CN116993489A - Method, device, equipment and medium for identifying financial fraud of marketing company - Google Patents
Method, device, equipment and medium for identifying financial fraud of marketing company Download PDFInfo
- Publication number
- CN116993489A CN116993489A CN202311023955.4A CN202311023955A CN116993489A CN 116993489 A CN116993489 A CN 116993489A CN 202311023955 A CN202311023955 A CN 202311023955A CN 116993489 A CN116993489 A CN 116993489A
- Authority
- CN
- China
- Prior art keywords
- financial
- fraud
- index data
- data
- financial index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 73
- 239000011159 matrix material Substances 0.000 claims abstract description 88
- 238000012545 processing Methods 0.000 claims abstract description 38
- 230000008569 process Effects 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims description 80
- 238000012795 verification Methods 0.000 claims description 54
- 230000007246 mechanism Effects 0.000 claims description 17
- 230000015654 memory Effects 0.000 claims description 17
- 238000004458 analytical method Methods 0.000 claims description 16
- 238000001914 filtration Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 7
- 238000010200 validation analysis Methods 0.000 claims description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 239000013598 vector Substances 0.000 description 13
- 230000002457 bidirectional effect Effects 0.000 description 9
- 230000008859 change Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 230000007787 long-term memory Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000006403 short-term memory Effects 0.000 description 3
- 230000007306 turnover Effects 0.000 description 3
- 230000007812 deficiency Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000005206 flow analysis Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Finance (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Accounting & Taxation (AREA)
- Technology Law (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a method, a device, equipment and a medium for identifying financial fraud of a marketing company, which can be applied to the field of artificial intelligence or the field of finance. In the method, the financial index data of the marketing company is obtained, and the financial index data is subjected to data processing, so that the processed financial index data can be obtained; then constructing a financial index homonymy matrix according to the processed financial index data; and finally, determining classification results of the listed companies according to the financial index comparability matrix and the financial fraud recognition model, wherein the classification results are used for indicating that the listed companies have financial fraud or the listed companies do not have the financial fraud. In the process, the classification result of whether the marketing company has the financial fraud can be obtained by utilizing the financial fraud identification model, so that the accuracy of identifying the financial fraud of the marketing company is improved.
Description
Technical Field
The application relates to the technical field of financial fraud prevention, in particular to a method, a device, equipment and a medium for identifying financial fraud of a marketing company.
Background
With the rapid development of technological economy, the current financial securities market is also vigorously developed, but the financial fraud problem of the marketing company is also caused. The marketing company maintains a good business image through accounting projects such as imaginary assets, income, profits and the like, thereby damaging the interests of investors, disturbing economic order and causing the social public to reduce the trust degree on capital markets. However, current conventional auditing methods are not effective in identifying financial fraud by a marketed company.
Disclosure of Invention
The application provides a method, a device, equipment and a medium for identifying financial fraud of a marketing company, which judge whether the marketing company has fraud or not through a fraud identification model, thereby improving the accuracy of identifying the financial fraud of the marketing company.
In a first aspect, the present application provides a method of identifying financial fraud by a marketable company, the method comprising:
acquiring financial index data of a marketing company;
data processing is carried out on the financial index data, and the processed financial index data are obtained;
constructing a financial index homonymy matrix according to the processed financial index data;
and determining classification results of the listed companies according to the financial index comparability matrix and the financial fraud recognition model, wherein the financial fraud recognition model is a model which is used for classifying the input financial index comparability matrix after training is completed.
Optionally, obtaining financial index data of a marketing company includes:
acquiring financial data of a marketing company according to the financial statement database;
acquiring index data of a marketing company according to the financial index analysis database;
and integrating the financial data and the index data to obtain the financial index data.
Optionally, performing data processing on the financial index data to obtain processed financial index data, including:
and carrying out missing value processing, index filtering and data standardization processing on the financial index data to obtain the processed financial index data.
Optionally, constructing a financial index homonymy matrix according to the processed financial index data, including:
obtaining a time sequence matrix according to the processed financial index data;
based on the time sequence matrix, constructing a financial index homonymy matrix.
