Account classification method and device thereof, electronic device, and storage medium

Abstract

The application discloses an account classification method and device, electronic equipment and storage medium, and relates to the field of financial technology or other related technical fields, wherein the method comprises: obtaining account data, and preprocessing the account data to obtain account text data; inputting the account text data into a natural language processing model, performing semantic analysis on the account text data through the natural language processing model, and extracting multi-dimensional account features; classifying the account based on the multi-dimensional account features through a rule classification engine to obtain an initial classification result; and inputting the initial classification result and the multi-dimensional account features into a clustering calibration engine, performing unsupervised calibration on the initial classification result by the clustering calibration engine through a clustering algorithm to obtain a final account classification result. The application solves the technical problems of low account classification accuracy and efficiency in the related art, which relies on manual division or account classification based on experience rule matching.

Description

Technical Field

[0001] This invention relates to the field of financial technology or other related technical fields, and more specifically, to an account classification method and apparatus, electronic device and storage medium. Background Technology

[0002] With the deepening of financial digital transformation and the rapid development of fintech businesses, account systems are becoming increasingly complex, and account profile data is growing exponentially. Account classification management is a fundamental step for financial institutions to build customer profiles, monitor anomalies, control risks, and conduct refined operations. Accurate and efficient account classification not only helps to quickly identify account nature, fund flows, and business scenarios, but also significantly improves profile retrieval efficiency, reduces compliance risks, and provides data support for subsequent business decisions.

[0003] In related technologies, account classification mainly relies on manual division or classification based on empirical rules, which is prone to misclassification and omission, resulting in low classification accuracy and efficiency.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides an account classification method, apparatus, electronic device, and storage medium to at least solve the technical problems of low accuracy and efficiency in account classification when relying on manual division or matching based on empirical rules in related technologies.

[0006] According to one aspect of the present invention, an account classification method is provided, comprising: acquiring account information and preprocessing the account information to obtain account text data; inputting the account text data into a natural language processing model, performing semantic parsing on the account text data through the natural language processing model to extract multi-dimensional account features; classifying the accounts based on the multi-dimensional account features using a rule-based classification engine to obtain an initial classification result; inputting the initial classification result and the multi-dimensional account features into a clustering calibration engine, and having the clustering calibration engine perform unsupervised calibration on the initial classification result using a clustering algorithm to obtain a final account classification result.

[0007] Furthermore, the preprocessing steps for the account information include: determining the data type of the account information; if the data type is non-text, extracting text from the account information to generate data text; and performing data cleaning and word segmentation on the data text to obtain the account text data.

[0008] Further, the step of semantically parsing the account text data and extracting multidimensional account features through the natural language processing model includes: inputting the account text data into the natural language processing model, generating a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and extracting the multidimensional account features based on the text vector representation through a feature extraction network, wherein the multidimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0009] Furthermore, the step of classifying accounts through the rule classification engine includes: loading preset classification rules from the rule base through the rule classification engine, wherein each classification rule consists of a logical condition expression based on account transaction features, account attribute features, and account association features; matching each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule through the rule classification engine, and determining the account category based on the matching result.

[0010] Further, the step of inputting the initial classification result and the multidimensional account features into the clustering calibration engine, and having the clustering calibration engine perform unsupervised calibration on the initial classification result using a clustering algorithm, includes: inputting the initial classification result and the multidimensional account features into the clustering calibration engine; binding the initial classification result and the multidimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multidimensional account features; performing unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; for each cluster, if the number of accounts with the same initial category label in the cluster is greater than a preset threshold, using the initial category labels with the same initial category label greater than the preset threshold as the dominant labels for each account within the cluster.

[0011] Furthermore, after using initial category labels with a number greater than a preset threshold as the dominant labels for each account within the cluster, the method further includes: in the case where the dominant label does not exist in the cluster, configuring new account labels for each account within the cluster based on the center vector of the cluster.

[0012] Furthermore, after obtaining the final account classification result, the method also includes: associating and storing account information and account tags based on the account classification result, and establishing an account profile index.

[0013] According to another aspect of the present invention, an account classification apparatus is also provided, comprising: an acquisition unit, configured to acquire account information and preprocess the account information to obtain account text data; a parsing unit, configured to input the account text data into a natural language processing model, and perform semantic parsing on the account text data through the natural language processing model to extract multi-dimensional account features; a classification unit, configured to classify accounts based on the multi-dimensional account features using a rule-based classification engine to obtain an initial classification result; and a calibration unit, configured to input the initial classification result and the multi-dimensional account features into a clustering calibration engine, and perform unsupervised calibration on the initial classification result using a clustering algorithm by the clustering calibration engine to obtain a final account classification result.

[0014] Furthermore, the acquisition unit includes: a first extraction module, used to determine the data type of the account data, and if the data type is a non-text type, to extract text from the account data to generate data text; and a first processing module, used to perform data cleaning and word segmentation on the data text to obtain the account text data.

