Accounting data processing system and method based on data mining

Abstract

The invention relates to the technical field of data processing, and particularly discloses an accounting data processing system and method based on data mining, and the method comprises the steps: carrying out the context semantic coding of fund flow data through a context encoder based on a converter, so as to obtain a plurality of fund flow feature vectors, and then, carrying out the segmentation of the fund flow feature vectors; fusing the plurality of fund flow feature vectors, performing local correlation coding to obtain a fund flow global semantic feature vector and a fund flow local correlation feature vector, and finally fusing the fund flow global semantic feature vector and the fund flow local correlation feature vector to obtain a fund flow classification feature vector. And a classification result for monitoring whether the fund flow data in the accounting data has potential fraudulent behaviors or not is obtained through the classifier, so that the financial safety and the operating benefits of an enterprise are protected.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and more specifically, to an accounting data processing system and method based on data mining. Background Art

[0002] Accounting is to unify various useful economic activities of an enterprise into monetary units, and provide economic information reflecting the financial status and operating results of the enterprise through a series of procedures such as bookkeeping, accounting, and reporting. It is an economic management activity that uses currency as the main unit of measurement and uses special methods to continuously, systematically, and comprehensively reflect and supervise the economic activities of enterprises, institutions, or other economic organizations. Specifically, accounting is the accounting and supervision of the economic activities of a certain subject, and provides corresponding accounting information to relevant parties.

[0003] In the existing technology, accounting data processing involves a large amount of data and information flow, and there are many links and processes. This complexity provides opportunities for fraud, because fraudsters can exploit loopholes or weaknesses in these links and processes to hide or tamper with data. With the advancement of technology, accounting data processing is increasingly dependent on computer systems and software. If there are loopholes in the information system of an enterprise, or if there are no appropriate information security measures, hackers or insiders may exploit these loopholes to tamper with data or conduct other fraudulent activities.

[0004] Therefore, an accounting data processing system and method based on data mining are desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an accounting data processing system and method based on data mining, which performs context semantic encoding on fund flow data through a converter-based context encoder to obtain multiple fund flow feature vectors, and then fuses and locally associates the multiple fund flow feature vectors to obtain fund flow global semantic feature vectors and fund flow local association feature vectors, and finally fuses the above two to obtain fund flow classification feature vectors, and obtains classification results for monitoring whether there is potential fraud in fund flow data in accounting data through a classifier, so as to protect the financial security and business interests of enterprises in this way.

[0006] According to one aspect of the present application, there is provided an accounting data processing system based on data mining, which comprises: A fund flow data collection module is used to obtain fund flow data, which includes three dimensions: fund source, fund destination and fund flow path; A fund flow data semantic encoding module, used for semantically encoding the fund flow data to obtain a plurality of fund flow feature vectors; A capital flow feature fusion module, used for fusing the multiple capital flow feature vectors to obtain a global semantic feature vector of capital flow; A local correlation coding module, used for performing local correlation coding of the plurality of capital flow feature vectors to obtain a local correlation feature vector of capital flow; A feature fusion module, used for performing feature matching fusion based on intrinsic attribute identification on the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow to obtain a classification feature vector of the capital flow; The result generation module is used to pass the fund flow classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

[0007] According to another aspect of the present application, there is provided an accounting data processing method based on data mining, which comprises: Obtaining capital flow data, the capital flow data including three dimensions: capital source, capital destination and capital flow path; semantically encoding the capital flow data to obtain a plurality of capital flow feature vectors; fusing the multiple capital flow feature vectors to obtain a global semantic feature vector of capital flow; Performing capital flow local correlation coding on the multiple capital flow feature vectors to obtain a capital flow local correlation feature vector; Performing feature matching and fusion based on intrinsic attribute identification on the global semantic feature vector of capital flow and the local correlation feature vector of capital flow to obtain a classification feature vector of capital flow; The fund flow classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

[0008] Compared with the prior art, the present application provides an accounting data processing system and method based on data mining, which performs context semantic encoding on cash flow data through a converter-based context encoder to obtain multiple cash flow feature vectors, and then fuses and locally associates the multiple cash flow feature vectors to obtain a global semantic feature vector of cash flow and a local association feature vector of cash flow. Finally, the two are fused to obtain a cash flow classification feature vector, and a classifier is used to obtain a classification result for monitoring whether there is potential fraud in the cash flow data in the accounting data. In this way, the financial security and operating interests of the enterprise are protected. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 The figure illustrates a block diagram of an accounting data processing system based on data mining according to an embodiment of the present application.