Optionally, the process of obtaining the financial fraud recognition model includes:
obtaining a fraud sample and a non-fraud sample in a sample training set according to the total list of illegal information, wherein the fraud sample is financial index data of a marketing company with fraud, and the non-fraud sample is financial index data of a marketing company without fraud;
constructing a training homonymy matrix according to the sample training set;
and constructing a first model according to a plurality of training homonymy matrixes of a plurality of samples in the sample training set, wherein the first model is a two-way long-short-term memory network of a training attention mechanism.
Optionally, the sample training set further comprises a validation set, the method further comprising:
adopting a fraud sample in the verification set and a non-fraud sample in the verification set to verify the first model to obtain a verification result;
and if the verification result indicates that the accuracy of the first model is higher than the preset value, the first model is used as a financial fraud identification model to obtain a classification result of a marketing company.
In a second aspect, the present application provides an apparatus for identifying financial fraud by a marketable company, the apparatus comprising:
the acquisition unit is used for acquiring financial index data of a marketing company;
the processing unit is used for carrying out data processing on the financial index data to obtain processed financial index data;
the construction unit is used for constructing a financial index homonymy matrix according to the processed financial index data;
and the determining unit is used for determining the classification result of the marketing company according to the financial index homonymy matrix and the financial fraud recognition model, wherein the financial fraud recognition model is a model which is used for classifying the input financial index homonymy matrix after training is completed.
Optionally, the acquiring unit is specifically configured to:
acquiring financial data of a marketing company according to the financial statement database;
acquiring index data of a marketing company according to the financial index analysis database;
and integrating the financial data and the index data to obtain the financial index data.
Optionally, the processing unit is specifically configured to:
and carrying out missing value processing, index filtering and data standardization processing on the financial index data to obtain the processed financial index data.
Optionally, the construction unit is specifically configured to:
obtaining a time sequence matrix according to the processed financial index data;
based on the time sequence matrix, constructing a financial index homonymy matrix.
Optionally, the means for identifying financial fraud by a marketer further comprises:
the sample obtaining unit is used for obtaining fraud samples and non-fraud samples in the sample training set according to the total table of the violation information, wherein the fraud samples are financial index data of a marketing company with fraud, and the non-fraud samples are financial index data of a marketing company without fraud;
the construction unit is also used for constructing a training homonymy matrix according to the sample training set;
and the model building unit is used for building a first model according to a plurality of training homonymy matrixes of a plurality of samples in the sample training set, wherein the first model is a two-way long-short-term memory network of the attention mechanism after training.
Optionally, the means for identifying financial fraud by a marketer further comprises:
the verification unit is used for verifying the first model by adopting a fraud sample in the verification set and a non-fraud sample in the verification set to obtain a verification result;
and the judging unit is used for taking the first model as the financial fraud identification model to obtain the classification result of the marketing company if the verification result indicates that the accuracy of the first model is higher than a preset value.
In a third aspect, an embodiment of the present application further provides an electronic device, including a memory and a processor:
the memory is used for storing a computer program;
the processor is configured to perform the method provided in the first aspect above according to a computer program.
In a fourth aspect, embodiments of the present application further provide a computer readable storage medium storing a computer program for executing the method provided in the first aspect.
From this, the application has the following beneficial effects:
the application provides a method, a device, equipment and a medium for identifying financial fraud of a marketing company, which are characterized in that firstly, financial index data of the marketing company are obtained, the financial index data are subjected to data processing, and the processed financial index data can be obtained; then constructing a financial index homonymy matrix according to the processed financial index data; and finally, determining classification results of the marketing companies according to the financial index homonymy matrix and the financial fraud recognition model, wherein the classification results of the marketing companies are used for indicating that the marketing companies have financial fraud or the marketing companies do not have the financial fraud. In the process, the financial index homonymy matrix is obtained through the financial index data of the marketing company, and is used as the input data of the financial fraud recognition model, so that the classification result of whether the financial fraud exists in the marketing company can be obtained, and the accuracy of recognizing the financial fraud existing in the marketing company is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to these drawings for those of ordinary skill in the art.