[0015] Further, the parsing unit includes: a first generation module, used to input the account text data into the natural language processing model, and generate a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and a second extraction module, used to extract the multi-dimensional account features based on the text vector representation through a feature extraction network, wherein the multi-dimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0016] Furthermore, the classification unit includes: a first loading module, used to load preset classification rules from the rule base through the rule classification engine, wherein each classification rule is composed of a logical condition expression based on account transaction features, account attribute features, and account association features; and a first matching module, used to match each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule through the rule classification engine, and determine the account category based on the matching result.

[0017] Further, the calibration unit includes: a first binding module, used to input the initial classification result and the multidimensional account features into a clustering calibration engine, and bind the initial classification result and the multidimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multidimensional account features; a first clustering module, used to perform unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; and a first associating module, used to, for each cluster, if the number of accounts with the same initial category label in the cluster is greater than a preset threshold, use the initial category labels with the same initial category label greater than the preset threshold as the dominant labels of each account in the cluster.

[0018] Furthermore, the account classification device further includes: a first configuration module, configured to configure new account labels for each account within the cluster based on the center vector of the cluster when the dominant label does not exist in the cluster.

[0019] Furthermore, the account classification device also includes: a first storage module, used to associate and store account information and account tags based on the account classification results, and to establish an account profile index.

[0020] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform any of the above-described account classification methods.

[0021] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any of the above-described account classification methods.

[0022] According to another aspect of the present invention, a computer program product is also provided, the computer program product comprising a computer program, wherein the computer program, when executed by a processor, implements any of the above-described account classification methods.

[0023] In this application, the following steps are performed: account information is obtained and preprocessed to obtain account text data; the account text data is input into a natural language processing model; the natural language processing model performs semantic parsing on the account text data to extract multi-dimensional account features; based on the multi-dimensional account features, the accounts are classified using a rule-based classification engine to obtain an initial classification result; the initial classification result and the multi-dimensional account features are input into a clustering calibration engine; the clustering calibration engine uses a clustering algorithm to perform unsupervised calibration on the initial classification result to obtain the final account classification result.

[0024] In this application, preprocessing techniques are used to transform acquired account information into standard account text data, eliminating format barriers of unstructured data. The text data is then input into a natural language processing model for deep semantic analysis to extract multi-dimensional account features, understand the account's business intent, and thus improve classification accuracy. The extracted multi-dimensional account features are input into a rule-based classification engine, which performs preliminary account classification based on preset business logic. To address the potential limitations and ambiguous boundaries of rule-based classification, the initial classification results and multi-dimensional account features are jointly input into a clustering calibration engine. A clustering algorithm is used to perform unsupervised learning calibration on the initial results, improving classification accuracy. Simultaneously, simplified feature matching and AI-assisted classification result calibration improve classification efficiency. This solves the technical problems of low accuracy and efficiency in account classification that rely on manual division or empirical rule matching in related technologies. Attached Figure Description

[0025] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0026] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing an account classification method is shown.

[0027] Figure 2 This is a flowchart of an optional account classification method according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of an optional account classification process according to an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of an optional account sorting device according to an embodiment of the present invention;

[0030] Figure 5 This is a hardware structure block diagram of an electronic device (or mobile device) that performs an optional account classification method according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] It should be noted that the account classification method and apparatus in this application can be used in the fintech field for classifying accounts based on rules and clustering algorithms, and can also be used in any field other than fintech for classifying accounts based on rules and clustering algorithms. This application does not limit the application field of the account classification method and apparatus.

[0034] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0035] The following embodiments of the present invention can be applied to various account classification systems / applications / devices. The present invention proposes an account classification technology based on a two-layer intelligent architecture combining rule-based classification and cluster calibration. It uses a natural language processing model for semantic recognition, extracts key account elements, and identifies business intent. A rule-based classification engine uses preset classification logic to perform preliminary account classification, while a clustering algorithm calibrates the preliminary classification results. This significantly improves classification accuracy, adaptability, scalability, and efficiency, completely breaking through the limitations of traditional manual and single-rule classification, and achieving efficient and accurate classification of massive account files.

[0036] The present invention will now be described in detail with reference to various embodiments.

[0037] Example 1

[0038] According to an embodiment of the present invention, an embodiment of an account classification method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0039] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing an account classification method is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0040] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0041] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the account classification method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby implementing the aforementioned account classification method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0042] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

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

[0044] Under the aforementioned operating environment, this application provides the following: Figure 2 The account classification method shown is implemented by an account classification system.

[0045] Figure 2 This is a flowchart of an optional account classification method according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes the following steps:

[0046] Step S201: Obtain account information and preprocess the account information to obtain account text data.

[0047] In step S201 above, the system obtains account information through a data interface. Account information refers to various original records and documents related to financial accounts, which may include: account attribute information, account transaction behavior information and background information for account business, account relationship information, etc. Among them, account attribute information may include data such as account type, account size, and account risk level; account transaction behavior information may include data such as transaction frequency, transaction channel, and transaction amount distribution; and account relationship information may include data such as fund transfer links, account business associations, and account organizational structure.

