterprise through a front-end webpage, and verifying the login request from the enterprise to obtain a verification result; When the verification result indicates that the enterprise has passed the verification, receiving authorization information of the enterprise, wherein the authorization information is credential information for obtaining the invoice data of the enterprise from the external data source; Generate a data acquisition request according to the HTTP protocol and the authorization information, and send the data acquisition request to the external data source; Receive the return data of the external data source in response to the data acquisition request, and determine the invoice data of the enterprise within a preset time period based on the return data.

6. The method according to claim 1, characterized in that The commodity information contained in the invoice data is extracted to obtain a commodity information set of each enterprise, including: Screening the invoice data to obtain sales invoice data; Determine the commodity names included in the sales invoice data, and obtain a commodity name set; Each product name in the product name set is sorted according to product attributes to obtain each sorted product name, and a product information set of each enterprise is determined according to each sorted product name, wherein the product attributes include at least one of the following: transaction frequency of the product corresponding to the product name, transaction amount of the product corresponding to the product name, and the product name.

7. The method according to claim 1, characterized in that Before converting the commodity information set of each enterprise into a first word vector set by using the trained word vector model, and converting the industry classification information of each enterprise into a second word vector by using the word vector model, the method further includes: Obtain Chinese corpus through preset channels to obtain a corpus set; Performing a preprocessing operation on the text information in the corpus to obtain a word set, wherein the preprocessing operation at least includes: font conversion and word segmentation processing; Inputting the word set into an untrained word vector model; The semantic similarity between words in the word set is analyzed by the untrained word vector model, a model parameter set is constructed, and the trained word vector model is constructed based on the model parameter set.

8. A device for identifying industry classification information of an enterprise, characterized in that: include: an extraction unit, for acquiring, for each of the N enterprises, invoice data of the enterprise from an external data source, and extracting commodity information contained in the invoice data to obtain a commodity information set of each enterprise, wherein N is a positive integer; A conversion unit, used to convert the commodity information set of each enterprise into a first word vector set through a trained word vector model, and convert the industry classification information of each enterprise into a second word vector through the word vector model; A calculation unit, configured to calculate the second word vector and each first word vector in the first word vector set by weighted similarity calculation to obtain the similarity between each enterprise and the industry classification information of each enterprise; The identification unit is used to identify whether the industry classification information of the N enterprises is correct based on the similarity between each enterprise and the industry classification information of each enterprise, and obtain the identification results of the N enterprises.

9. An electronic device, characterized in that: include: A memory storing an executable program; A processor is used to run the program, wherein the program, when running, executes the method for identifying the industry classification information of an enterprise as described in any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the steps of the method for identifying the industry classification information of an enterprise as described in any one of claims 1 to 7 are implemented.

Images