LLM and LSTM driven intelligent business financial classification and identification management method

Abstract

According to the LLM and LSTM-driven intelligent business and financial classification and identification management method provided by the invention, the LLM large model has strong language understanding and information extraction capabilities, sales orders in various formats can be accurately analyzed, and errors caused by manual analysis are avoided. The LSTM industry financial intelligent classification driving model is trained based on a large amount of historical data, customer attributes and prediction rules can be identified more accurately, and the accuracy of order settlement and delivery is improved. Interference of human factors is reduced in the automatic data processing and delivery process, and accuracy and consistency of data are guaranteed. The automatic business process reduces the dependence on manpower and reduces the labor cost. Meanwhile, due to the fact that errors and delay are reduced, extra cost expenditure is avoided, and economic benefits of enterprises are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent industry and finance tax processing, in particular to an LLM and LSTM driven industry and finance intelligent classification identification management method, an LLM and LSTM driven industry and finance intelligent classification identification management device, an electronic device and a computer readable storage medium. BACKGROUND

[0002] The traditional method usually relies on manual entry of sales order information into the business system. The staff needs to carefully check the contents on the order, such as customer name, order amount, product details, etc., and then fill in the fields one by one in the system. As for the preprocessing of the order, it may only be a simple format check, such as checking whether the date format is correct, whether the amount is a number, etc., and these checks are mostly done manually or rely on simple script programs.

[0003] Therefore, the traditional method lacks intelligent analysis means and mainly relies on manual identification and extraction of key information in the order. Employees need to find out the document number, customer information and order document information from a large amount of order text. This process not only consumes time, but also is prone to errors, especially for orders with non-standard formats. Even if some rule engines are used to assist in analysis, they can only handle some fixed format orders and have poor adaptability to complex and variable order formats.

[0004] In addition, the traditional method often identifies customer attributes based on pre-set rules and limited historical data. For example, customers are divided into different levels according to their purchase frequency and purchase amount, but this classification method is rough and cannot fully reflect the true attributes of customers. For the determination of settlement rules and delivery rules, experience or fixed templates are usually used, lacking consideration of personalized needs of customers and unable to dynamically adjust according to real-time business conditions.

[0005] Traditional order settlement mainly relies on manual calculation, and financial personnel need to manually calculate order revenue settlement data according to order information and settlement rules. This process involves a lot of numerical calculations and checking work, which is prone to calculation errors and low efficiency. Due to the delay in data processing, the settlement period may be prolonged, affecting the accuracy of the enterprise's cash flow and financial statements.

[0006] Under the traditional method, the delivery of order revenue settlement data is usually completed by manual operation, such as printing reports and sending emails. This not only increases labor costs, but also is prone to problems such as delayed and inaccurate data transmission. The execution of delivery rules also lacks effective supervision and automation mechanisms, which may lead to delayed delivery or failure to meet customer requirements. SUMMARY

[0007] To address the technical problems existing in the prior art, the present invention provides the following technical solution: On the one hand, an LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters is provided. This method is implemented by an electronic device and includes: S1. Sales orders are input into the business system from the front end and pre-processed; S2. The business system calls the LLM large model to parse the sales order input into the business system and automatically obtains the document number, order customer information and order document information. S3. Input the order customer information and the order document information into the LSTM business and finance intelligent classification driving model pre-deployed on the business system. The LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes. S4. Based on the settlement rules, settle the sales order to obtain the corresponding order revenue settlement data; S5. Feed back the order revenue settlement data to the front end and simultaneously inform the delivery rules, so that the front end can automatically deliver the order revenue settlement data to the client according to the delivery rules.

[0008] Preferably, in S2, before calling the LLM large model to parse the sales order input into the business system and automatically obtain the order customer information and order document information, the following steps are also included: An LLM API is deployed on the business system. The LLM API is used to log in to a third-party LLM platform and request to invoke the LLM large model.

[0009] Preferably, in S2, after calling the LLM large model to parse the sales order input into the business system and automatically obtaining the order customer information and order document information, the method further includes: The business system imports the order customer information and the order document information into the dashboard of the business system, and the dashboard visually displays the order customer information and the order document information of the sales order.

[0010] Preferably, the method for constructing the LSTM business and financial intelligent classification-driven model includes: Collect sales orders, settlement records, and delivery records from several order customers from the accounts receivable / payable system and the initial data system; Feature engineering is performed on the settlement records and delivery records to extract corresponding settlement rule features and delivery behavior rule features. The settlement rule features include the following: customer attributes, settlement amount, settlement quantity, unit price, discount rate, tax rate, whether the settlement includes tax, and settlement date. The customer attributes indicate whether the customer is a one-time customer: if so, a fixed discount rate, tax rate, and tax-inclusive settlement rule are automatically associated; otherwise, a recursive discount rate is applied based on the customer's transaction batches. The delivery behavior rule features include delivery progress, delivery date, delivery amount, delivery quantity, delivery ratio, and delivery document number. The settlement rule features and delivery behavior rule features of each order customer are statistically analyzed, and the component is a feature set consisting of {settlement rule features, delivery behavior rule features}. The feature set is divided into a training set, a validation set, and a test set according to a preset ratio; The feature set is input into the initial LSTM model for initial training and learning, generating the initial LSTM business intelligence classification-driven model. The LSTM business and financial intelligent classification-driven model was tested using a test set. The test passed, and the accuracy of the LSTM business and finance intelligent classification-driven model in identifying customer attributes and predicting the corresponding settlement and delivery rules was verified using the validation set. If the verification is successful, the LSTM business and financial intelligent classification-driven model will be deployed and applied to the business system. If the verification fails, repeat the above steps to reconstruct the LSTM business and financial intelligent classification driven model.