[0011] Figure 2 The diagram illustrates a block diagram of a semantic encoding module for fund flow data in an accounting data processing system based on data mining according to an embodiment of the present application.

[0012] Figure 3 The figure illustrates a flow chart of an accounting data processing method based on data mining according to an embodiment of the present application.

[0013] Figure 4 A block diagram of an electronic device according to an embodiment of the present application is illustrated. DETAILED DESCRIPTION

[0014] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0015] Figure 1 The figure shows a block diagram of an accounting data processing system based on data mining according to an embodiment of the present application. Figure 1As shown, the accounting data processing system 100 based on data mining according to the embodiment of the present application includes: a capital flow data acquisition module 110, which is used to obtain capital flow data, and the capital flow data includes three dimensions of capital source, capital destination and capital flow path; a capital flow data semantic encoding module 120, which is used to semantically encode the capital flow data to obtain multiple capital flow feature vectors; a capital flow feature fusion module 130, which is used to fuse the multiple capital flow feature vectors to obtain a global semantic feature vector of capital flow; a local association encoding module 140, which is used to perform capital flow local association encoding on the multiple capital flow feature vectors to obtain a capital flow local association feature vector; a feature fusion module 150, which is used to perform feature matching and fusion based on intrinsic attribute identification on the global semantic feature vector of capital flow and the local association feature vector of capital flow to obtain a capital flow classification feature vector; a result generation module 160, which is used to pass the capital flow classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

[0016] In an embodiment of the present application, the fund flow data acquisition module 110 is used to obtain fund flow data, and the fund flow data includes three dimensions: source of funds, destination of funds, and path of fund flow. It should be understood that fund flow data is crucial for risk management. By acquiring fund flow data and analyzing the patterns and trends therein, potential risks and abnormal situations can be discovered. For example, abnormal fund flow patterns may indicate potential fraud or misappropriation of funds. Timely identification and management of these risks can help protect the financial security and operating interests of the enterprise. It can make full use of the characteristics and laws of the data itself to adaptively determine appropriate weights, thereby comprehensively and objectively monitoring fund flow data.

[0017] In the embodiment of the present application, the fund flow data semantic encoding module 120 is used to semantically encode the fund flow data to obtain multiple fund flow feature vectors. Specifically, Figure 2 The figure shows a block diagram of a semantic coding module for capital flow data in an accounting data processing system based on data mining according to an embodiment of the present application. Figure 2 As shown, the fund flow data semantic encoding module 120 includes: an embedding conversion unit 121, which is used to pass the fund flow data through an embedding layer to convert each dimensional data in the fund flow data into an embedding vector to obtain a sequence of embedded vectors, wherein the embedding layer uses a learnable embedding matrix to embed the each dimensional data; an encoding unit 122, which is used to pass the sequence of embedded vectors through a converter-based context encoder to obtain the multiple fund flow feature vectors.

[0018] It should be understood that converting the capital flow data into an embedding vector can convert the original high-dimensional data representation into a low-dimensional vector representation. This can reduce the dimension of the data, simplify the representation and processing of the data, and improve computational efficiency. Through the transformer-based context encoder, the capital flow data can be converted into an embedding vector with semantic information. The encoder can learn the key features and patterns in the data and encode them into vector representations. These feature vectors can better capture the important information in the capital flow data and provide more representative feature representations.

[0019] Furthermore, in an embodiment of the present application, the context encoder uses a transformer-based Bert model to perform global context semantic encoding on the sequence of embedded vectors. In particular, the Bert model performs global context encoding on each embedded vector in the sequence of embedded vectors based on the intrinsic mask structure of the transformer, with the global context of the sequence of embedded vectors as the semantic background, to obtain the multiple capital flow feature vectors. Each feature vector in the multiple capital flow feature vectors corresponds to an embedded vector, that is, a high-dimensional feature expression corresponding to a time-domain statistical feature.