FIG. 1 is a flow chart of a method for identifying financial fraud by a marketable company according to an embodiment of the present application;
FIG. 2 is a schematic diagram of a financial fraud recognition model in a method for recognizing financial fraud of a listed company according to an embodiment of the present application;
FIG. 3 is a flow chart of an embodiment of a method for identifying financial fraud by a marketer in accordance with the present application;
FIG. 4 is a schematic diagram of an apparatus for identifying financial fraud by a marketer in accordance with an embodiment of the present application;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.
The method, the device, the equipment and the medium for identifying financial fraud of the marketing company can be used in the artificial intelligence field or the financial field. The foregoing is merely exemplary, and is not intended to limit the application fields of the method, apparatus, device, and medium for identifying financial fraud by a marketable company.
At present, a traditional auditing method is used for judging whether a marketing company has financial fraud or not. However, traditional auditing methods rely primarily on inspection, observation, interrogation, function verification, recalculation, re-execution, and analysis procedures, with only a small portion of the financial fraud being found, and a large portion of the financial fraud being found by internal staff and external partners reporting.
In the embodiment of the application, firstly, the financial index data of the marketing company is obtained, then the financial index homonymy matrix is constructed according to the financial index data, and finally, the financial index homonymy matrix is input into the fraud recognition model to determine the classification result of the marketing company. The classification result of the marketing company is used for indicating that the marketing company has financial fraud or the marketing company does not have financial fraud. In particular implementations, the method may include, for example: firstly, acquiring financial index data of a marketing company through a financial statement database and a financial index analysis database; and then carrying out data processing on the financial index data, after obtaining the processed financial index data, constructing a financial index homonymy matrix according to the processed financial index data, and finally determining a classification result of a marketing company according to the financial index homonymy matrix and a financial fraud recognition model, wherein the financial fraud recognition model is a model which is used for classifying the input financial index homonymy matrix after training is completed.
Therefore, the method provided by the implementation of the application can determine the classification result of the marketing company according to the financial index data and the financial fraud identification model of the marketing company, and the classification result is used for indicating that the marketing company has financial fraud or the marketing company does not have financial fraud, so that the accuracy of identifying the financial fraud of the marketing company is improved.
In order to facilitate understanding of a specific implementation of a method for identifying financial fraud of a marketing company according to an embodiment of the present application, the following description will be made with reference to the accompanying drawings.
It should be noted that, the main body implementing the method for identifying financial fraud of a listed company may be a device for identifying financial fraud of a listed company provided by the embodiment of the present application, and the device for identifying financial fraud of a listed company may be carried in an electronic device or a functional module of an electronic device. The electronic device in the embodiment of the present application may be any device capable of implementing a method for identifying financial fraud of a marketing company in the embodiment of the present application, for example, may be an internet of things (Internet of Things, ioT) device.
Fig. 1 is a flowchart of a method for identifying financial fraud of a marketing company according to an embodiment of the present application. The method may also be applied to a means for identifying financial fraud of a listed company, such as a means 400 for identifying financial fraud of a listed company as shown in fig. 4, or the means for identifying financial fraud of a listed company may be a functional module integrated in the electronic device 500 as shown in fig. 5.
As shown in fig. 1, the method includes the following S101 to S104:
s101: and acquiring financial index data of a marketing company.
In order to obtain the classification result of the marketing company, firstly, the financial index data of the marketing company is required to be obtained, then the financial index data is subjected to data processing, the processed financial index data can be obtained, a financial index homonymy matrix is constructed according to the processed financial index data, and finally, the financial index homonymy matrix is input into a financial fraud recognition model, so that the classification result of the marketing company can be determined. Therefore, the embodiment of the application provides a precondition for obtaining the classification result of the listed company by acquiring the financial index data of the listed company through S101.