[0011] Preferably, in step S3, after the LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes, the method further includes: The business system calls the LLM big model to parse the settlement rules and obtain the corresponding settlement keywords and their settlement relationships; A corresponding settlement template is generated based on the settlement keywords and the settlement relationship, wherein the settlement template is used for: After settling the sales order based on the settlement rules and obtaining the corresponding order revenue settlement data, the order revenue settlement data is written and saved as the corresponding settlement record; Both the settlement template and the settlement record are bound to the document number.

[0012] Preferably, in S5, after feeding back the order revenue settlement data to the front end and simultaneously informing it of the delivery rules, and the front end automatically delivers the order revenue settlement data to the client according to the delivery rules, the method further includes: The business system generates corresponding delivery logs based on the delivery process, and generates corresponding delivery records based on the delivery logs and the delivery rules. The delivery record is linked to the document number.

[0013] Preferably, in S5, after feeding back the order revenue settlement data to the front end and simultaneously informing it of the delivery rules, and the front end automatically delivers the order revenue settlement data to the client according to the delivery rules, the method further includes: The client confirms and provides feedback on the revenue settlement data of the delivered order; The front end forwards the client's confirmation feedback on the order revenue settlement data to the business system; After the business system receives confirmation feedback from the client regarding the order revenue settlement data, it reads the settlement record and delivery record bound to the document number and imports them into the dashboard of the business system, where the dashboard visually displays the settlement record and delivery record of the sales order.

[0014] On the other hand, an LLM and LSTM-driven intelligent classification and recognition management device for business and finance is provided. This device is used to implement the aforementioned LLM and LSTM-driven intelligent classification and recognition management method for business and finance. The device includes: The front end is used to input sales orders into the business system and preprocess them; The business system is used to call the LLM large model to parse the sales orders input into the business system, automatically obtaining the order number, order customer information, and order document information; and inputting the order customer information and the order document information into the LSTM business and financial intelligent classification driven model pre-deployed on the business system, whereby the LSTM business and financial intelligent classification driven model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes; and, based on the settlement rules, settles the sales orders to obtain the corresponding order revenue settlement data; and, feeds back the order revenue settlement data to the front end and simultaneously informs the delivery rules, whereby the front end automatically delivers the order revenue settlement data to the client according to the delivery rules. The client is used to populate the sales order back to the front end and to receive the order revenue settlement data; The client communicates with the front-end. The front end is connected to the business system.

[0015] On the other hand, an electronic device is provided, comprising: a processor; and a memory storing computer-readable instructions, which, when executed by the processor, implement any one of the above-described LLM and LSTM-driven intelligent classification and recognition management methods for business and finance.

[0016] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the methods in the above-described LLM and LSTM-driven intelligent classification and recognition management methods for business and finance. The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention automates the input of sales orders at the front end and utilizes an LLM (Large-Scale Management) model for parsing, enabling rapid and accurate acquisition of key order information and significantly reducing order processing time. The LSTM (Large-Scale Management) business and finance intelligent classification-driven model automatically identifies customer attributes and predicts settlement and delivery rules, reducing the time spent on manual judgment and decision-making and improving the efficiency of business processes. The entire business process is highly automated, with each step from order input to data delivery completed quickly, accelerating the company's cash flow and business response speed.

[0017] The LLM large-scale model boasts powerful language understanding and information extraction capabilities, accurately parsing sales orders in various formats and avoiding errors from manual parsing. The LSTM business and finance intelligent classification-driven model, trained on extensive historical data, more accurately identifies customer attributes and prediction rules, improving the accuracy of order settlement and delivery. Automated data processing and delivery minimize human interference, ensuring data accuracy and consistency.

[0018] The LLM large-scale model and LSTM business and financial intelligent classification-driven model of this invention have good adaptability and scalability, and can be adjusted and optimized according to different business scenarios and customer needs. Settlement and delivery rules can be adjusted in real time based on market changes and customer feedback to meet personalized customer needs and improve the company's market competitiveness. Automated business processes reduce reliance on manual labor and lower labor costs. At the same time, by reducing errors and delays, additional cost expenditures are avoided, improving the company's economic efficiency. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of an LLM and LSTM-driven intelligent classification and recognition management method for business and finance provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the training mechanism of an LSTM business intelligence classification driven model provided in an embodiment of the present invention; Figure 3 This is a block diagram of an LLM and LSTM driven intelligent classification and recognition management device provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0022] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0023] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0024] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0025] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0026] This invention provides an LLM and LSTM-driven intelligent classification and recognition management method for business and financial information. This method can be implemented by an electronic device, which can be a terminal or a server. Figure 1 The flowchart shown is for an LLM and LSTM-driven intelligent classification and recognition management method for business and finance. The processing flow of this method may include the following steps: S1. Sales orders are input into the business system from the front end and pre-processed; S2. Deploy an LLM API on the business system. The LLM API is used to log in to a third-party LLM platform and request to call the LLM large model. The business system calls the LLM large model to parse the sales order input into the business system, and automatically obtains the document number, order customer information and order document information. S3. Input the order customer information and the order document information into the LSTM business and finance intelligent classification driving model pre-deployed on the business system. The LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes. S4. Based on the settlement rules, settle the sales order to obtain the corresponding order revenue settlement data; S5. Feed back the order revenue settlement data to the front end and simultaneously inform the delivery rules, so that the front end can automatically deliver the order revenue settlement data to the client according to the delivery rules.

[0027] Combined with appendix Figure 2 As shown, this section utilizes the business system for adaptive identification and execution of business, financial, and tax matters. It combines LLM and LSTM-driven intelligent identification and application technologies to achieve intelligent processing and efficient execution of business, financial, and tax information. The implementation process of each step will be described in detail below.

[0028] S1. Sales orders are input from the front end into the business system and preprocessed. The front-end serves as the interface for users to interact with the business system. Users enter relevant information about sales orders on the front-end, such as basic order information, product details, and customer information. After receiving this data, the business system performs preprocessing operations. The purpose of this preprocessing is to convert the input data into a format that the business system can effectively process, remove invalid data, fill in missing data, and standardize the data format, thus preparing for subsequent processing flows.