[0020] In an embodiment of the present application, the fund flow feature fusion module 130 is used to fuse the multiple fund flow feature vectors to obtain a global semantic feature vector of fund flow. Specifically, in an embodiment of the present application, the fund flow feature fusion module is used to: cascade the multiple fund flow feature vectors to obtain the global semantic feature vector of fund flow. It should be understood that cascading multiple feature vectors can capture the global semantics of the entire fund flow data. Each feature vector provides a local perspective, and the cascading operation integrates these local perspectives to form a more comprehensive global perspective. This can better understand and represent the overall characteristics and semantics of the fund flow data, and provide more comprehensive information for subsequent analysis and decision-making. Furthermore, cascading multiple feature vectors can merge multiple low-dimensional vectors into a higher-dimensional vector. This can reduce dimensionality loss to a certain extent and retain more information. Higher-dimensional global semantic feature vectors of fund flows can provide richer feature representations, which help to more accurately describe and distinguish different fund flow patterns and trends.

[0021] In the embodiment of the present application, the local association encoding module 140 is used to perform local association encoding of the multiple capital flow feature vectors to obtain the local association feature vector of the capital flow. Specifically, in the embodiment of the present application, the local association encoding module is used to: after the multiple capital flow feature vectors are arranged in two dimensions into a feature matrix, the first convolutional neural network model as a filter is used to obtain the local association feature vector of the capital flow. It should be understood that although the use of the converter-based context encoder can extract the global context semantic information of each capital flow data, its performance in extracting local capital flow association features is relatively weak. Therefore, by arranging multiple feature vectors into a feature matrix, the local association information of the capital flow data can be encoded. Each element in the feature matrix corresponds to a feature vector, and the arrangement of the feature matrix can reflect the spatial or temporal relationship in the capital flow data. Through the convolutional neural network model, the local association information in the feature matrix can be extracted by using the convolution operation to capture the pattern and structure in the capital flow data. More importantly, the convolutional neural network model acts as a filter, and the feature matrix can be extracted through the convolution operation. The convolution operation can extract local features on the feature matrix in the form of a sliding window. In this way, important patterns and correlation information in the feature matrix can be identified, thereby obtaining the local correlation feature vector of capital flow. That is, a convolutional neural network model with excellent performance in the field of local feature extraction is used as a feature extractor to capture the high-dimensional implicit feature representation of the local area in the feature matrix formed by the two-dimensional arrangement of the plurality of capital flow feature vectors, that is, the high-dimensional feature representation of the correlation features between different capital flow feature vectors.

[0022] In an embodiment of the present application, the feature fusion module 150 is used to: perform feature matching fusion based on intrinsic attribute identification on the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow to obtain a classification feature vector of the capital flow. It should be understood that the semantic feature vector of the capital flow and the local correlation feature vector of the capital flow respectively capture information at different levels. The global semantic feature vector of the capital flow may contain overall semantic information, and the local correlation feature vector of the capital flow may capture feature information of local correlation. In order to integrate information at different levels and make the final classification feature vector more representative and expressive, it is necessary to fuse the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow.

[0023] In particular, in the technical solution of the present application, considering that the global semantic feature vector of capital flow and the local correlation feature vector of capital flow come from different conditions, direct merging may cause information distortion or loss. This is because the two have differences in scale, semantics, time series and spatial distribution characteristics, and may contain information redundancy or conflict. Specifically, if the numerical range or scale of the two feature vectors is inconsistent, if they are directly merged without proper normalization, the impact of one feature vector on the result will be much greater than the other, which will cause the model training to favor certain features and ignore other equally important features. In addition, since the global semantic feature vector captures the overall capital flow pattern, such as the overall trend and frequency of transactions, and the local correlation feature vector pays more attention to the correlation between specific transactions, such as frequent counterparties or abnormal large transfers within a certain period of time, these two features represent different perspectives and levels of information. Direct merging without considering their semantic differences may confuse the model's understanding of this information and fail to accurately distinguish between global patterns and local anomalies. Furthermore, capital flow data often has obvious time series characteristics (i.e., trends that change over time) and spatial distribution characteristics (such as geographical distribution, network structure, etc.). The feature vectors generated under different conditions emphasize different time or space dimensions respectively. Direct merging may destroy the original time and space relationship and cause information distortion. Not only that, there may be information overlap or contradictions between the feature vectors generated under different conditions. Direct merging may lead to repeated calculations or cancel out effective information. Finally, directly merging feature vectors under different conditions without proper processing will increase the complexity of the model, making it difficult for the learning algorithm to extract effective decision boundaries from them, thereby reducing the generalization ability and prediction accuracy of the model. Based on this, in the technical solution of the present application, feature matching fusion based on intrinsic attribute identification is performed on the global semantic feature vector of capital flow and the local correlation feature vector of capital flow.