As one example, S101 may include: s1011, acquiring financial data of a marketing company according to a financial statement database, wherein the financial data of the marketing company specifically comprises: financial data of the liability statement, the profit statement and the cash flow statement in the financial statement database; s1012, acquiring index data of a marketing company according to the financial index analysis database, wherein the index data of the marketing company specifically comprises index data in repayment capacity, development capacity, risk level, operation capacity, cash flow analysis and profitability in the financial index analysis database; s1013, integrating the financial data and index data to obtain financial index data; the specific integration process comprises the following steps: and after integrating the financial data and the index data, eliminating three indexes of stock codes, accounting practices and report types and corresponding data, thereby obtaining the financial index data.
In the process, financial data and index data of a marketing company can be obtained through the financial statement database and the financial index analysis database, the financial data and the index data are integrated, irrelevant data are removed, and therefore accuracy of a financial fraud recognition model can be improved.
S102: and carrying out data processing on the financial index data to obtain processed financial index data.
As one example, S102 may include: and carrying out missing value processing, index filtering and data standardization processing on the financial index data to obtain the processed financial index data. Wherein, the process of carrying out the missing value processing on the financial index data is as follows: and checking the proportion of the deficiency value of the financial index, and removing the index with the deficiency value proportion being more than 20% from the financial index data. If the remaining index has a missing value, 0 is used as a replacement. After the finance index data is subjected to missing value processing, three to four indexes such as an inventory turnover rate A, an inventory turnover rate B, an inventory turnover rate C and the like can be derived from the same finance index due to different finance index calculation methods and different selected time nodes. For this case, index filtering may be performed, wherein the process of index filtering is as follows: the Euclidean distance is used for measuring the distance between the indexes, and one index is deleted randomly under the condition that the Euclidean distance of the numerical values of the two indexes is smaller than a certain threshold value. After the repeated identical financial indexes are filtered, as different orders of magnitude and dimensions exist among the financial indexes, in order to avoid the influence of the orders of magnitude on the effect of the subsequent financial fraud recognition model, the financial index data needs to be standardized, so that the processed financial index data is obtained.
In the process, in order to avoid the influence of unprocessed financial index data on the effect of a subsequent financial fraud recognition model, missing value processing, index filtering and data standardization processing are required to be carried out on the financial index data, so that the accuracy of the financial fraud recognition model can be improved.
S103: and constructing a financial index homonymy matrix according to the processed financial index data.
As one example, S103 may include: according to the processed financial index data, a time sequence matrix can be obtained, and then the calculation method of the same-ratio increase/decrease rate is used for the time sequence matrix, so that the financial index same-ratio matrix can be obtained. Wherein, the elements in the financial index homonymy matrix represent the homonymy change rate of the financial index data between two adjacent years, and if the element values are positive numbers, the element values represent the growth rate; if the element value is negative, the drop rate is represented; if the element value is 0, no change or information missing is indicated.
It should be noted that the timing matrix may be t= [ T ] 1 ,t 2 ,…,t n ] T Representing that the elements in the time matrix are arranged in time order, i.e. t i-1 Is t i Financial index data from the previous year. Wherein, the ith row in the time sequence matrixIs the i-th financial index data set of the marketing company, and the inner element represents one kind of financial index data.
It should be noted that the financial index homography matrix may adopt s= [ S ] 1 ,s 2 ,…,s n ] T Representing the ith row in the financial index homography matrixIs the same ratio change rate of the financial index data between two adjacent years of the marketing company. But when i=1, the change rate calculation result is erroneous because the history data of the last year has no comparable base period data, and the present embodiment uses 0 instead. Thus, s 1 Is an m-dimensional all-zero vector, has no practical meaning, and the final financial index homonymy matrix is S= [ S ] 2 ,…s n ] T . Wherein the elements in the financial index homography matrix can be obtained by the following formula (1), for example, formula (1) is as follows:
wherein t is ij Jth financial index data, t, representing the ith year of the company on the market (i-1)j Jth financial index data, s, representing the ith-1 st year of the company on the market ij Represents the rate of change of the same ratio of the jth financial index data between the ith year and the ith-1 th year of the marketing company.