[0048] Account data comes in various forms, including structured database text records as well as unstructured scanned documents, image-formatted electronic receipts, and images of account opening applications. The system supports batch import or real-time upload to receive these materials, ensuring coverage of the entire account file data. Because raw account data may contain noise, inconsistent formats, or unstructured features, directly inputting it into the model will affect the processing results. Therefore, the acquired account data needs to be preprocessed to obtain account text data.

[0049] By obtaining account information through the above steps, a data entry point for the system is established, ensuring that the system can receive multi-source heterogeneous data from various business systems within the financial institution, thus providing a data foundation for automated classification.

[0050] In one optional embodiment, the preprocessing steps for account information include: determining the data type of the account information; if the data type is not text, extracting text from the account information to generate data text; and performing data cleaning and word segmentation on the data text to obtain account text data.

[0051] Furthermore, the preprocessing step for account information aims to transform heterogeneous raw data into a standardized text format that can be efficiently processed by computers. The specific execution process is as follows.

[0052] The system first determines the data type of the acquired account information. Data type refers to the data storage format of the account information, mainly divided into text and non-text types. These two types can be distinguished by checking the file header information or file extension, thereby determining the subsequent processing path.

[0053] When the data type is determined to be non-text, the system performs text extraction to generate the data text. Since non-text account data contains pixel information rather than character encoding, it cannot be directly understood by natural language processing models; therefore, optical character recognition (OCR) technology is needed for conversion. The system first preprocesses the image, including grayscale conversion, noise reduction, and binarization, to enhance the contrast between text and background. Then, a text detection model is used to locate text regions in the image, and a text recognition model converts the image features of these text regions into character sequences. After post-processing correction, the recognized character sequences ultimately form continuous data text. This achieves the transition from visual images to semantic text, ensuring that unstructured data such as scanned documents and photographs can be incorporated into subsequent automated processing.

[0054] When the data type is determined to be text, or after the text extraction steps described above have generated the data text, the system performs data cleaning and word segmentation. Data cleaning aims to remove noise data from the text and improve its purity. This may include removing redundant content. Word segmentation aims to divide a continuous text sequence into the smallest semantically independent units. For Chinese account data, the system uses word segmentation tools to segment sentences into word sequences; for English data, segmentation is based on spaces or punctuation marks. The segmented results form the basic units for subsequent natural language processing (NLP) model processing. After data cleaning and word segmentation, the system obtains account text data, which serves as the direct input source for subsequent NLP model semantic analysis and feature extraction.

[0055] Step S202: Input the account text data into the natural language processing model, and perform semantic parsing on the account text data through the natural language processing model to extract multi-dimensional account features.

[0056] In step S202 above, natural language processing (NLP) techniques are used to perform deep semantic mining on the standardized data, aiming to extract structured features with business meaning from unstructured text. The system first inputs the account text data obtained in step S201 into the NLP model. This NLP model is a pre-trained language model based on an attention mechanism. Its bidirectional attention mechanism can capture long-distance dependencies and contextual relationships between words in the text, making it suitable for processing complex text semantics related to financial accounts. The NLP model performs semantic parsing on the account text data and extracts multi-dimensional account features based on the parsing results. These multi-dimensional account features comprehensively and accurately reflect the business logic behind the account text, providing high-dimensional data support for subsequent classification decisions.

[0057] In one optional embodiment, the step of semantically parsing account text data and extracting multidimensional account features through a natural language processing model includes: inputting the account text data into a natural language processing model, generating a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and extracting multidimensional account features based on the text vector representation through a feature extraction network, wherein the multidimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0058] Furthermore, the system first inputs the account text data into a natural language processing model, which includes a word embedding layer and a multi-layer bidirectional encoder. The natural language processing model first uses the word embedding layer to convert the segmented text sequence into a low-dimensional dense vector representation, i.e., word vectors, so that semantically similar words are closer together in the vector space. Subsequently, the text vectors enter the multi-layer bidirectional encoder, which captures the contextual information to the left and right of each word in the text simultaneously through a self-attention mechanism. Through multi-layer stacking, it captures long-distance dependencies from local phrases to the global sentence, thereby generating a text vector representation containing rich contextual semantic information. The text vector representation can accurately reflect the deep semantics of the account text in a specific business context.

[0059] Based on the generated text vector representation, a feature extraction network is invoked for subsequent processing. The feature extraction network is a fully connected layer, convolutional neural network layer, or attention mechanism layer appended to the natural language processing model to reduce the dimensionality of the high-dimensional text vector representation, fuse elements, or adapt it for specific tasks, thereby extracting multi-dimensional account features. The extracted multi-dimensional account features can include: account transaction features, reflecting account transaction behavior patterns; account attribute features, reflecting the account's basic identity and static attributes; and account association features, reflecting the relationship between the account and other entities (such as counterparties or affiliated companies).