[0024] Specifically, in an embodiment of the present application, the feature fusion module includes: a vector normalization conversion unit, which is used to perform a norm-constrained normalization conversion on the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow to obtain a pre-aligned global semantic feature vector of the capital flow and a pre-aligned local correlation feature vector of the capital flow; a vector intrinsic attribute identification unit, which is used to perform intrinsic attribute identification on the pre-aligned global semantic feature vector of the capital flow and the pre-aligned local correlation feature vector of the capital flow to obtain a first principal component pre-aligned global semantic feature vector of the capital flow and a second principal component pre-aligned local correlation feature vector of the capital flow; a vector intrinsic structure matching unit, which is used to perform intrinsic structure matching and fusion on the first principal component pre-aligned global semantic feature vector of the capital flow and the second principal component pre-aligned local correlation feature vector of the capital flow to obtain the capital flow classification feature vector.

[0025] More specifically, in an embodiment of the present application, the vector normalization conversion unit is used to: calculate the square of the binary norm of the global semantic feature vector of the capital flow to obtain a global semantic feature strength value of the capital flow; calculate the natural exponential function value for each eigenvalue of the global semantic feature vector of the capital flow to obtain an indexed global semantic feature vector of the capital flow; divide each eigenvalue of the indexed global semantic feature vector of the capital flow by the sum of the indexed global semantic feature vectors of the capital flow to obtain a normalized global semantic feature vector of the capital flow; multiply each eigenvalue of the normalized global semantic feature vector of the capital flow by the global semantic feature strength value of the capital flow to obtain the pre-aligned global semantic feature vector of the capital flow.

[0026] More specifically, in an embodiment of the present application, the vector intrinsic attribute identification unit is used to: perform eigenvalue-based decomposition on the pre-aligned funds flow global semantic feature vector to obtain a first intrinsic decomposition sequence of pre-aligned funds flow global semantic feature vectors and a first diagonal matrix; cascade the first eigenvector and the second eigenvector in the sequence of the first intrinsic decomposition pre-aligned funds flow global semantic feature vector to obtain the first principal component pre-aligned funds flow global semantic feature vector.

[0027] In the embodiment of the present application, specifically, the formula implementation expression of the feature fusion module can be: the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow are processed by the following formula to obtain the capital flow classification feature vector; wherein the formula is: ; ; ; ; ; in, Represents the global semantic feature vector of capital flow, represents the local correlation eigenvector of capital flow, represents the square of the vector's two-norm, represents the natural exponential function, The first one represents the global semantic feature vector of capital flow The eigenvalues ​​at the positions, The first eigenvector representing the local correlation eigenvector of capital flow The eigenvalues ​​at the positions, represents the global semantic feature vector of pre-aligned capital flow, represents the local correlation feature vector of pre-aligned fund flows, Indicates intrinsic attribute identification, represents the sequence of global semantic feature vectors of pre-aligned capital flows of the first eigendecomposition, represents the first diagonal matrix, and Represent the first position and the first diagonal matrix respectively. The eigenvalues ​​at the positions, represents the transpose of a vector, represents vector concatenation, , , denote the first, second and third sequences of the global semantic feature vectors of the pre-aligned capital flow of the first eigendecomposition. feature vectors, represents the global semantic feature vector of capital flow pre-aligned by the first principal component, represents the sequence of local correlation eigenvectors of the second eigendecomposition pre-aligned fund flows, represents the second diagonal matrix, and Represent the first and second positions of the second diagonal matrix respectively. The eigenvalues ​​at the positions, represents vector concatenation, , , denote the first, second and third sequences of the local correlation eigenvectors of the pre-aligned capital flows of the second eigendecomposition. feature vectors, represents the local correlation eigenvector of the second principal component pre-aligned capital flow, It means subtracting by position. It means point multiplication by position. It means adding by position. represents point convolution, Represents the capital flow classification feature vector.