In this process, the timing matrix includes both the timing information of the financial index data and the rate of change of the current financial index data in the same year as the previous years, so that the "signal" of the financial fraud can be mined from the historical financial index data by using the timing matrix. The obtained financial index homonymy matrix can dynamically describe the financial condition of the company, and then the fluctuation process of the financial index data is obtained according to time sequencing, wherein the fluctuation process comprises time sequence and variation amplitude of the financial index data, the financial condition of the company is described in a multi-dimensional manner, and the two-dimensional matrix can use a more complex deep learning model.
S104: and determining classification results of the listed companies according to the financial index comparability matrix and the financial fraud recognition model, wherein the financial fraud recognition model is a model which is used for classifying the input financial index comparability matrix after training is completed.
As one example, S104 may include: s1041, in an input layer of the financial fraud recognition model, inputting each row of elements of the financial index homonymy matrix into a bidirectional node of the financial fraud recognition model according to a time sequence; s1042, in a two-way long and short Term Memory network (Bidirectional Long Short-Term Memory, biLSTM) layer of the financial fraud recognition model, calculating a financial index homonymy matrix through two-way nodes to obtain a two-way hidden state of each time node; s1043, in the attention layer of the financial fraud recognition model, splicing the bidirectional hidden states on each time node into hidden vectors, and inputting the hidden vectors into the attention mechanism of the financial fraud recognition model to obtain an output result; s1044, inputting the output result into a full connection layer of the financial fraud recognition model in the output layer of the financial fraud recognition model, and obtaining a classification result of a marketing company, wherein the classification result is 0, which indicates that no financial fraud exists; a classification result of 1 indicates the presence of financial fraud.
It should be noted that, in S1043, the bidirectional hidden states on each node may be spliced into a hidden vector by the following formula (2), where the formula (2) is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,and->Representing the bidirectional hidden state of the t-th time node, h t A hidden vector representing the t-th time node.
And the calculation process of the attention mechanism employed in S1043 includes: the similarity weight may first be obtained by using the hidden vector and the attention scoring function by the following equation (3), for example, equation (3) as follows:
u t ＝v T tanh(Wh t ) (3)
wherein W and v T Represents a weight matrix, h t Hidden vector representing the t-th time node, tanh () representing the activation function, u t Representing the similarity weight of the t-th time node.
The similarity weights are then normalized using a softmax function by the following equation (4), for example:
wherein u is t Similarity weight representing the t-th time node, u i Phases representing the ith time nodeSimilarity weight, T represents the number of time nodes of the financial fraud recognition model, a t Representing the normalized similarity weight of the t-th time node.
Finally, the obtained normalized similarity weight and the corresponding hidden vector can be weighted and summed through the following formula (5), and an output result is obtained, wherein the formula (5) is as follows:
wherein a is t Normalized similarity weight, h, representing the t-th time node t And (3) a hidden vector representing a T-th time node, wherein T represents the number of time nodes of the financial fraud identification model.
The process of obtaining the classification result of the listed companies by using the financial fraud recognition model can also refer to the schematic diagram of the financial fraud recognition model shown in fig. 2, wherein the model is divided into four layers: the input layer inputs the financial index homonymy matrix to the bidirectional nodes of the BiLSTM model according to time sequence, and it is to be noted that the financial index homonymy matrix is transposed in the model schematic diagram for convenience; when the financial index homonymy matrix enters the BiLSTM layer, the hidden state of each time node is obtained through bidirectional node operation; splicing the bidirectional hidden states on each time node into a hidden vector at the attention layer, and then inputting the hidden vector into an attention mechanism to obtain an output result; and inputting the output result into the full-connection layer at the output layer to obtain a classification result.
In the process, the financial index homonymy matrix can be input into the BiLSTM layer, and the bi-directional information flow is integrated to identify financial fraud by utilizing the bi-directional result of the BiLSTM. The attention weight of each time node can be calculated, the important time node information can be given higher weight, and the unimportant time node information can be given lower weight, so that the purposes of retaining important information and reducing irrelevant information are achieved, and the accuracy of classification results is improved.