[0060] Step S203: Based on multi-dimensional account features, the accounts are classified using a rule-based classification engine to obtain initial classification results.

[0061] In step S203 above, the account features after semantic parsing are initially categorized using preset business logic. This aims to combine the semantic understanding capabilities of artificial intelligence with the interpretability of traditional rules to achieve an efficient and deterministic initial classification. The multidimensional account features are input into a rule classification engine, which is a classification module based on a preset business rule base and mapping logic. The rule classification engine maintains a real-time updated set of classification rules. During processing, the rule classification engine uses semantic vector information from the multidimensional account features for rule matching, allowing the rule matching to tolerate certain expression differences and improving the robustness of the matching. After the matching operation, the rule classification engine outputs a definite classification label, which is the initial classification result.

[0062] In one optional embodiment, the step of classifying accounts using a rule classification engine includes: loading preset classification rules from a rule base using the rule classification engine, wherein each classification rule consists of a logical condition expression based on account transaction features, account attribute features, and account association features; matching each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule using the rule classification engine, and determining the account category based on the matching results.

[0063] Furthermore, the specific steps for preliminary account classification using a rule-based classification engine can include: first, loading pre-defined classification rules from the rule base using the rule-based classification engine. Each classification rule is not a simple keyword match, but rather consists of logical conditional expressions based on account transaction characteristics, account attribute characteristics, and account association characteristics. These logical conditional expressions use Boolean logic to combine feature values ​​from different dimensions to form complex judgment logic, thereby accurately characterizing specific business scenarios.

[0064] Subsequently, the system uses a rule classification engine to match each feature value in the account's multidimensional feature vector with the logical condition expression of each rule. The multidimensional feature vector is a vector output by the natural language processing model, containing numerical values ​​for various dimensions such as account transaction features, account attribute features, and account association features. The matching process is based on logical evaluation of numerical values ​​or labels, traversing each classification rule in the rule base, extracting the logical conditions defined in the rule, and extracting the corresponding feature values ​​from the current account's multidimensional feature vector. During the matching process, the rule classification engine calculates the satisfaction level of each rule. If all logical conditions of a rule are satisfied, i.e., the feature values ​​conform to the logical range or label defined by the rule, then the rule is activated. The rule classification engine uses a priority mechanism or confidence scoring mechanism to select the category corresponding to the rule with the highest priority or highest confidence score from all activated rules as the initial classification of the account. If multiple rules are satisfied simultaneously, the rule classification engine can determine the final initial classification result according to a preset conflict resolution strategy, such as weight allocation based on feature dimensions.

[0065] Step S204: Input the initial classification results and multidimensional account features into the clustering calibration engine. The clustering calibration engine uses a clustering algorithm to perform unsupervised calibration on the initial classification results to obtain the final account classification results.

[0066] In step S204 above, a clustering algorithm based on an unsupervised learning mechanism is introduced to perform secondary verification and correction on the initial classification results generated based on rules. This addresses the potential issues of ambiguous boundaries, missed classifications, or insufficient adaptability to new scenarios that may arise from single-rule classification. The aforementioned clustering calibration engine is a processing unit that integrates a clustering algorithm module. Its core function is to perform unsupervised or semi-supervised group feature analysis on account data to discover the inherent natural distribution patterns of the data and to calibrate predefined classification labels.

[0067] The calibration process mainly includes the following two aspects: First, identifying abnormal or misclassified accounts. If the multidimensional account features of an account are too far from the cluster center corresponding to its initial classification result in the feature space, or if the account forms an independent, small outlier cluster, the system will determine that the account may be misclassified or belong to a new category not covered by the rules, thereby triggering the calibration mechanism. Second, discovering new categories. If the clustering algorithm identifies significant data clusters that are not covered by existing rules, the system can dynamically label these clusters and suggest that new classification rules need to be added or that they be classified as categories to be reviewed.

[0068] After analysis and calibration using clustering algorithms, the system combines the deterministic results of rule-based classification with the latent structures discovered through clustering analysis to correct the classification labels, resulting in the final account classification. This result not only retains the efficiency and interpretability of rule-based classification but also incorporates data-driven dynamic adaptability, improving the overall accuracy and refinement of the classification.

[0069] In one optional embodiment, the step of inputting the initial classification result and multidimensional account features into a clustering calibration engine, and having the clustering calibration engine perform unsupervised calibration on the initial classification result using a clustering algorithm, includes: inputting the initial classification result and multidimensional account features into the clustering calibration engine; binding the initial classification result and multidimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multidimensional account features; performing unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; for each cluster, if the number of accounts with the same initial category label in the cluster is greater than a preset threshold, using the initial category labels with the same initial category label greater than the preset threshold as the dominant labels for each account within the cluster.

