A method, device, equipment and medium for generating intelligent financial statements

Abstract

The present invention relates to the field of computer technology, and proposes an intelligent financial report generation method, device, equipment and medium. The method includes: obtaining financial data stored in an ERP system; training a report generation model to classify financial data based on corresponding classification rules to obtain corresponding classification results; generating corresponding financial reports based on the corresponding classification results, and calculating the matching degree between the generated financial reports and the target business scenario based on matching degree calculation rules; updating the report generation model based on the calculated matching degree and returning to the step of classifying the financial data until the preset conditions are met, and generating financial reports matching the business scenario based on the trained report generation model. The solution of the present invention adaptively generates financial reports matching different business scenarios, meets the needs of different business scenarios, improves user experience, and improves the adaptability and flexibility of the ERP system in generating financial reports.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an intelligent financial report generation method, device, equipment and medium. Background Art

[0002] The ERP (Enterprise Resource Planning) system is built on the basis of information technology. It uses a systematic management concept to provide a management platform for enterprise decision-makers and employees to make decisions. The financial accounting module in the ERP system is the core of enterprise financial management. Traditional ERP systems rely on a lot of manual operations in accounting processing and report generation, which is prone to human errors and low efficiency. With the development of artificial intelligence and machine learning technology, the existing ERP system has been combined with intelligent technology, which has significantly improved the automation level and data processing efficiency of the ERP system.

[0003] However, the existing ERP system integrates a fixed financial report generation model, which is difficult to match with different business scenarios and meet the needs of different business scenarios. It is easy to lead to a poor user experience and reduce the adaptability and flexibility of the ERP system in generating financial reports.

[0004] Therefore, there is an urgent need for an intelligent financial report generation method in an ERP system that can be applied to different business scenarios. Summary of the invention

[0005] In view of this, in the prior art, the ERP system integrates a fixed financial report generation mode, which is difficult to match with different business scenarios and meet the needs of different business scenarios, which easily leads to a poor user experience and reduces the adaptability and flexibility of the ERP system in generating financial reports. The present disclosure provides an intelligent financial report generation method, device, equipment and medium.

[0006] According to a first aspect of the present invention, there is provided a method for generating an intelligent financial report, comprising:

[0007] Access financial data stored in the ERP system;

[0008] The training report generation model classifies the financial data based on corresponding classification rules to obtain corresponding classification results;

[0009] Generate a corresponding financial statement based on the corresponding classification result, and calculate the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule;

[0010] The report generation model is updated based on the calculated matching degree and the step of classifying the financial data is returned until a preset condition is met, and a financial report matching the business scenario is generated based on the trained report generation model.

[0011] In some embodiments, the step of calculating the matching degree between the generated financial statements and the target business scenario based on the matching degree calculation rule includes:

[0012] Assign corresponding weight values ​​to multiple dimensional indicators for evaluating the matching degree between financial statements and business scenarios;

[0013] The statistically generated financial statements and the target business scenarios correspond to the corresponding evaluation values ​​under each dimension indicator, which are multiplied by the corresponding assigned weight values ​​and then added together to obtain the corresponding matching degree.

[0014] In some embodiments, the step of updating the report generation model based on the calculated matching degree and returning to classify the financial data until a preset condition is met, and the step of generating a financial report matching the business scenario based on the trained report generation model includes:

[0015] A loss function is established based on the calculated matching degree and matching degree target value;

[0016] The report generation model is updated based on the loss function and the financial data is classified, so as to perform several rounds of training on the report generation model, and the report generation model is updated based on the loss function corresponding to each round of training until the loss function is minimized;

[0017] The report generation model obtained based on the final round of training outputs financial reports that best match the business scenario.

[0018] In some embodiments, the step of updating the report generation model based on the loss function corresponding to each round of training includes:

[0019] Based on the loss function obtained in each round of training, analyze the impact of each item in the corresponding financial statement generated in this round on the calculated matching degree, whether it is a positive or negative impact, and the corresponding impact weight;

[0020] The parameters of the report generation model are updated based on the analysis results and the corresponding classification rule selected from the preset multiple classification rules is adjusted to update the report generation model.

[0021] In some embodiments, the step of training the report generation model to classify the financial data based on the corresponding classification rules includes:

[0022] The training report generation model selects a corresponding classification rule from a plurality of preset classification rules and classifies the financial data based on the corresponding classification rule, wherein the plurality of preset classification rules include a plurality of classification rules formed based on one or more combinations of transaction type, transaction amount, and accounting subject.