In order to construct the financial fraud recognition model, it is first necessary to obtain a method of completing the first model of training: firstly, dividing a marketing company into a sample training set and a verification set according to a total list of illegal information, wherein the sample training set and the verification set both comprise fraud samples and non-fraud samples, the fraud samples are financial index data of the marketing company with fraud, and the non-fraud samples are financial index data of the marketing company without fraud; then constructing a training homonymy matrix according to the sample training set; and finally, constructing a first model according to a plurality of training homonymy matrixes of a plurality of samples in the sample training set, wherein the first model is a two-way long-short-term memory network of the attention mechanism after training.
Wherein, the process of obtaining the fraud sample and the non-fraud sample includes: according to the total list of the violation information, using a marketing company with the violation type conforming to the fraud type as a fraud company, and acquiring a corresponding fraud sample based on the fraud company, wherein the fraud sample comprises the following components: fraud company name and date fraud occurred, the fraud types including: one or more of fictional profit, fictional column asset, false documentations (misleading statements), significant omission, overt (others) and fraudulent marketing; according to the total list of the illegal information, using a marketed company with the illegal type not conforming to the fraud type as a non-fraud company, and selecting a non-fraud sample corresponding to the non-fraud company according to the quantity of fraud samples of 1:1, wherein the non-fraud sample comprises: a non-fraudulent company name and a date when the violation occurred.
It should be noted that, the process of constructing the training homonymy matrix according to the sample training set is the same as the process of S101 to S103 in the embodiment of the present application.
Then adopting the verification set to verify the first model, and obtaining the financial fraud identification model comprises the following steps: adopting a fraud sample in the verification set and a non-fraud sample in the verification set to verify the first model to obtain a verification result; and if the verification result indicates that the accuracy of the first model is higher than the preset value, the first model is used as a financial fraud identification model, and the financial fraud identification model is used for obtaining the classification result of the marketing company. In the process, the first model is verified through the verification set, the accuracy of the first model can be obtained through the verification result, if the accuracy of the first model is higher than a preset value, the first model is trained, and the first model can be used as a financial fraud recognition model, so that the classification accuracy of the financial fraud recognition model is improved.
Therefore, according to the embodiment of the application, the financial index data of the marketing company is obtained according to the financial statement database and the financial index analysis database; then, carrying out data processing on the financial index data to obtain processed financial index data, and constructing a financial index homonymy matrix according to the processed financial index data; and finally, inputting the financial index comparability matrix into a financial fraud recognition model to determine the classification result of the marketing company. Wherein the classification of the listed company indicates that the listed company has financial fraud or that the listed company does not have financial fraud. In the process, the classification result of the listed companies can be obtained by utilizing the financial index data and the financial fraud identification model of the listed companies, so that the accuracy of identifying the financial fraud existing in the listed companies is improved.
In order to make the method provided by the embodiment of the present application clearer and easier to understand, a specific example of a method for constructing a model for identifying financial fraud will be described with reference to fig. 3.
As shown in fig. 3, this embodiment may include S301 to S310:
s301: dividing a marketing company into a sample training set and a verification set according to a total table of illegal information, wherein the sample training set and the verification set both comprise fraud samples and non-fraud samples, the fraud samples are financial index data of the marketing company with fraud, and the non-fraud samples are financial index data of the marketing company without fraud.
As one example, S301 may include: firstly, part of marketing companies are selected as sample training sets in the violation information summary list, and then part of marketing companies are selected as verification sets, wherein the number of the marketing companies in the sample training sets and the verification sets can be the same or different, and the method is not limited in this regard. And selecting a fraud sample and a non-fraud sample according to whether the violation types of the marketing companies accord with the fraud types, wherein the fraud types comprise: one or more of fictional profit, fictional column asset, false documentations (misleading statements), significant omission, overt (others) and fraudulent marketing. Taking a marketing company with the violation type conforming to the fraud type as a fraud company, and acquiring corresponding fraud samples based on the fraud company, wherein the fraud samples comprise: fraud company name and date fraud occurred. Taking a marketing company with a violation type which does not accord with a fraud type as a non-fraud company, and selecting a non-fraud sample corresponding to the non-fraud company, wherein the non-fraud sample comprises: a non-fraudulent company name and a date when the violation occurred.