[0029] The front-end can use HTML, CSS, and JavaScript to build the user interface, providing forms for users to enter sales order information. Front-end frameworks such as Vue.js and React can be used to improve development efficiency and user experience. Data input from the front-end is sent to the back-end server of the business system via the HTTP protocol. Asynchronous data transmission can be achieved using AJAX technology, avoiding page refreshes.

[0030] In the backend of the business system, data processing scripts are written using programming languages ​​such as Python and Java. Regular expressions and data cleaning libraries (such as Pandas) are used to clean and transform the input data.

[0031] Suppose the front end is an order entry page for an e-commerce website, where users fill in information such as product name, quantity, customer name, and contact information. The front end sends this information to the back end server in JSON format. After receiving the data, the back end server checks if the customer name is empty; if so, it prompts the user to fill in the information; it converts the product quantity to an integer type; and it removes special characters from the product name, etc.

[0032] Preprocessing removes invalid data and fills in missing data, improving data accuracy and completeness and providing a reliable data foundation for subsequent analysis and processing. Standardizing data formats enables business systems to better handle data from different sources, improving system compatibility and scalability.

[0033] S2. Deploy an LLM API on the business system. The LLM API is used to log in to a third-party LLM platform and request to call the LLM big model. The business system calls the LLM big model to parse the sales order input into the business system and automatically obtain the document number, order customer information, and order document information. The business system establishes a connection with a third-party LLM platform by deploying an LLM API. The LLM API is responsible for completing the authentication process for logging into the third-party platform and obtaining access permissions. The business system uses sales order data as input and sends it to the large model of the third-party LLM platform via the LLM API. The large model utilizes its powerful natural language processing capabilities to understand and analyze the sales order text, extracting key information such as document number, order customer information, and order document information.

[0034] An LLM API can be deployed on the backend server of the business system. API services can be built using programming languages ​​such as Python's Flask and Django frameworks. Authentication is performed using API keys, OAuth, or other methods according to the authentication requirements of the third-party LLM platform to ensure legitimate access. Sales order data is encapsulated into a format conforming to the LLM API requirements (such as JSON) and sent via HTTP requests to the third-party LLM platform. The parsed results returned by the platform are received and processed.

[0035] For example, a business system might use the Flask framework in Python to deploy an LLM API. When calling the API, authentication is performed using an API key provided by the third-party LLM platform. The text information of the sales order is sent as the request body to the LLM API, which forwards the request to the third-party platform. The third-party platform's big data model parses the order text and returns JSON data containing information such as the order number, customer name, and order amount. The business system receives this data and stores it in its database.

[0036] Large LLM models possess powerful natural language understanding capabilities, enabling them to quickly and accurately extract key information from sales order text, reducing manual processing workload and error rates. The input text and request parameters can be adjusted to adapt to various sales order formats and content based on different business needs. Utilizing large models from third-party LLM platforms eliminates the need to train and maintain complex natural language processing models independently, saving significant development time and costs.

[0037] S3. Input the order customer information and the order document information into the LSTM business and finance intelligent classification driving model pre-deployed on the business system. The LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes. Combined with appendix Figure 3 As shown, preferably, the method for constructing the LSTM business and financial intelligent classification-driven model includes: Collect sales orders, settlement records, and delivery records from several order customers from the accounts receivable / payable system and the initial data system; Feature engineering is performed on the settlement records and delivery records to extract corresponding settlement rule features and delivery behavior rule features. The settlement rule features include the following: customer attributes, settlement amount, settlement quantity, unit price, discount rate, tax rate, whether the settlement includes tax, and settlement date. The customer attributes indicate whether the customer is a one-time customer: if so, a fixed discount rate, tax rate, and tax-inclusive settlement rule are automatically associated; otherwise, a recursive discount rate is applied based on the customer's transaction batches. The delivery behavior rule features include delivery progress, delivery date, delivery amount, delivery quantity, delivery ratio, and delivery document number. The settlement rule features and delivery behavior rule features of each order customer are statistically analyzed, and the component is a feature set consisting of {settlement rule features, delivery behavior rule features}. The feature set is divided into a training set, a validation set, and a test set according to a preset ratio; The feature set is input into the initial LSTM model for initial training and learning, generating the initial LSTM business intelligence classification-driven model. The LSTM business and financial intelligent classification-driven model was tested using a test set. The test passed, and the accuracy of the LSTM business and finance intelligent classification-driven model in identifying customer attributes and predicting the corresponding settlement and delivery rules was verified using the validation set. If the verification is successful, the LSTM business and financial intelligent classification-driven model will be deployed and applied to the business system. If the verification fails, repeat the above steps to reconstruct the LSTM business and financial intelligent classification driven model.

[0038] The following will describe in detail the specific steps for building a model based on LSTM: (a) Data collection Collect sales orders, settlement records, and delivery records from several order customers from the accounts receivable / payable system and the initial data system.

[0039] Data source: Clarify the data structure and storage method of these two systems. For example, the accounts receivable / payable system may store settlement-related information in the form of database tables, while the initial data system may store basic order information.

[0040] Scope of data collection: Determine the scope of customers to be collected from. This can be done by filtering based on factors such as time and customer size to ensure that the collected data is representative.

[0041] Data cleaning: Perform preliminary cleaning on the collected data to remove duplicate, erroneous, or missing data, ensuring data quality.

[0042] (ii) Feature Engineering Feature engineering is performed on the settlement records and delivery records to extract the corresponding settlement rule features and delivery behavior rule features.

[0043] Settlement rule features Customer Attributes: Determine if a customer is a one-time customer. This can be done by querying the customer's transaction history. If the customer has only one transaction record, they are considered a one-time customer. For one-time customers, a fixed discount rate, tax rate, and tax-inclusive settlement rule are applied. For non-one-time customers, the discount rate is recursively applied based on the number of transactions.