[0023] In some embodiments, the method further comprises:

[0024] Scan the financial voucher to be input into the ERP system into an image file, and extract key field information from the image file based on the OCR model;

[0025] Convert the key field information into corresponding structured data based on a preset regular expression;

[0026] The structured data is imported into the ERP system based on the RPA tool to generate the financial data based on it.

[0027] In some embodiments, the method further comprises:

[0028] Performing data cleaning on the structured data, and detecting whether there is abnormal transaction data exceeding a preset threshold or occurring at an abnormal time in the structured data based on a pre-trained anomaly detection algorithm;

[0029] In response to the absence of the abnormal transaction data in the structured data, the structured data is stored in the form of accounting entries as the financial data.

[0030] According to a second aspect of the present invention, there is provided an intelligent financial statement generating device, the device comprising:

[0031] The first module is used to obtain financial data stored in the ERP system;

[0032] The second module is used to train the report generation model to classify the financial data based on the corresponding classification rules to obtain the corresponding classification results;

[0033] A third module is used to generate a corresponding financial statement based on the corresponding classification result, and calculate the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule;

[0034] The fourth module is used to update the report generation model based on the calculated matching degree and return to the second module until the preset conditions are met, and generate a financial report matching the business scenario based on the trained report generation model.

[0035] According to a third aspect of the present invention, there is further provided an electronic device, the electronic device comprising:

[0036] At least one processor; and a memory, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the program, the aforementioned intelligent financial report generation method is executed.

[0037] According to a fourth aspect of the present invention, there is further provided a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the computer program executes any one of the above-mentioned intelligent financial report generation methods.

[0038] The above-mentioned intelligent financial report generation method obtains financial data from the ERP system, trains a report generation model, classifies the financial data based on the corresponding classification rules, obtains the corresponding classification results, and then generates the corresponding financial report based on the classification results, calculates the matching degree between the generated financial report and the target business scenario based on the matching degree calculation rules, and updates the report generation model based on the matching degree calculation rules to generate financial reports that match the business scenario. The report generation model trained in the ERP system is used to adaptively generate financial reports that match different business scenarios, meet the needs of different business scenarios, improve user experience, and improve the adaptability and flexibility of the ERP system in generating financial reports.

[0039] In addition, the present invention also provides an intelligent financial statement generation device, an electronic device and a computer-readable storage medium, which can also achieve the above-mentioned technical effects, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other embodiments can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of a method for generating an intelligent financial report provided by one embodiment of the present invention;

[0042] Figure 2 A schematic diagram of the structure of an intelligent financial report generation device provided by another embodiment of the present invention;

[0043] Figure 3 is an internal structure diagram of an electronic device in another embodiment of the present invention;

[0044] Figure 4 It is a schematic diagram of the structure of a computer-readable storage medium in another embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the embodiments of the present invention are further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0046] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are for distinguishing two non-identical entities with the same name or non-identical parameters. It can be seen that "first" and "second" are only for the convenience of expression and should not be understood as limitations on the embodiments of the present invention. The subsequent embodiments will not explain this one by one.

[0047] In one embodiment, please refer to Figure 1 As shown, the present invention provides an intelligent financial statement generation method 100, specifically, comprising the following steps:

[0048] Step 101, obtaining financial data stored in the ERP system;

[0049] Step 102, the training report generation model classifies the financial data based on the corresponding classification rules to obtain the corresponding classification results;

[0050] Step 103, generating a corresponding financial statement based on the corresponding classification result, and calculating the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule;

[0051] Step 104, based on the calculated matching degree, the report generation model is updated and the step of classifying the financial data is returned until the preset conditions are met, and a financial report matching the business scenario is generated based on the trained report generation model.

[0052] Through the report generation model trained in the ERP system, financial reports matching different business scenarios can be adaptively generated to meet the needs of different business scenarios, improve user experience, and improve the adaptability and flexibility of the ERP system in generating financial reports.