S302: and acquiring training index data in the sample training set according to the financial statement database and the financial index analysis database.
As one example, S302 may include: acquiring financial data of a marketing company in a sample training set according to the financial report database; then according to the financial index analysis database, index data of a sample training set marketing company is obtained; and finally integrating the financial data of the sample training set marketing company with the index data of the sample training set marketing company to obtain training index data of the sample training set.
S303: and carrying out data processing on the training index data to obtain the processed training index data.
As one example, S303 may include: and carrying out missing value processing, index filtering and data standardization processing on the training index data to obtain the processed training index data.
S304: and constructing a training homonymy matrix according to the processed training index data.
As one example, S304 may include: according to the processed training index data, a corresponding training time sequence matrix is obtained, and then the training time sequence matrix is utilized to use a calculation method of the same-ratio increasing/decreasing rate, so that a training same-ratio matrix can be obtained.
S305: the training homonymy matrix is input into a two-way long-short-term memory network of the attention mechanism for training, so that a model 1 is obtained, wherein the model 1 is the two-way long-term memory network of the attention mechanism after training.
As one example, S305 may include: inputting training homonymy matrix into two-way nodes of a two-way long-short-term memory network of an attention mechanism according to time sequence at an input layer; performing bidirectional node operation on the training homonymy matrix to obtain the hidden state of each time node; splicing the two-way hidden states on each time node into a hidden vector, and then inputting the hidden vector into an attention mechanism in a two-way long-short-term memory network of the attention mechanism to obtain an output result; the output layer in the two-way long and short term memory network of the attention mechanism inputs the output result into the fully connected layer, resulting in a classification result, thus obtaining model 1, wherein model 1 may be, for example, the first model in the above method.
S306: and acquiring verification index data in the verification set according to the financial statement database and the financial index analysis database.
S307: and carrying out data processing on the verification index data to obtain the processed verification index data.
S308: and constructing a verification homonymy matrix according to the processed verification index data.
S309: and inputting the verification homonymy matrix into the model 1 for verification to obtain a verification result.
S310: and if the verification result indicates that the accuracy rate of the model 1 is higher than the preset value, taking the model 1 as a financial fraud recognition model, wherein the financial fraud recognition model is used for obtaining the classification result of the marketing company.
If the verification result indicates that the accurate value of the model 1 is higher than the preset value, which means that the model 1 can be put into use, the model 1 can be used as a financial fraud recognition model.
The embodiment provides a method for obtaining a financial fraud recognition model, which divides a marketing company into a sample training set and a verification set according to a violation information summary table, and obtains training index data of the sample training set and verification index data of the verification set according to a financial statement database and a financial index analysis database. And respectively carrying out data processing on the training index data and the verification index data to obtain processed training index data and processed verification index data. And respectively constructing a corresponding training homonymy matrix and a corresponding verification homonymy matrix according to the processed training index data and the processed verification index data. The training homonymy matrix is input into a two-way long-short term memory network of the attention mechanism for training so as to obtain a model 1. And inputting the verification homonymy matrix into the model 1 for verification to obtain a verification result. And if the verification result indicates that the accuracy of the model 1 is higher than the preset value, taking the model 1 as a financial fraud recognition model. In the process, the training homonymy matrix is input into a two-way long-short-term memory network of an attention mechanism for training, so that a model 1 is obtained, and the two-way long-term memory network of the attention mechanism can integrate two-way information flow to identify financial fraud. The aim of keeping important information and reducing irrelevant information can be achieved by calculating the attention weight of each time node, so that the accuracy of the classification result is improved.
Referring to fig. 4, an embodiment of the present application provides an apparatus 400 for identifying financial fraud by a marketable company, the apparatus comprising:
an acquisition unit 401 for acquiring financial index data of a listed company;
a processing unit 402, configured to perform data processing on the financial index data, and obtain processed financial index data;
a construction unit 403, configured to construct a financial index homography matrix according to the processed financial index data;
and the determining unit 404 is configured to determine a classification result of the listed company according to the financial index homonymy matrix and a financial fraud recognition model, where the financial fraud recognition model is a model that is used for classifying the input financial index homonymy matrix after training is completed.