[0028] In the technical solution of the present application, the global semantic feature vector of the capital flow and the local associated feature vector of the capital flow are subjected to feature matching fusion based on intrinsic attribute identification to obtain a capital flow classification feature vector. First, the global semantic feature vector of the capital flow and the local associated feature vector of the capital flow are subjected to normalization transformation based on norm constraints to obtain a pre-aligned global semantic feature vector of the capital flow and a pre-aligned local associated feature vector of the capital flow. This process is not just for simple scale adjustment, but an important technology in feature engineering, which aims to solve the problem of inconsistent scales between features in a data set. By calculating the modulus of the feature vector and dividing each vector by its own modulus, it can be ensured that all feature vectors have the same length, thereby eliminating the impact of scale differences between different feature vectors. Normalization processing not only contributes to numerical stability in subsequent processing, but also makes the comparison between feature vectors more fair and reasonable, laying the foundation for subsequent feature alignment.

[0029] Next, the pre-aligned global semantic feature vector of capital flow and the pre-aligned local correlation feature vector of capital flow are subjected to intrinsic attribute identification to obtain the first principal component pre-aligned global semantic feature vector of capital flow and the second principal component pre-aligned local correlation feature vector of capital flow. That is, after the normalization modulation is completed, the pre-aligned global semantic feature vector of capital flow and the local correlation feature vector of capital flow are subjected to intrinsic attribute identification. As an unsupervised linear dimensionality reduction method, intrinsic attribute identification is widely used in data compression and feature extraction. This process determines the principal components by finding the direction of the maximum variance of the data. These principal components are new coordinate axes in the original feature space and can capture the structure of the data with minimal information loss. It is worth mentioning that by selecting the eigenvectors corresponding to the first few largest eigenvalues, the dimension of the feature space can be effectively reduced while maintaining the key information of the data. This dimensionality reduction can not only accelerate the model training process, but also help the model to generalize better and reduce the impact of noise, thereby improving the performance of the model.

[0030] Finally, the first principal component pre-aligned global semantic feature vector of capital flow and the second principal component pre-aligned local correlation feature vector of capital flow are fused by intrinsic structural matching to obtain the capital flow classification feature vector. The goal of this step is to find the best matching point or area between the two principal component feature vectors to achieve alignment between the two. Fine-grained alignment goes beyond the traditional rough fusion methods such as simple weighted average or splicing, and needs to consider the time series characteristics or spatial distribution characteristics of the feature vector. The fusion after alignment is not a simple numerical operation, but requires the design of a specific mechanism to determine how to combine the information of the two feature vectors. This content-based intelligent fusion method can more effectively utilize multi-source information and improve the representation ability and decision-making quality of the model.

[0031] In the embodiment of the present application, the result generation module 160 is used to pass the fund flow classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud. It should be understood that the classification result can be used to indicate whether there is potential fraud. According to the output of the classifier, the input data can be divided into fraud or non-fraud categories. This classification result can be used to predict potential fraud risks and help take corresponding measures to prevent and deal with fraud.

[0032] In one embodiment of the present application, the result generation module 160 includes: a fully connected encoding unit, used to use the fully connected layer of the classifier to fully connect the funds flow classification feature vector to obtain a fully connected encoded feature vector; a probability acquisition unit, used to pass the fully connected encoded feature vector through the Softmax classification function of the classifier to obtain a first probability attributable to the existence of fraud and a second probability attributable to the absence of fraud; a classification result determination unit, used to determine the classification result based on a comparison between the first probability and the second probability.

[0033] In summary, the accounting data processing system based on data mining based on the embodiment of the present application is clearly explained, which performs context semantic encoding on the cash flow data through a converter-based context encoder to obtain multiple cash flow feature vectors, and then, respectively fuses and locally associates the multiple cash flow feature vectors to obtain a cash flow global semantic feature vector and a cash flow local association feature vector. Finally, the above two are fused to obtain a cash flow classification feature vector, and a classifier is used to obtain a classification result for monitoring whether there is potential fraud in the cash flow data in the accounting data. In this way, the financial security and operating interests of the enterprise are protected.