[0053] According to some embodiments of the present invention, the method further includes: scanning the financial voucher to be input into the ERP system as an image file, extracting key field information in the image file based on an OCR (Optical Character Recognition) model, and converting the key field information into corresponding structured data based on a preset regular expression, and then importing the structured data into the ERP system based on an RPA (Robotic process automation) tool to generate the financial data based thereon. By combining the OCR model with the RPA tool, the manual data entry and verification by financial personnel in the accounting process is reduced, and human errors are reduced and data accuracy is improved through automation and intelligent functions.

[0054] According to some embodiments of the present invention, the method further includes: performing data cleaning on the structured data, detecting whether there is abnormal transaction data exceeding a preset threshold or at an abnormal time in the structured data based on a pre-trained anomaly detection algorithm; in response to the absence of abnormal transaction data in the structured data, storing the structured data in the form of accounting entries as the financial data. The accounting data is monitored in real time through a machine learning model, abnormal transactions are automatically detected, and financial personnel are reminded to review to ensure the accuracy of the financial data in the ERP system.

[0055] According to several embodiments of the present invention, the step of calculating the matching degree between the generated financial statements and the target business scenarios based on the matching degree calculation rules includes: assigning corresponding weight values ​​to the multiple dimensional indicators for evaluating the matching degree between the financial statements and the business scenarios; calculating the corresponding evaluation values ​​of the generated financial statements and the target business scenarios corresponding to each dimensional indicator, multiplying them with the corresponding assigned weight values ​​and then adding them up to obtain the corresponding matching degree. The method of establishing a loss function based on the calculated matching degree and the matching degree target value, and performing multiple rounds of training on the report generation model, improves the accuracy of the financial statements and the relevance to the specific business scenarios, enhances the user experience, and provides strong support for enterprise decision-making.

[0056] According to several embodiments of the present invention, the report generation model is updated based on the calculated matching degree and the step of returning to classify the financial data until the preset conditions are met, and the step of generating financial statements matching the business scenario based on the trained report generation model includes: establishing a loss function based on the calculated matching degree and the matching degree target value; updating the report generation model based on the loss function and returning to classify the financial data, so as to train the report generation model for several rounds, and updating the report generation model based on the loss function corresponding to each round of training until the loss function is minimized, and the report generation model obtained from the final round of training outputs the financial statements with the highest matching degree with the business scenario. By minimizing the loss function, financial statements matching specific business scenarios can be obtained to the greatest extent, thereby improving the overall adaptability and flexibility of the ERP system.

[0057] According to several embodiments of the present invention, the step of updating the report generation model based on the loss function corresponding to each round of training includes: based on the loss function corresponding to each round of training, analyzing whether the impact of each item in the corresponding financial report generated in this round on the calculated matching degree is positive or negative, and the corresponding impact weight; based on the analysis results, updating the parameters of the report generation model and adjusting the corresponding classification rules selected from the preset multiple classification rules to update the report generation model. By analyzing the loss function and its relationship with each item in the financial report, it is possible to deeply understand which items have a positive or negative impact on the matching degree (i.e., the degree of conformity between the model prediction and the actual needs of the business scenario), as well as the specific weights of these impacts, so as to improve the accuracy of the financial reports automatically generated by ERP and enhance the overall adaptability of the ERP system.

[0058] According to several embodiments of the present invention, the step of training a report generation model to classify financial data based on corresponding classification rules includes: the training report generation model selects corresponding classification rules from a plurality of preset classification rules and classifies the financial data based thereon, wherein the plurality of preset classification rules include a plurality of classification rules formed based on one or more combinations of transaction type, transaction amount, and accounting subject. The report generation model is trained to select a classification rule from a plurality of different classification rules, and the financial data is classified based on the selected classification rule, and after multiple rounds of training, the classification rule that is most suitable for the target business scenario is found to meet the needs of different business scenarios and improve the user experience.

[0059] In order to further understand the intelligent financial statement generation method provided by the present invention, the technical solution of the present invention is further elaborated in detail in a specific example below.