[0070] Furthermore, the system first inputs the initial classification results and multidimensional account features into the clustering calibration engine. Within the engine, a data association operation is performed, binding the initial classification results with the multidimensional account features. Specifically, for each account to be processed, the system combines the initial classification results output in step S203 with the multidimensional account features generated in step S202 in a one-to-one correspondence. After binding, the system forms a sample set with initial category labels. The sample set refers to a collection containing all account data to be calibrated, where each account sample in the sample set is a data unit containing an account feature vector generated based on the multidimensional account features and the corresponding initial classification result label.

[0071] Subsequently, an unsupervised clustering algorithm is invoked to perform clustering on the account samples. In the feature space, the clustering calibration engine groups similar account samples together based on the geometric distance or density distribution between account feature vectors. Through iterative optimization using density-based region growing, the entire sample set is divided into several non-overlapping groups, thus generating multiple clusters. Each cluster represents a set of accounts that are highly similar in the multidimensional feature space, reflecting the group characteristics of accounts in terms of business behavior, attributes, or relationships.

[0072] Finally, the cluster calibration engine performs label analysis on each generated cluster. For any given cluster, it statistically analyzes the distribution of initial classification result labels carried by all account samples within that cluster. If, within a cluster, there exists an initial classification result label (i.e., an initial category label) whose frequency or quantity within the cluster exceeds a preset threshold, the system determines that this label has significant dominance. In this case, the system uses this initial category label, which has a quantity exceeding the preset threshold, as the dominant label for all account samples within that cluster, and uses this dominant label as the classification label for all accounts within the cluster, thus obtaining the corrected classification result. Through this majority voting-based mechanism, the consistency within the cluster is used to verify and correct the initial classification result of a single account, thereby achieving the calibration of the classification result.

[0073] In one optional embodiment, after using initial category labels with a number greater than a preset threshold as the dominant labels for each account within a cluster, the method further includes: configuring new account labels for each account within a cluster based on the center vector of the cluster when no dominant label exists in the cluster.

[0074] Furthermore, after determining the dominant label, when analyzing a specific cluster, if statistical analysis shows that the frequency of occurrence of all initial category labels within that cluster does not exceed a preset threshold, the system determines that there is no dominant label in that cluster. This indicates that although the accounts within the cluster have similar characteristics, their initial classification results are scattered and contradictory, or that the cluster represents a group of accounts in a new business scenario not covered by the rule base. In this case, the system executes a label generation strategy based on the feature space.

[0075] The system first calculates the center vector of the cluster, which is the geometric center or point of mass of the account feature vectors of all account samples within the cluster in the multidimensional feature space. Then, this center vector can be input into a pre-trained label prediction model, or, based on its position in the feature space, matched to the nearest business classification prototype. The resulting new account labels deduce new classification labels based on the actual feature distribution of the current account group. These new account labels can reflect the unique business attributes or emerging account transaction patterns of the accounts in the cluster. This enhances the system's adaptability to unknown business scenarios, ensuring that all accounts receive a reasonable and accurate classification.

[0076] In one optional embodiment, after obtaining the final account classification result, the method further includes: associating and storing account information and account tags based on the account classification result, and establishing an account profile index.

[0077] Furthermore, the system establishes a persistent storage and efficient retrieval mechanism for the classified data to ensure that the classification results can be quickly accessed by the system and support subsequent query and management operations. The system first receives the final account classification results output by the clustering calibration engine, binding unstructured account information with structured account tags to ensure that the corresponding account tags can be directly retrieved when reading account information. After completing the associated storage, the system uses key fields such as account tags, account number, account name, and transaction time as index keys, and the physical storage location or logical reference of the account information as index values ​​to construct an inverted index or hash index.

[0078] Through the above steps, account information is obtained and preprocessed to obtain account text data. The account text data is then input into a natural language processing (NLP) model, which performs semantic parsing to extract multi-dimensional account features. Based on these multi-dimensional features, the accounts are classified using a rule-based classification engine to obtain initial classification results. These initial classification results and multi-dimensional account features are then input into a clustering calibration engine, which uses a clustering algorithm to perform unsupervised calibration on the initial classification results, resulting in the final account classification results.

[0079] In this embodiment, preprocessing techniques are used to convert the acquired account information into standard account text data, eliminating the format barriers of unstructured data. The text data is then input into a natural language processing model for deep semantic analysis to extract multi-dimensional account features, understand the business intent of the accounts, and thus improve classification accuracy. The extracted multi-dimensional account features are input into a rule-based classification engine, which performs preliminary classification of accounts according to preset business logic. To address the potential limitations and ambiguous boundaries of rule-based classification, the initial classification results and multi-dimensional account features are jointly input into a clustering calibration engine. A clustering algorithm is used to perform unsupervised learning calibration on the initial results, improving classification accuracy. Simultaneously, based on simplified feature matching and AI-assisted classification result calibration, classification efficiency is improved. This solves the technical problem of low accuracy and efficiency in account classification that relies on manual division or empirical rule matching in related technologies.

[0080] The following describes in detail another optional implementation method.