[0034] As described above, the accounting data processing system 100 based on data mining according to the embodiment of the present application can be implemented in various terminal devices, such as a server of the accounting data processing system based on data mining. In one example, the accounting data processing system 100 based on data mining can be integrated into the terminal device as a software module and / or a hardware module. For example, the accounting data processing system 100 based on data mining can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the accounting data processing system 100 based on data mining can also be one of the many hardware modules of the terminal device.

[0035] Alternatively, in another example, the data mining-based accounting data processing system 100 and the terminal device may also be separate devices, and the data mining-based accounting data processing system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0036] Figure 3 The figure shows a flow chart of an accounting data processing method based on data mining according to an embodiment of the present application. Figure 3 As shown, the accounting data processing method based on data mining according to the embodiment of the present application includes: S110, obtaining cash flow data, the cash flow data including three dimensions of fund source, fund destination and fund flow path; S120, semantically encoding the cash flow data to obtain multiple cash flow feature vectors; S130, fusing the multiple cash flow feature vectors to obtain a global semantic feature vector of cash flow; S140, performing cash flow local association encoding on the multiple cash flow feature vectors to obtain a cash flow local association feature vector; S150, performing feature matching fusion based on intrinsic attribute identification on the global semantic feature vector of cash flow and the local association feature vector of cash flow to obtain a cash flow classification feature vector; S160, passing the cash flow classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

[0037] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned accounting data processing method based on data mining have been referred to above. Figure 1 to Figure 2 The description of the accounting data processing system based on data mining has been introduced in detail, and therefore, its repeated description will be omitted.

[0038] Below, reference Figure 4 To describe an electronic device according to an embodiment of the present application.

[0039] Figure 4 A block diagram of an electronic device according to an embodiment of the present application is illustrated.

[0040] like Figure 4 As shown, the electronic device 10 includes one or more processors 11 and a memory 12 .

[0041] The processor 11 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 10 to perform desired functions.

[0042] The memory 12 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 11 may run the program instructions to implement the accounting data processing based on data mining and / or other desired functions of the various embodiments of the present application described above. Various contents such as data of various dimensions of capital flow data may also be stored in the computer-readable storage medium.

[0043] In one example, the electronic device 10 may further include: an input device 13 and an output device 14 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0044] The input device 13 may include, for example, a keyboard, a mouse, etc.

[0045] The output device 14 can output various information to the outside, including classification results, etc. The output device 14 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, and the like.

[0046] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the accounting data processing method based on data mining according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0047] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0048] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the steps of accounting data processing based on data mining according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0049] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0050] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0051] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0052] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0053] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0054] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. An accounting data processing system based on data mining, characterized in that: include: A fund flow data collection module is used to obtain fund flow data, which includes three dimensions: fund source, fund destination and fund flow path; A fund flow data semantic encoding module, used for semantically encoding the fund flow data to obtain a plurality of fund flow feature vectors; A capital flow feature fusion module, used for fusing the multiple capital flow feature vectors to obtain a global semantic feature vector of capital flow; A local correlation coding module, used for performing local correlation coding of the plurality of capital flow feature vectors to obtain a local correlation feature vector of capital flow; A feature fusion module, used for performing feature matching fusion based on intrinsic attribute identification on the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow to obtain a classification feature vector of the capital flow; The result generation module is used to pass the fund flow classification feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

2. The accounting data processing system based on data mining according to claim 1 is characterized in that: The fund flow data semantic encoding module includes: An embedding conversion unit, configured to pass the fund flow data through an embedding layer to convert each dimension data in the fund flow data into an embedding vector to obtain a sequence of embedding vectors, wherein the embedding layer uses a learnable embedding matrix to embed the each dimension data; An encoding unit is used to pass the sequence of embedded vectors through a transformer-based context encoder to obtain the multiple capital flow feature vectors.

3. The accounting data processing system based on data mining according to claim 2 is characterized in that: The capital flow feature fusion module is used to: cascade the multiple capital flow feature vectors to obtain the capital flow global semantic feature vector.