[0060] First, the financial data to be imported into the ERP system is preprocessed and detected for anomalies. After no problems are found in the anomaly detection, the financial data to be imported is stored in the ERP system in the form of accounting entries. Specifically, high-precision scanning equipment is used to scan financial documents such as bills and invoices into image files, and the images are preprocessed, such as denoising, edge detection, image correction and other preprocessing methods, to ensure that the subsequent OCR model can accurately recognize the content in the image. Then, the preprocessed image file is input into the OCR model, and the key field information in the image file, such as invoice number, amount, date, tax number, etc., is extracted through the OCR model. Then, the extracted data is preliminarily structured using predefined templates and regular expressions to obtain structured data that can be used by the RPA tool. The data recognized by the OCR is automatically filled into the ERP system through the RPA tool. The RPA tool can realize the automatic import of data into the ERP system and synchronize data in multiple ERP systems through automated scripts, reducing manual operations and improving the accuracy and efficiency of data entry into the ERP system. For example, use Tesseract or Google Cloud Vision API (Application Programming Interface) to automatically identify and enter financial documents (such as bills, invoices, etc.). Tesseract is an image recognition based on convolutional neural networks, and its formula is as follows:

[0061] (Formula 1)

[0062] Among them, σ is the activation function, w k is the convolution kernel, x k is the pixel of the input image, b is the bias, y is the image recognition result, σ is used to determine the response intensity of each pixel to help identify the character morphology in the image, and w k and x k The convolution operation is used to scan the image and extract feature information such as the edges and curves of the characters, thereby helping the OCR model to accurately identify the content of the bill.

[0063] The ERP system performs anomaly detection on the imported structured data, and performs a preliminary automatic check on the entered data with the existing financial records in the ERP system. If inconsistent data is found, the system will mark these abnormal items and generate reports to remind financial personnel to conduct manual review. The system also performs anomaly detection on the entered data based on the anomaly detection model trained in the ERP system. The machine learning model monitors the accounting data in real time, automatically detects abnormal transactions and reminds financial personnel to review. According to the set threshold or anomaly detection rules, it automatically identifies abnormal transaction data in the input data, such as transaction amounts exceeding the set threshold, transaction amounts during non-working hours, etc. If abnormal transaction data is detected, an abnormal reminder will be generated, such as through email, text messages or system notifications, to remind financial personnel to review and process in real time to ensure data accuracy. For example, the LSTM (Long Short-Term Memory) network is used to monitor the accounting data in real time, automatically detect abnormal transactions and remind financial personnel to review. The formula of the LSTM network is as follows:

[0064]

[0065] (Formula 2)

[0066] in, are the activation values ​​of the forget gate, input gate, and output gate, respectively. is the candidate cell state, C t is the cell state, h t is a hidden state. The activation values ​​of the forget gate, input gate, and output gate are used to control the memory or forgetfulness of each transaction in the account data, helping the model to predict whether future transactions are abnormal. t and h t Used to capture transaction patterns in time series of financial data and mark abnormal transactions.

[0067] During the training of the anomaly detection model, historical transaction data is extracted from the ERP system, including transaction data of different dimensions such as amount, time, and transaction category. The data is cleaned to remove invalid records and fill in missing values ​​to ensure the accuracy of model training. At the same time, a suitable machine learning algorithm is selected to train anomaly detection models for financial transaction anomalies. Common algorithms include Isolation Forest and density-based anomaly detection algorithms, and the features used in model training include transaction amount, transaction frequency, user behavior patterns, etc.

[0068] Furthermore, in the process of training the report generation model, historical accounting entries are extracted from the financial data stored in the ERP system, and corresponding classification rules are selected from the preset multiple classification rules to classify the extracted historical accounting entries, such as classification according to one or more combinations of transaction type, amount, and subject. According to the financial system and business process of the enterprise, the training model learns the rules of automatic classification, such as automatically matching the corresponding subject and amount according to the transaction type. These rules can be based on the business logic preset by the ERP system, such as procurement, sales, payment, etc. They can also be automatically summarized from the data through machine learning methods. The matching degree of the generated financial statements and the target business scenario is calculated based on the matching degree calculation rules, and the report generation model is continuously trained based on the matching degree to generate financial statements suitable for different business scenarios. The architecture of the report generation model is a large natural language processing model based on neural networks. It understands and generates natural language text by training massive text data. The core technology is deep neural network, especially Transformer architecture, which processes text data through self-attention mechanism to capture long-distance dependencies in text. The self-attention mechanism includes pre-training steps and fine-tuning steps. In the pre-training stage, the model is trained on large-scale unsupervised text data to learn the basic structure and patterns of the language; in the fine-tuning stage, the model is further trained on data of specific tasks to adapt to specific business scenarios. Common large language models include the GPT (Generative Pre-trained Transformer) series, BERT (Bidirectional Encoder Representations from Transformers), and Turing-NLG. Based on the trained report generation model, when new transaction data enters the ERP system, the system automatically generates accounting entries based on historical entry data, business rules, and NLP (Neuro Linguistic Programming) analysis results. In order to further ensure data security, the automatically generated financial statements will be verified by the system's preset verification rules to ensure that they meet financial requirements. For example, the automatic generation of accounting entries through NLP technology is achieved, and the formula is as follows,