[0044] Settlement Amount: The total settlement amount is extracted directly from the settlement records.

[0045] Settlement Quantity: The quantity of goods or services for which settlement has been obtained.

[0046] Unit price: The unit price is obtained by dividing the settlement amount by the settlement quantity.

[0047] Discount Rate: A fixed discount rate is used for one-time customers; for non-one-time customers, a recursive discount rate is calculated based on the number of transaction batches. For example, the discount rate increases by a certain percentage for each additional transaction batch.

[0048] Tax rate: A fixed tax rate is used for one-time customers, while the tax rate for non-one-time customers is determined according to business rules.

[0049] Does the settlement include tax? Clarify from the settlement records whether the settlement includes tax.

[0050] Settlement Date: Records the specific date on which the settlement occurred.

[0051] Delivery behavior rules and characteristics Delivery progress: Delivery progress can be calculated by comparing the actual delivered quantity with the order required delivery quantity, such as Delivery progress = Actual delivered quantity / Order required delivery quantity.

[0052] Delivery Date: Records the actual delivery date.

[0053] Delivery Amount: The total amount to be delivered.

[0054] Delivery quantity: The quantity of goods or services actually delivered.

[0055] Delivery ratio: the ratio of the number of deliveries to the number of orders.

[0056] Delivery document number: Retrieve a unique delivery document number from the delivery record.

[0057] (III) Feature Set Construction We statistically analyze the settlement rule characteristics and delivery behavior rule characteristics of each order customer, constructing a feature set consisting of {settlement rule characteristics, delivery behavior rule characteristics}. We then combine the settlement rule characteristics and delivery behavior rule characteristics of each order customer to form a complete feature set for subsequent model training.

[0058] (iv) Dataset partitioning The feature set is divided into training, validation, and test sets according to a preset ratio. A typical ratio is 70%, 15%, and 15%.

[0059] Training set: Used to train the LSTM model, allowing the model to learn patterns and rules in the data.

[0060] Validation set: Used during model training to evaluate model performance, adjust model hyperparameters, and prevent overfitting.

[0061] Test set: After the model training is completed, it is used to finally evaluate the model's performance and test the model's generalization ability.

[0062] (v) Model Training The feature set is input into the initial LSTM model for initial training and learning, generating the initial LSTM business intelligence classification-driven model.

[0063] Model initialization: Select a suitable LSTM model structure and determine hyperparameters such as the number of layers and neurons.

[0064] Training process: The model is trained using the training set, and the weights are continuously adjusted through backpropagation to minimize the loss function. Mean squared error (MSE) can be chosen as the loss function, and stochastic gradient descent (SGD) or the Adam optimizer can be used for optimization.

[0065] Number of training epochs: Determine the appropriate number of training epochs to avoid overfitting or underfitting the model. The optimal number of training epochs can be determined by observing the loss function values ​​on the validation set.

[0066] (vi) Model Testing The LSTM business and financial intelligent classification-driven model was tested using a test set.

[0067] Test metrics: Select appropriate test metrics, such as accuracy, recall, F1 score, etc., to evaluate the model's performance on the test set.

[0068] Testing process: Input the features of the test set into the trained model, obtain the model's prediction results, compare them with the true labels, and calculate the test metrics.

[0069] (vii) Model Validation After the test is passed, the accuracy of the LSTM business and finance intelligent classification-driven model in identifying customer attributes and predicting the settlement rules and delivery rules corresponding to the customer attributes is verified using the validation set.

[0070] Validation metrics: The same metrics as accuracy, recall, and F1 score are used to evaluate the model's performance on the validation set.

[0071] Validation process: Input the features of the validation set into the model to determine whether the model accurately identifies customer attributes and whether the predicted settlement and delivery rules conform to the actual situation.

[0072] (viii) Model Deployment and Reconfiguration Validation passed: If the validation is successful, the LSTM business and financial intelligent classification-driven model will be deployed and applied to the business system, allowing the model to play a role in actual business operations.

[0073] Validation failed: If validation fails, repeat the above steps to reconstruct the LSTM business intelligence classification-driven model. Model performance can be improved by adjusting hyperparameters, increasing training data, and improving feature engineering.

[0074] The following is an example: Suppose that sales orders, settlement records, and delivery records of 1,000 order customers were collected from the accounts receivable / payable system and the initial data system, with a time range of the past year.

[0075] Perform feature engineering: (1) Settlement rule characteristics For customer A, a review of the transaction history reveals that they are a one-time customer with a fixed discount rate of 0.9%, a tax rate of 0.13%, a settlement amount of 1000 yuan, a settlement quantity of 10 items, a unit price of 100 yuan, a settlement including tax, and a settlement date of October 1, 2023.

[0076] For customer B, the transaction batch consists of 3 transactions. According to the recursive discount rule, the discount rate is 0.85, the tax rate is determined to be 0.1% according to the business rules, the settlement amount is 2,000 yuan, the settlement quantity is 20 pieces, the unit price is 100 yuan, the settlement includes tax, and the settlement date is November 1, 2023.

[0077] (2) Characteristics of delivery behavior rules Customer A's order requested delivery of 10 items. 10 items were actually delivered, the delivery progress is 100%, the delivery date is October 5, 2023, the delivery amount is 1000 yuan, the delivery quantity is 10 items, the delivery ratio is 100%, and the delivery document number is 001.

[0078] Customer B's order requested delivery of 20 items, but 18 items were actually delivered, representing a delivery progress of 90%. The delivery date was November 5, 2023, the delivery amount was 1800 yuan, the delivery quantity was 18 items, the delivery ratio was 90%, and the delivery document number was 002.

[0079] The settlement rule features and delivery behavior rule features of each customer are combined to form a feature set containing 1000 samples. The feature set is then divided into a training set (700 samples), a validation set (150 samples), and a test set (150 samples) in a ratio of 70%, 15%, and 15%, respectively.