4. The accounting data processing system based on data mining according to claim 3 is characterized in that: The local correlation encoding module is used to: arrange the multiple capital flow feature vectors in two dimensions into a feature matrix and then pass it through the first convolutional neural network model as a filter to obtain the capital flow local correlation feature vector.

5. The accounting data processing system based on data mining according to claim 4 is characterized in that: The feature fusion module comprises: A vector normalization conversion unit, used for performing a normalization conversion based on a norm constraint on the global semantic feature vector of the capital flow and the local correlation feature vector of the capital flow to obtain a pre-aligned global semantic feature vector of the capital flow and a pre-aligned local correlation feature vector of the capital flow; A vector intrinsic attribute identification unit, used for performing intrinsic attribute identification on the pre-aligned capital flow global semantic feature vector and the pre-aligned capital flow local correlation feature vector to obtain a first principal component pre-aligned capital flow global semantic feature vector and a second principal component pre-aligned capital flow local correlation feature vector; The vector intrinsic structure matching unit is used to perform intrinsic structure matching and fusion on the first principal component pre-aligned capital flow global semantic feature vector and the second principal component pre-aligned capital flow local correlation feature vector to obtain the capital flow classification feature vector.

6. The accounting data processing system based on data mining according to claim 5 is characterized in that: The vector normalization conversion unit is used for: Calculating the square of the second norm of the global semantic feature vector of the capital flow to obtain the global semantic feature strength value of the capital flow; Calculating the natural exponential function value for each eigenvalue of the global semantic feature vector of the capital flow to obtain an indexed global semantic feature vector of the capital flow; Dividing each feature value of the indexed capital flow global semantic feature vector by the sum of the indexed capital flow global semantic feature vectors to obtain a normalized capital flow global semantic feature vector; Each feature value of the normalized capital flow global semantic feature vector is multiplied by the capital flow global semantic feature strength value to obtain the pre-aligned capital flow global semantic feature vector.

7. The accounting data processing system based on data mining according to claim 6 is characterized in that: The vector intrinsic attribute identification unit is used to: Performing eigenvalue-based decomposition on the pre-aligned capital flow global semantic feature vector to obtain a sequence of first eigen-decomposed pre-aligned capital flow global semantic feature vectors and a first diagonal matrix; The first feature vector and the second feature vector in the sequence of the first intrinsic decomposition pre-aligned capital flow global semantic feature vector are cascaded to obtain the first principal component pre-aligned capital flow global semantic feature vector.

8. The accounting data processing system based on data mining according to claim 7 is characterized in that: The result generation module comprises: A fully connected encoding unit, used for performing fully connected encoding on the capital flow classification feature vector using the fully connected layer of the classifier to obtain a fully connected encoded feature vector; A probability obtaining unit, configured to pass the fully connected encoded feature vector through a Softmax classification function of the classifier to obtain a first probability attributable to the existence of fraudulent behavior and a second probability attributable to the absence of fraudulent behavior; A classification result determination unit is used to determine the classification result based on a comparison between the first probability and the second probability.

9. An accounting data processing method based on data mining, characterized in that: include: Obtaining capital flow data, the capital flow data including three dimensions: capital source, capital destination and capital flow path; semantically encoding the capital flow data to obtain a plurality of capital flow feature vectors; fusing the multiple capital flow feature vectors to obtain a global semantic feature vector of capital flow; Performing capital flow local correlation coding on the multiple capital flow feature vectors to obtain a capital flow local correlation feature vector; Performing feature matching and fusion based on intrinsic attribute identification on the global semantic feature vector of capital flow and the local correlation feature vector of capital flow to obtain a classification feature vector of capital flow; The fund flow classification feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether there is potential fraud.

10. The accounting data processing method based on data mining according to claim 9 is characterized in that: The fund flow classification feature vector is passed through a classifier to obtain a classification result, which is used to indicate whether there is potential fraud, including: Using the fully connected layer of the classifier to perform fully connected encoding on the capital flow classification feature vector to obtain a fully connected encoded feature vector; Passing the fully connected encoded feature vector through the Softmax classification function of the classifier to obtain a first probability attributable to the existence of fraudulent behavior and a second probability attributable to the absence of fraudulent behavior; The classification result is determined based on a comparison between the first probability and the second probability.

Images