[0081] Figure 3 This is a schematic diagram of an optional account classification process according to an embodiment of the present invention, such as... Figure 3 As shown, the account classification process includes:

[0082] Step 1, Begin;

[0083] Step two, obtain account information;

[0084] Step 3: Determine the document type. If the document type is an image or scanned document, proceed to Step 4. If it is text, skip Step 4 and proceed to Step 5.

[0085] Step 4: Perform text recognition on non-text materials such as images or scanned documents;

[0086] Step 5, Feature Extraction;

[0087] Step 6: Rule determination, obtaining initial classification results;

[0088] Step 7: Cluster calibration to obtain the final classification results;

[0089] Step eight, end.

[0090] This invention proposes an account classification technology based on a two-layer intelligent architecture combining rule-based classification and cluster calibration. It uses a natural language processing model for semantic recognition, extracts key account elements, and identifies business intent. A rule-based classification engine performs preliminary account classification using preset classification logic, while a clustering algorithm calibrates the preliminary classification results. This significantly improves classification accuracy, adaptability, scalability, and efficiency, completely breaking through the limitations of traditional manual and single-rule classification, and achieving efficient and accurate classification of massive account files.

[0091] The following is a detailed description with reference to another embodiment.

[0092] Example 2

[0093] The account classification device provided in this embodiment includes multiple implementation units, each of which corresponds to a specific implementation step in the above embodiment one. The specific implementation method and beneficial effects can be referred to the aforementioned method embodiment, and will not be repeated here.

[0094] Figure 4 This is a schematic diagram of an optional account classification device according to an embodiment of the present invention, such as... Figure 4 As shown, the account classification device may include: an acquisition unit 41, a parsing unit 42, a classification unit 43, and a calibration unit 44, wherein,

[0095] The acquisition unit 41 is used to acquire account information and preprocess the account information to obtain account text data;

[0096] Parsing unit 42 is used to input account text data into the natural language processing model, and to perform semantic parsing on the account text data through the natural language processing model to extract multi-dimensional account features;

[0097] Classification unit 43 is used to classify accounts based on multidimensional account features using a rule-based classification engine to obtain initial classification results;

[0098] The calibration unit 44 is used to input the initial classification results and multidimensional account features into the clustering calibration engine, which then uses a clustering algorithm to perform unsupervised calibration on the initial classification results to obtain the final account classification results.

[0099] The aforementioned account classification device acquires account information through acquisition unit 41 and preprocesses the account information to obtain account text data; it then inputs the account text data into a natural language processing model through parsing unit 42, which performs semantic parsing on the account text data and extracts multi-dimensional account features; based on the multi-dimensional account features, the classification unit 43 classifies the accounts using a rule-based classification engine to obtain an initial classification result; and finally, the calibration unit 44 inputs the initial classification result and the multi-dimensional account features into a clustering calibration engine, which uses a clustering algorithm to perform unsupervised calibration on the initial classification result to obtain the final account classification result.

[0100] In this embodiment, preprocessing techniques are used to convert the acquired account information into standard account text data, eliminating the format barriers of unstructured data. The text data is then input into a natural language processing model for deep semantic analysis to extract multi-dimensional account features, understand the business intent of the accounts, and thus improve classification accuracy. The extracted multi-dimensional account features are input into a rule-based classification engine, which performs preliminary classification of accounts according to preset business logic. To address the potential limitations and ambiguous boundaries of rule-based classification, the initial classification results and multi-dimensional account features are jointly input into a clustering calibration engine. A clustering algorithm is used to perform unsupervised learning calibration on the initial results, improving classification accuracy. Simultaneously, based on simplified feature matching and AI-assisted classification result calibration, classification efficiency is improved. This solves the technical problem of low accuracy and efficiency in account classification that relies on manual division or empirical rule matching in related technologies.

[0101] Furthermore, the acquisition unit includes: a first extraction module, used to determine the data type of the account information; if the data type is not text, it extracts text from the account information to generate data text; and a first processing module, used to perform data cleaning and word segmentation on the data text to obtain account text data.

[0102] Furthermore, the parsing unit includes: a first generation module, used to input account text data into a natural language processing model, and generate a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and a second extraction module, used to extract multi-dimensional account features based on the text vector representation through a feature extraction network, wherein the multi-dimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0103] Furthermore, the classification unit includes: a first loading module, used to load preset classification rules from the rule base through a rule classification engine, wherein each classification rule consists of a logical condition expression based on account transaction features, account attribute features, and account association features; and a first matching module, used to match each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule through the rule classification engine, and determine the account category based on the matching result.

[0104] Furthermore, the calibration unit includes: a first binding module, used to input the initial classification result and multi-dimensional account features into the clustering calibration engine, and bind the initial classification result with the multi-dimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multi-dimensional account features; a first clustering module, used to perform unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; and a first assignment module, used to, for each cluster, if the number of accounts with the same initial category label in the cluster is greater than a preset threshold, use the initial category labels with the same initial category label greater than the preset threshold as the dominant labels of each account in the cluster.