[0080] Choose a 2-layer LSTM model with 64 neurons in each layer. Train the model using the training set for 50 epochs with a learning rate of 0.001, using the Adam optimizer and the mean squared error loss function.

[0081] The features of the test set are input into the trained model to obtain the model's prediction results. Precision, recall, and F1 score are calculated, assuming a precision of 0.85, a recall of 0.82, and an F1 score of 0.83.

[0082] The model's accuracy in identifying customer attributes and predicting corresponding settlement and delivery rules was validated using a validation set. Assuming an accuracy of 0.8, recall of 0.78, and an F1 score of 0.79, the validation failed. The model's hyperparameters were adjusted by increasing the number of layers in the LSTM model to 3 and the number of neurons per layer to 128. The model was then retrained, tested, and validated until successful, at which point it was deployed to the business system.

[0083] Therefore, the LSTM (Long Short-Term Memory) business and financial intelligent classification-driven model is a deep learning-based model capable of processing sequential data. Taking order customer information and order document information as input sequences, the model learns patterns and rules from historical data to identify customer attributes (such as whether they are one-time customers) and predicts corresponding settlement rules (such as discount rates, tax rates, settlement dates, etc.) and delivery rules (such as delivery schedule, delivery date, etc.) based on these customer attributes.

[0084] Therefore, following the given construction method, sales orders, settlement records, and delivery records of order customers can be collected from the accounts receivable / payable system and the initial data system. Feature engineering is then performed on this data to extract settlement rule features and delivery behavior rule features, forming a feature set. The feature set is divided into training, validation, and test sets. An initial LSTM model is constructed using Python deep learning frameworks such as TensorFlow and PyTorch. The training set is then input into the model for training and learning, generating an LSTM business and finance intelligent classification-driven model.

[0085] The trained LSTM business and financial intelligent classification-driven model can be deployed to the backend server of the business system using Docker containerization technology to ensure the model's portability and stability. Within the business system, order customer information and order document information are converted into an input format acceptable to the model, and the deployed LSTM model is then invoked for prediction.

[0086] For example, the business system has already trained an LSTM business and financial intelligent classification-driven model. When a new sales order is received, customer information (such as the number of historical transactions) and order document information (such as the order amount) are extracted from the order. This information is then input into the model, which determines that the customer is a one-time customer and predicts the corresponding settlement rules as a fixed discount rate of 10%, a tax rate of 13%, and tax-inclusive settlement, with delivery within 3 business days after order confirmation.

[0087] Therefore, LSTM models can capture long-term dependencies in sequence data, accurately predict customer attributes and rules, and improve the accuracy of business decisions. The model can automatically identify customer attributes and predict rules, reducing manual intervention and improving business processing efficiency.

[0088] S4. Based on the settlement rules, settle the sales order to obtain the corresponding order revenue settlement data. The business system processes sales orders based on settlement rules predicted by the LSTM business intelligence classification-driven model. These settlement rules include discount rates, tax rates, and whether the settlement includes tax. The system calculates the actual revenue for each order based on these rules, such as deducting discounts and calculating taxes, ultimately obtaining the order revenue settlement data. The specific operations are as follows: Rule analysis: Extract key parameters from the settlement rules, such as discount rate and tax rate; Calculation logic: The calculation is based on information such as the sales order's product amount and quantity, combined with the settlement rules. For example, if the settlement rule is tax-inclusive and the tax rate is 13%, then the order revenue = product amount * (1 + tax rate) - discount amount; Data storage: The calculated order revenue settlement data is stored in the database of the business system.

[0089] Assuming a sales order has a total value of 1000 yuan, with a 10% discount and a 13% tax rate (tax included), the discount amount is 1000 * 10% = 100 yuan, the tax is (1000 - 100) * 13% = 117 yuan, and the order revenue settlement is (1000 - 100) + 117 = 1017 yuan. The business system stores this data in the database.

[0090] Therefore, calculations are performed based on the settlement rules recommended by the model to ensure the accuracy of order revenue settlement data and avoid human calculation errors. Unified settlement rules and calculation logic guarantee consistency in settlement across different orders, facilitating financial accounting and management. The automated settlement process improves settlement efficiency and reduces manual processing time and workload.

[0091] S5. Feed the order revenue settlement data back to the front end and simultaneously inform it of the delivery rules. The front end then automatically delivers the order revenue settlement data to the client according to the delivery rules. The business system sends the calculated order revenue settlement data and predicted delivery rules to the front end. After receiving this information, the front end delivers the order revenue settlement data to the client in an appropriate manner, such as by sending emails or SMS notifications, according to the delivery rules (e.g., delivery schedule, delivery date).

[0092] Order revenue settlement data and delivery rules can be sent from the backend of the business system to the frontend using the HTTP protocol, with real-time data push implemented using WebSocket. After receiving the data, the frontend uses JavaScript to write scripts to generate corresponding notification content, such as email templates or SMS templates, based on the delivery rules. The notification content is then sent to the client by calling email sending APIs (such as SendGrid) or SMS sending APIs (such as Alibaba Cloud SMS Service).

[0093] For example, the business system can send the order revenue settlement data of 1017 yuan and the delivery rule (delivery within 3 business days after order confirmation) to the front end. The front end generates an email with the content "Dear customer, your order has been settled, the order revenue is 1017 yuan, and delivery is expected to be completed within 3 business days after order confirmation." The email is then sent to the customer's inbox via an email sending API.

[0094] Timely feedback of order revenue settlement data and delivery rules to the client improves customer satisfaction and trust. The front end automatically delivers data to the client based on delivery rules, reducing manual operations and improving delivery efficiency.

[0095] Preferably, in S2, after calling the LLM large model to parse the sales order input into the business system and automatically obtaining the order customer information and order document information, the method further includes: The business system imports the order customer information and the order document information into the dashboard of the business system, and the dashboard visually displays the order customer information and the order document information of the sales order.