[0069] (Formula 3)

[0070] Among them, h T is the final hidden state, W y and b y are the weights and biases of the output layer, h T and W yIn the natural language model, it is used to extract key information from the input financial data and map this information to accounting subjects and amounts. By preprocessing historical accounting data, building training sets, training large language models, and learning accounting processing rules, the model automatically generates financial statements based on the input financial data when applied in real time.

[0071] The financial statements are automatically generated based on the trained report generation model, and stream processing technologies such as Kafka are used to achieve real-time synchronization of financial data and ensure the timeliness and consistency of report data. The working principle of Kafka is based on the publish-subscribe model, and its formula is as follows:

[0072] (Formula 4)

[0073] Where P(x) is the probability of receiving a message, π i The weights of different topics, is the conditional probability of a message under a specific topic.

[0074] Furthermore, the ERP system provides a variety of intelligent report templates based on the trained report generation model for adaptive matching in different business scenarios. Users can customize the report format according to their needs. The system automatically integrates data and generates financial reports. Then, through enhanced retrieval technology, such as Elasticsearch technology, the required data can be quickly located and extracted. Elasticsearch uses inverted indexes for fast data retrieval. The formula is as follows:

[0075] (Formula 5)

[0076] Among them, S(q, d) is the matching score between query q and document d, N is the total number of documents, df t is the number of documents containing term t, is the frequency of term t in document d, k1 and b are adjustment parameters, |d| is the document length, and avgdl is the average document length.

[0077] Furthermore, it also provides a combination of deep learning models, such as a deep learning model based on the Transformer model, to conduct multi-dimensional analysis of future financial data, provide trend forecasting and decision support. The calculation formula of the Transformer model is as follows:

[0078] (Formula 6)

[0079] Attention is the self-attention mechanism of the Transformer model. Q, K, and V represent query, key, and value matrices, respectively. k is the dimension of the key vector.

[0080] Based on the trained report generation model, the ERP system receives the user's query request. For example, the user enters the query statement "generate a certain financial report". The ERP system retrieves the document library based on the DPR model and vectorizes the retrieved related documents. It calculates the similarity between the retrieved related documents and all documents in the document library, selects the k documents with the highest similarity, and generates text y based on the retrieval results by using the BART (Bidirectional and Auto-Regressive Transformers) model. Specifically, the retrieved related documents are concatenated into an input sequence, which is input into the encoder of the generation model to generate hidden representation, and then the decoder is used to generate natural language text. The generated text is output as the final report content for users to view and use.

[0081] According to some embodiments of the present invention, please refer to Figure 2 As shown, the present invention also provides an intelligent financial statement generation device 200, the device comprising:

[0082] The first module 201 is used to obtain financial data stored in the ERP system;

[0083] The second module 202 is used to train the report generation model to classify the financial data based on the corresponding classification rules to obtain the corresponding classification results;

[0084] The third module 203 is used to generate a corresponding financial statement based on the corresponding classification result, and calculate the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule;

[0085] The fourth module 204 is used to update the report generation model based on the calculated matching degree and return to the second module until the preset conditions are met, and generate a financial report matching the business scenario based on the trained report generation model.

[0086] It should be noted that each module in the above-mentioned intelligent financial statement generation device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in an electronic device in the form of hardware, or can be stored in a memory in an electronic device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0087] According to another aspect of the present invention, an electronic device is provided. The electronic device may be a server. Figure 3As shown. The electronic device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the electronic device is used to store data. The network interface of the electronic device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the above-mentioned intelligent financial report generation method is implemented.

[0088] According to another aspect of the present invention, a computer readable storage medium is provided. Figure 4 As shown, a computer program is stored thereon, and when the computer program is executed by a processor, the above-mentioned intelligent financial statement generation method is implemented.