[0105] Furthermore, the account classification device also includes: a first configuration module, used to configure new account labels for each account within a cluster based on the center vector of the cluster when there is no dominant label in the cluster.

[0106] Furthermore, the account classification device also includes: a first storage module, used to associate and store account information and account tags based on the account classification results, and to establish an account profile index.

[0107] It should be noted that the acquisition unit 41, parsing unit 42, classification unit 43, and calibration unit 44 mentioned above correspond to steps S201 to S204 in Embodiment 1. The instances and application scenarios implemented by the above units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules or units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0108] The invention will now be described in conjunction with another alternative embodiment.

[0109] Example 3

[0110] The present invention can also provide an electronic device. Figure 5This is a hardware structure block diagram of an electronic device (or mobile device) for implementing an optional account classification method according to an embodiment of the present invention, such as... Figure 5 As shown, the electronic device may include: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0111] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0112] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: acquire account information and preprocess the account information to obtain account text data; input the account text data into a natural language processing model, perform semantic parsing on the account text data through the natural language processing model, and extract multi-dimensional account features; classify the accounts based on the multi-dimensional account features through a rule-based classification engine to obtain an initial classification result; input the initial classification result and the multi-dimensional account features into a clustering calibration engine, and have the clustering calibration engine perform unsupervised calibration on the initial classification result using a clustering algorithm to obtain the final account classification result.

[0113] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: determine the data type of the account information; if the data type is not text, extract the text from the account information to generate data text; perform data cleaning and word segmentation on the data text to obtain account text data.

[0114] The processor can invoke the information and application stored in the memory through the transmission device to perform the following steps: inputting account text data into a natural language processing model, generating a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and extracting multi-dimensional account features based on the text vector representation through a feature extraction network, wherein the multi-dimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0115] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: load preset classification rules from the rule base through a rule classification engine, wherein each classification rule consists of a logical condition expression based on account transaction features, account attribute features, and account association features; match each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule through the rule classification engine, and determine the account category based on the matching results.

[0116] The processor can access the information and application programs stored in the memory via the transmission device to perform the following steps: inputting the initial classification results and multidimensional account features into the clustering calibration engine; binding the initial classification results with the multidimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multidimensional account features; performing unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; for each cluster, if the number of accounts with the same initial category label in the cluster exceeds a preset threshold, using the initial category labels with the same initial category label exceeding the preset threshold as the dominant labels for each account within the cluster.

[0117] The processor can access information and applications stored in memory via a transmission device to perform the following steps: in the absence of a dominant label in a cluster, configure new account labels for each account within the cluster based on the cluster's center vector.

[0118] The processor can access information and applications stored in memory via a transmission device to perform the following steps: associate and store account information and account tags based on account classification results, and establish an account profile index.

[0119] This invention provides an account classification scheme. Preprocessing techniques are used to convert acquired account information into standard account text data, eliminating format barriers associated with unstructured data. The text data is then input into a natural language processing model for deep semantic analysis, extracting multi-dimensional account features to understand the account's business intent and improve classification accuracy. These extracted multi-dimensional account features are then input into a rule-based classification engine, which performs preliminary account classification based on preset business logic. To address the limitations and ambiguous boundaries of rule-based classification, the initial classification results and multi-dimensional account features are jointly input into a clustering calibration engine. A clustering algorithm is used to perform unsupervised learning calibration on the initial results, improving classification accuracy. Simultaneously, simplified feature matching and AI-assisted classification result calibration improve classification efficiency. This solves the technical problems of low accuracy and efficiency in account classification that rely on manual division or empirical rule matching in related technologies.

[0120] Those skilled in the art will understand that Figure 5 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0121] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0122] The invention will now be described in conjunction with another alternative embodiment.

[0123] Example 4

[0124] This invention also provides a computer-readable storage medium. Optionally, in this invention, the computer-readable storage medium can be used to store the program code executed by the account classification method provided in Embodiment 1.

[0125] Optionally, in this embodiment of the invention, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0126] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the following steps of an account classification method: acquiring account information and preprocessing the account information to obtain account text data; inputting the account text data into a natural language processing model, performing semantic parsing on the account text data through the natural language processing model, and extracting multi-dimensional account features; classifying accounts based on the multi-dimensional account features through a rule-based classification engine to obtain an initial classification result; inputting the initial classification result and the multi-dimensional account features into a clustering calibration engine, which uses a clustering algorithm to perform unsupervised calibration on the initial classification result to obtain the final account classification result.

[0127] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of an account classification method: determining the data type of the account information; if the data type is not text, extracting text from the account information to generate data text; and performing data cleaning and word segmentation on the data text to obtain account text data.