[0096] Dashboards are system visualization components that categorize and display imported information. Business systems choose a suitable dashboard tool to visualize their data. Open-source tools such as Grafana and Metabase can be used, as well as commercial tools such as Tableau and Power BI. When selecting a tool, factors such as functionality, ease of use, compatibility, and cost need to be considered.

[0097] Integrate the selected dashboard tool into the business system, enabling it to interact with the business system's data. This may require using the API or SDK provided by the dashboard tool to programmatically implement data transfer and display. For example, use a RESTful API to send order customer information and order document information from the business system to the dashboard tool.

[0098] Configure the dashboard tools to create corresponding visualization components to display order customer information and order document information. Different visualization types such as bar charts, line charts, pie charts, and tables can be used to display different data. For example, a bar chart can be used to show the distribution of order amounts for different customers, and a table can be used to display detailed order information. At the same time, you need to set the data source, data filtering conditions, and sorting rules for the visualization components to ensure that the displayed data is accurate and clear.

[0099] To ensure the real-time nature of the data displayed on the dashboard, a data update mechanism needs to be set up. Different update frequencies can be configured according to business needs, such as real-time updates and scheduled updates (e.g., daily, hourly). For real-time updates, message queues or real-time data transmission protocols (such as WebSocket) can be used to push data from the business system to the dashboard tool in real time; for scheduled updates, scheduled task scheduling tools (such as Cron or Windows Task Scheduler) can be used to periodically execute the data import program.

[0100] Preferably, in step S3, after the LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes, the method further includes: The business system calls the LLM big model to parse the settlement rules and obtain the corresponding settlement keywords and their settlement relationships; A corresponding settlement template is generated based on the settlement keywords and the settlement relationship, wherein the settlement template is used for: After settling the sales order based on the settlement rules and obtaining the corresponding order revenue settlement data, the order revenue settlement data is written and saved as the corresponding settlement record; Both the settlement template and the settlement record are bound to the document number.

[0101] The business system establishes an interface connection with the LLM (Local Management Model) big data model, transmitting settlement rules to the model in text form. Leveraging its powerful natural language processing capabilities, the LLM big data model performs deep semantic analysis on the settlement rules. It identifies key settlement-related information in the rule text, extracts settlement keywords, and analyzes the logical relationships between these keywords, such as their order and conditional dependencies, thereby determining the settlement relationships. For example, if the settlement rule is "When the product type of a sales order is A and the sales quantity is greater than 10, settlement will be made at 90% of the unit price," the LLM big data model will extract settlement keywords such as "product type," "sales quantity," "unit price," and "90% discount," and determine that "product type A" and "sales quantity greater than 10" are the conditions that trigger "90% discount settlement."

[0102] The business system constructs a structured settlement template based on the settlement keywords and relationships obtained from parsing the LLM model. This template defines the position and role of each settlement keyword in the settlement process, as well as how to perform calculations and data population based on settlement relationships. The template reserves some fields for subsequent writing of order revenue settlement data. The settlement template generated by the above settlement rules may include fields such as "Product Type," "Sales Quantity," "Unit Price," "Discount," and "Settlement Amount." The "Discount" field is determined based on the conditional relationship between "Product Type" and "Sales Quantity," and the "Settlement Amount" is calculated by multiplying the "Unit Price" by the "Sales Quantity" and then by the "Discount."

[0103] The business system retrieves detailed sales order information and matches the relevant data in the order with settlement rules and settlement templates. Based on the conditions and calculation methods defined in the settlement rules, the sales order is settled, yielding order revenue settlement data. This settlement data is then written into the corresponding fields of the previously generated settlement template. For example, consider a sales order with product type A, sales quantity of 15, and unit price of 100 yuan. According to the settlement rules, the "discount" is 0.9, and the "settlement amount" is 15 * 100 * 0.9 = 1350 yuan. These data are then written into the "product type," "sales quantity," "unit price," "discount," and "settlement amount" fields of the settlement template, respectively.

[0104] The business system saves the settlement template containing order revenue settlement data as a settlement record. This settlement record can be stored in the database for subsequent querying, statistics, and auditing operations. The settlement template containing the settlement data is inserted into the "Settlement Record Table" as a record, and each record has a unique identifier.

[0105] The business system associates settlement templates and records with the document numbers of sales orders while saving settlement records. This binding can be achieved by adding corresponding fields to the database. For example, a "Document Number" field can be added to the "Settlement Records Table," and the document number of the sales order can be filled into this field. For instance, if the document number of a sales order is "SO001," then when saving the settlement record, "SO001" will be written into the "Document Number" field of the "Settlement Records Table," thereby binding the settlement template, settlement record, and document number.

[0106] The LLM large-scale model, based on deep learning technology, is trained on a large amount of text data, learning rich linguistic knowledge and semantic representations. It can perform lexical, syntactic, and semantic analysis on the input settlement rule text, identifying key entities (settlement keywords) and the logical relationships between them (settlement relationships). The neural network structure within the model can capture contextual information in the text, thus accurately understanding the meaning of the settlement rules. This provides accurate foundational information for subsequent settlement template generation, avoiding errors and inaccuracies that may occur during manual interpretation of settlement rules.

[0107] The settlement template is generated based on a structured representation of settlement keywords and relationships. Fields and formats are defined according to the type and function of the keywords, and the calculation logic and data flow between fields are determined based on the settlement relationships. This transforms complex settlement rules into an operable template, facilitating subsequent settlement operations and data processing. It improves the standardization and normalization of the settlement process, reduces settlement errors caused by human factors, and enables unified settlement processing for different sales orders.

[0108] The business system processes and calculates sales order data based on settlement rules and the calculation logic defined in the settlement template. By matching order data with the rules in the template, the value of each settlement keyword is determined, and the order revenue settlement data is derived according to the calculation method. This data is then written into the corresponding fields of the template, thus achieving data population and recording.