[0089] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0090] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0091] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for generating intelligent financial statements, characterized in that: include: Access financial data stored in the ERP system; The training report generation model classifies the financial data based on corresponding classification rules to obtain corresponding classification results; Generate a corresponding financial statement based on the corresponding classification result, and calculate the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule; The report generation model is updated based on the calculated matching degree and the step of classifying the financial data is returned until a preset condition is met, and a financial report matching the business scenario is generated based on the trained report generation model; The step of updating the report generation model based on the calculated matching degree and returning to classify the financial data until the preset condition is met includes: A loss function is established based on the calculated matching degree and matching degree target value; The report generation model is updated based on the loss function and the financial data is classified, so as to perform several rounds of training on the report generation model, and the report generation model is updated based on the loss function corresponding to each round of training until the loss function is minimized; The step of updating the report generation model based on the loss function corresponding to each round of training includes: Based on the loss function obtained in each round of training, analyze the impact of each item in the corresponding financial statement generated in this round on the calculated matching degree, whether it is a positive or negative impact, and the corresponding impact weight; The parameters of the report generation model are updated based on the analysis results and the corresponding classification rule selected from the preset multiple classification rules is adjusted to update the report generation model.

2. The intelligent financial report generation method according to claim 1, characterized in that: The step of calculating the matching degree between the generated financial statements and the target business scenario based on the matching degree calculation rule includes: Assign corresponding weight values ​​to multiple dimensional indicators for evaluating the matching degree between financial statements and business scenarios; The statistically generated financial statements and the target business scenarios correspond to the corresponding evaluation values ​​under each dimension indicator, which are multiplied by the corresponding assigned weight values ​​and then added together to obtain the corresponding matching degree.

3. The intelligent financial report generation method according to claim 2, characterized in that: The step of generating a financial report matching the business scenario based on the trained report generation model includes: The report generation model obtained based on the final round of training outputs financial reports that best match the business scenario.

4. The intelligent financial report generation method according to claim 1, characterized in that: The step of training the report generation model to classify the financial data based on the corresponding classification rules includes: The training report generation model selects a corresponding classification rule from a plurality of preset classification rules and classifies the financial data based on the corresponding classification rule, wherein the plurality of preset classification rules include a plurality of classification rules formed based on one or more combinations of transaction type, transaction amount, and accounting subject.

5. The intelligent financial report generation method according to claim 1, characterized in that: The method further comprises: Scan the financial voucher to be input into the ERP system into an image file, and extract key field information from the image file based on the OCR model; Convert the key field information into corresponding structured data based on a preset regular expression; The structured data is imported into the ERP system based on the RPA tool to generate the financial data based on it.

6. The intelligent financial report generation method according to claim 5, characterized in that: The method further comprises: Performing data cleaning on the structured data, and detecting whether there is abnormal transaction data exceeding a preset threshold or occurring at an abnormal time in the structured data based on a pre-trained anomaly detection algorithm; In response to the absence of the abnormal transaction data in the structured data, the structured data is stored in the form of accounting entries as the financial data.

7. An intelligent financial report generation device, characterized in that: The device comprises: The first module is used to obtain financial data stored in the ERP system; The second module is used to train the report generation model to classify the financial data based on the corresponding classification rules to obtain the corresponding classification results; A third module is used to generate a corresponding financial statement based on the corresponding classification result, and calculate the matching degree between the generated financial statement and the target business scenario based on the matching degree calculation rule; A fourth module is used to update the report generation model based on the calculated matching degree and return to the second module until a preset condition is met, and generate a financial report matching the business scenario based on the trained report generation model; The fourth module is further used for: A loss function is established based on the calculated matching degree and matching degree target value; The report generation model is updated based on the loss function and the financial data is classified, so as to perform several rounds of training on the report generation model, and the report generation model is updated based on the loss function corresponding to each round of training until the loss function is minimized; The step of updating the report generation model based on the loss function corresponding to each round of training includes: Based on the loss function obtained in each round of training, analyze the impact of each item in the corresponding financial statement generated in this round on the calculated matching degree, whether it is a positive or negative impact, and the corresponding impact weight; The parameters of the report generation model are updated based on the analysis results and the corresponding classification rule selected from the preset multiple classification rules is adjusted to update the report generation model.

8. An electronic device, characterized in that: include: at least one processor; and a memory, wherein the memory stores a computer program that can be executed in the processor, and when the processor executes the program, the intelligent financial statement generation method described in any one of claims 1-6 is executed.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the intelligent financial statement generation method according to any one of claims 1 to 6 is performed.

Images