[0128] This invention also provides a computer program product, which, when executed on a data processing device, is a program suitable for performing the steps of an account classification method: inputting account text data into a natural language processing model, generating a text vector representation containing contextual semantic information through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model; and extracting multi-dimensional account features based on the text vector representation through a feature extraction network, wherein the multi-dimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

[0129] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of an account classification method: loading preset classification rules from a rule base through a rule classification engine, wherein each classification rule consists of a logical condition expression based on account transaction features, account attribute features, and account association features; matching each feature value in the multidimensional feature vector of the account with the logical condition expression of each rule through the rule classification engine, and determining the account category based on the matching result.

[0130] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of an account classification method: inputting initial classification results and multidimensional account features into a clustering calibration engine, binding the initial classification results with multidimensional account features through the clustering calibration engine to form a sample set with initial category labels, wherein each account sample in the sample set contains an account feature vector generated based on the multidimensional account features; performing unsupervised clustering on the account samples using a clustering algorithm to generate multiple clusters; for each cluster, if the number of accounts with the same initial category label in the cluster is greater than a preset threshold, using the initial category labels with the same initial category label greater than the preset threshold as the dominant labels for each account within the cluster.

[0131] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of an account classification method: in the absence of a dominant label in a cluster, configuring new account labels for each account within the cluster based on the center vector of the cluster.

[0132] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for performing the steps of an account classification method: associating and storing account information and account tags based on the account classification results, and establishing an account profile index.

[0133] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0134] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0135] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0136] 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 units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0137] Furthermore, 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. The integrated unit can be implemented in hardware or as a software functional unit.

[0138] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. 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 the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0139] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An account classification method, characterized in that, include: Obtain account information and preprocess the account information to obtain account text data; The account text data is input into a natural language processing model, and the natural language processing model performs semantic parsing on the account text data to extract multi-dimensional account features; Based on the aforementioned multidimensional account features, the accounts are classified using a rule-based classification engine to obtain initial classification results; The initial classification result and the multidimensional account features are input into the clustering calibration engine, which uses a clustering algorithm to perform unsupervised calibration on the initial classification result to obtain the final account classification result.

2. The method according to claim 1, characterized in that, The steps for preprocessing the account information include: Determine the data type of the account information. If the data type is not text, extract the text from the account information to generate data text. The data text is cleaned and segmented to obtain the account text data.

3. The method according to claim 1, characterized in that, The steps of semantically parsing the account text data and extracting multidimensional account features using the natural language processing model include: The account text data is input into the natural language processing model, and a text vector representation containing contextual semantic information is generated through the word embedding layer and multi-layer bidirectional encoder of the natural language processing model. Based on the text vector representation, the multidimensional account features are extracted by a feature extraction network, wherein the multidimensional account features include at least one of the following: account transaction features, account attribute features, and account association features.

4. The method according to claim 1, characterized in that, The steps for classifying accounts using a rule-based classification engine include: The rule classification engine loads preset classification rules from the rule base, wherein each classification rule consists of a logical conditional expression based on account transaction characteristics, account attribute characteristics, and account association characteristics. The rule classification engine matches each feature value in the multidimensional feature vector of an account with the logical condition expression of each rule, and determines the account category based on the matching results.

5. The method according to claim 1, characterized in that, The steps of inputting the initial classification result and the multidimensional account features into the clustering calibration engine, and having the clustering calibration engine perform unsupervised calibration on the initial classification result using a clustering algorithm, include: The initial classification result and the multidimensional account features are input into the clustering calibration engine. The clustering calibration engine binds the initial classification result and the multidimensional account features to form a sample set with initial category labels. Each account sample in the sample set contains an account feature vector generated based on the multidimensional account features. Unsupervised clustering was performed on the account samples using a clustering algorithm to generate multiple clusters; For each cluster, if the number of initial category labels in the cluster is greater than a preset threshold, the initial category labels with the same initial category label being greater than the preset threshold shall be used as the dominant labels for each account within the cluster.

6. The method according to claim 5, characterized in that, After using initial category labels where the number of identical initial category labels exceeds a preset threshold as the dominant labels for each account within the cluster, the process further includes: If the dominant label does not exist in the cluster, a new account label is configured for each account within the cluster based on the center vector of the cluster.

7. The method according to claim 1, characterized in that, After obtaining the final account classification results, the following is also included: Based on the account classification results, account information and account tags are associated and stored, and an account profile index is established.

8. An account sorting device, characterized in that, include: The acquisition unit is used to acquire account information and preprocess the account information to obtain account text data; The parsing unit is used to input the account text data into the natural language processing model, and perform semantic parsing on the account text data through the natural language processing model to extract multi-dimensional account features. The classification unit is used to classify accounts based on the multidimensional account features using a rule-based classification engine to obtain an initial classification result; The calibration unit is used to input the initial classification result and the multidimensional account features into the clustering calibration engine, and the clustering calibration engine uses a clustering algorithm to perform unsupervised calibration on the initial classification result to obtain the final account classification result.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the account classification method according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the account classification method according to any one of claims 1 to 7.

11. A computer program product, characterized in that, The computer program product includes a computer program, wherein the computer program, when executed by a processor, implements the account classification method according to any one of claims 1 to 7.

Images