[0109] Saving settlement records to a database is for long-term storage and management of settlement data. By binding settlement templates and records with document numbers, a link can be established between sales orders and settlement information, facilitating subsequent retrieval and tracing of settlement records based on document numbers, and also enabling statistical analysis of settlement data.

[0110] Preferably, in S5, after feeding back the order revenue settlement data to the front end and simultaneously informing it of the delivery rules, and the front end automatically delivers the order revenue settlement data to the client according to the delivery rules, the method further includes: The business system generates corresponding delivery logs based on the delivery process, and generates corresponding delivery records based on the delivery logs and the delivery rules. Bind the delivery record to the document number; The client confirms and provides feedback on the revenue settlement data of the delivered order; The front end forwards the client's confirmation feedback on the order revenue settlement data to the business system; After the business system receives confirmation feedback from the client regarding the order revenue settlement data, it reads the settlement record and delivery record bound to the document number and imports them into the dashboard of the business system, where the dashboard visually displays the settlement record and delivery record of the sales order.

[0111] In the business system, a corresponding event listening mechanism is set up for each delivery process. When key nodes in the delivery process occur, such as the start of order delivery, partial delivery completion, and full delivery completion, a log recording operation is triggered.

[0112] The log entries should include key information such as the delivery time, details of the delivered items or services, and the person responsible for the delivery. This information can be obtained from relevant database tables in the business system, such as the order table and inventory table.

[0113] Log records are stored in the business system's log database in a structured data format for easy subsequent querying and processing.

[0114] Delivery rules are predefined in the business system. These rules can be configured according to different business scenarios and needs, such as delivery quantity and delivery time range. The business system periodically (or immediately after delivery) reads delivery logs from the log database and filters and analyzes them according to the delivery rules. Delivery logs that meet the delivery rules are integrated to generate delivery records. Delivery records should include key information such as the order number, the quantity of items or services delivered, and the delivery time. Delivery records are also stored in the business system's database in a structured data format.

[0115] After generating a delivery record, the business system retrieves the corresponding document number from the order database. The document number is a crucial piece of information that uniquely identifies an order and is typically generated when the order is created. Through database operations, the delivery record and the document number are associated, establishing a one-to-one or one-to-many binding relationship. This can be achieved by adding a field to the delivery record table to store the document number, or by creating an intermediate table to record this binding relationship.

[0116] The client interface provides users with a function to display order revenue settlement data. This settlement data can be obtained from the business system and updated in real time, including information such as total order revenue, settled amount, and unsettled amount. The client provides a confirmation button or option, allowing users to confirm the order revenue settlement data. When the user clicks the confirmation button, the client encapsulates the confirmation feedback information into a specific data format, such as JSON. The client then sends the confirmation feedback information to the front-end server via a network request.

[0117] After receiving the confirmation feedback from the client, the front-end server verifies and parses the information. It verifies the integrity and legality of the information to ensure it hasn't been tampered with. The front-end server then forwards the confirmation feedback to the business system. This can be done via HTTP requests, message queues, or other methods. Upon receiving the forwarded confirmation feedback, the business system records and processes it. Information such as the time and source of the confirmation feedback can be recorded in the business system's logs.

[0118] Once the business system receives confirmation from the client regarding order revenue settlement data, it retrieves the settlement and delivery records associated with that document number from the database. The business system then organizes and formats these records to conform to the data requirements of the dashboard. For example, it converts the data into a format suitable for chart display.

[0119] The business system imports the compiled settlement and delivery records into the dashboard system. The dashboard system can be an internal module of the business system or a standalone third-party system. Upon receiving the settlement and delivery records, the dashboard system displays the data in charts, tables, and other formats based on pre-defined visualization templates. Users can intuitively view the settlement and delivery records of sales orders through the dashboard, such as order delivery progress and revenue settlement status.

[0120] Combined with appendix Figure 3 As shown, on the other hand, an LLM and LSTM-driven intelligent classification and recognition management device for business and finance is provided. This device is used to implement the aforementioned LLM and LSTM-driven intelligent classification and recognition management method for business and finance. The device includes: The front end is used to input sales orders into the business system and preprocess them; The business system is used to call the LLM large model to parse the sales orders input into the business system, automatically obtaining the order number, order customer information, and order document information; and inputting the order customer information and the order document information into the LSTM business and financial intelligent classification driven model pre-deployed on the business system, whereby the LSTM business and financial intelligent classification driven model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes; and, based on the settlement rules, settles the sales orders to obtain the corresponding order revenue settlement data; and, feeds back the order revenue settlement data to the front end and simultaneously informs the delivery rules, whereby the front end automatically delivers the order revenue settlement data to the client according to the delivery rules. The client is used to populate the sales order back to the front end and to receive the order revenue settlement data; The client communicates with the front-end. The front end is connected to the business system.

[0121] The functions and interactions of the various components in the above device can be understood by referring to the previous steps and methods; they will not be repeated here.

[0122] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 4 As shown, electronic device 410 may include a first processor 2001.

[0123] Optionally, the electronic device 410 may also include a memory 2002 and a transceiver 2003.

[0124] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0125] The following is combined Figure 4 A detailed description of each component of electronic device 410 is provided below: The first processor 2001 is the control center of the electronic device 410. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0126] Optionally, the first processor 2001 can perform various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0127] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0128] In a specific implementation, as one example, the electronic device 410 may also include multiple processors, for example... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0129] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0130] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0131] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0132] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0133] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected via the interface circuit of the electronic device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0134] It should be noted that, Figure 4 The structure of the electronic device 410 shown does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0135] Furthermore, the technical effects of the electronic device 410 can be referenced from the technical effects of the LLM and LSTM driven intelligent classification and recognition management method described in the above method embodiments, and will not be repeated here.

[0136] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0137] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0138] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0139] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0140] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0141] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0142] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0143] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0144] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0145] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0146] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0147] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0148] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A business and financial intelligent classification and recognition management method driven by LLM and LSTM, characterized in that, The method includes: S1. Sales orders are input into the business system from the front end and pre-processed; S2. The business system calls the LLM large model to parse the sales order input into the business system and automatically obtains the document number, order customer information and order document information. S3. Input the order customer information and the order document information into the LSTM business and finance intelligent classification driving model pre-deployed on the business system. The LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes. S4. Based on the settlement rules, settle the sales order to obtain the corresponding order revenue settlement data; S5. Feed back the order revenue settlement data to the front end and simultaneously inform the delivery rules. The front end then automatically delivers the order revenue settlement data to the client according to the delivery rules.

2. The LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters according to claim 1, characterized in that, In S2, before calling the LLM large model to parse the sales order input into the business system and automatically obtain the order customer information and order document information, the following steps are also included: An LLM API is deployed on the business system. The LLM API is used to log in to a third-party LLM platform and request to invoke the LLM large model.

3. The LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters according to claim 1, characterized in that, In S2, after calling the LLM large model to parse the sales order input into the business system and automatically obtaining the order customer information and order document information, it also includes: The business system imports the order customer information and the order document information into the dashboard of the business system, and the dashboard visually displays the order customer information and the order document information of the sales order.

4. The LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters according to claim 1, characterized in that, The method for constructing the LSTM business and financial intelligent classification-driven model includes: Collect sales orders, settlement records, and delivery records from several order customers from the accounts receivable / payable system and the initial data system; Feature engineering is performed on the settlement records and delivery records to extract corresponding settlement rule features and delivery behavior rule features. The settlement rule features include the following: customer attributes, settlement amount, settlement quantity, unit price, discount rate, tax rate, whether the settlement includes tax, and settlement date. The customer attributes indicate whether the customer is a one-time customer: if so, a fixed discount rate, tax rate, and tax-inclusive settlement rule are automatically associated; otherwise, a recursive discount rate is applied based on the customer's transaction batches. The delivery behavior rule features include delivery progress, delivery date, delivery amount, delivery quantity, delivery ratio, and delivery document number. The settlement rule features and delivery behavior rule features of each order customer are statistically analyzed, and the component is a feature set consisting of {settlement rule features, delivery behavior rule features}. The feature set is divided into a training set, a validation set, and a test set according to a preset ratio; The feature set is input into the initial LSTM model for initial training and learning, generating the initial LSTM business intelligence classification-driven model. The LSTM business and financial intelligent classification-driven model was tested using a test set. The test passed, and the accuracy of the LSTM business and finance intelligent classification-driven model in identifying customer attributes and predicting the corresponding settlement and delivery rules was verified using the validation set. If the verification is successful, the LSTM business and financial intelligent classification-driven model will be deployed and applied to the business system. If the verification fails, repeat the above steps to reconstruct the LSTM business and financial intelligent classification driven model.

5. The LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters according to claim 1, characterized in that, In S3, after the LSTM business and finance intelligent classification driving model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes, it also includes: The business system calls the LLM big model to parse the settlement rules and obtain the corresponding settlement keywords and their settlement relationships; A corresponding settlement template is generated based on the settlement keywords and the settlement relationship, wherein the settlement template is used for: After settling the sales order based on the settlement rules and obtaining the corresponding order revenue settlement data, the order revenue settlement data is written and saved as the corresponding settlement record; Both the settlement template and the settlement record are bound to the document number.

6. The LLM and LSTM-driven intelligent classification and recognition management method for business and financial matters according to claim 5, characterized in that, In S5, after feeding back the order revenue settlement data to the front end and simultaneously informing it of the delivery rules, and the front end automatically delivers the order revenue settlement data to the client according to the delivery rules, the process further includes: The business system generates corresponding delivery logs based on the delivery process, and generates corresponding delivery records based on the delivery logs and the delivery rules. The delivery record is linked to the document number.

7. The LLM and LSTM-driven intelligent classification and recognition management method for business and finance as described in claim 6, characterized in that, In S5, after feeding back the order revenue settlement data to the front end and simultaneously informing it of the delivery rules, and the front end automatically delivers the order revenue settlement data to the client according to the delivery rules, the process further includes: The client confirms and provides feedback on the revenue settlement data of the delivered order; The front end forwards the client's confirmation feedback on the order revenue settlement data to the business system; After the business system receives confirmation feedback from the client regarding the order revenue settlement data, it reads the settlement record and delivery record bound to the document number and imports them into the dashboard of the business system, where the dashboard visually displays the settlement record and delivery record of the sales order.

8. An LLM and LSTM-driven intelligent classification and recognition management device for business and finance, wherein the LLM and LSTM-driven intelligent classification and recognition management device is used to implement the LLM and LSTM-driven intelligent classification and recognition management method for business and finance as described in any one of claims 1-7, characterized in that, The device includes: The front end is used to input sales orders into the business system and preprocess them; The business system is used to call the LLM large model to parse the sales orders input into the business system, automatically obtaining the order number, order customer information, and order document information; and inputting the order customer information and the order document information into the LSTM business and financial intelligent classification driven model pre-deployed on the business system, whereby the LSTM business and financial intelligent classification driven model identifies customer attributes and predicts the settlement rules and delivery rules corresponding to the customer attributes; and, based on the settlement rules, settles the sales orders to obtain the corresponding order revenue settlement data; and, feeds back the order revenue settlement data to the front end and simultaneously informs the delivery rules, whereby the front end automatically delivers the order revenue settlement data to the client according to the delivery rules. The client is used to populate the sales order back to the front end and to receive the order revenue settlement data; The client communicates with the front-end. The front end is connected to the business system.

9. An electronic device, characterized in that, The electronic device includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images