Alternatively, the obtaining unit 401 is specifically configured to:
acquiring financial data of a marketing company according to the financial statement database;
acquiring index data of a marketing company according to the financial index analysis database;
and integrating the financial data and the index data to obtain the financial index data.
Optionally, the processing unit 402 is specifically configured to:
and carrying out missing value processing, index filtering and data standardization processing on the financial index data to obtain the processed financial index data.
Optionally, the construction unit 403 is specifically configured to:
obtaining a time sequence matrix according to the processed financial index data;
based on the time sequence matrix, constructing a financial index homonymy matrix.
Optionally, the apparatus 400 for identifying financial fraud of a marketing company further comprises:
the sample obtaining unit is used for obtaining fraud samples and non-fraud samples in the sample training set according to the total table of the violation information, wherein the fraud samples are financial index data of a marketing company with fraud, and the non-fraud samples are financial index data of a marketing company without fraud;
the construction unit 403 is further configured to construct a training homonymy matrix according to the sample training set;
and the model building unit is used for building a first model according to a plurality of training homonymy matrixes of a plurality of samples in the sample training set, wherein the first model is a two-way long-short-term memory network of the attention mechanism after training.
Optionally, the apparatus 400 for identifying financial fraud of a marketing company further comprises:
the verification unit is used for verifying the first model by adopting a fraud sample in the verification set and a non-fraud sample in the verification set to obtain a verification result;
and the judging unit is used for taking the first model as the financial fraud identification model to obtain the classification result of the marketing company if the verification result indicates that the accuracy of the first model is higher than a preset value.
It should be noted that, the specific implementation manner and the achieved effect of the apparatus 400 for identifying financial fraud of a marketing company may be referred to the related description in the method provided in fig. 1 or fig. 3, and will not be repeated here.
The embodiment of the application further provides an electronic device 500, as shown in fig. 5, the device 500 includes a memory 501 and a processor 502:
the memory 501 is used for storing a computer program;
the processor 502 is configured to perform the methods provided in fig. 1 or 3 described above in accordance with a computer program.
Furthermore, the present application provides a computer readable storage medium for storing a computer program for executing the method provided in fig. 1 or fig. 3.
From the above description of embodiments, it will be apparent to those skilled in the art that all or part of the steps of the above described example methods may be implemented in software plus general hardware platforms. Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a storage medium, such as a read-only memory (ROM)/RAM, a magnetic disk, an optical disk, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network communication device such as a router) to perform the method according to the embodiments or some parts of the embodiments of the present application.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the device embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and reference is made to the description of the method embodiments for relevant points. The apparatus embodiments described above are merely illustrative, in which the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, i.e., may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the objective of the embodiment. Those of ordinary skill in the art will understand and implement the present application without undue burden.
The foregoing description of the exemplary embodiments of the application is merely illustrative of the application and is not intended to limit the scope of the application.
Claims (10)
1. A method of identifying financial fraud by a listing company, comprising:
acquiring financial index data of a marketing company;
performing data processing on the financial index data to obtain processed financial index data;
constructing a financial index homonymy matrix according to the processed financial index data;
and determining the classification result of the marketing company according to the financial index comparability matrix and the financial fraud recognition model, wherein the financial fraud recognition model is a model which is used for classifying the input financial index comparability matrix after training is completed.
2. The method of claim 1, wherein the obtaining financial index data for a listing company comprises:
acquiring financial data of the marketing company according to a financial statement database;
acquiring index data of the marketing company according to a financial index analysis database;
and integrating the financial data and the index data to obtain the financial index data.
3. The method of claim 1, wherein the data processing the financial index data to obtain processed financial index data comprises:
and carrying out missing value processing, index filtering and data standardization processing on the financial index data to obtain the processed financial index data.
4. The method of claim 1, wherein constructing a financial index homography matrix from the processed financial index data comprises:
obtaining a time sequence matrix according to the processed financial index data;
and constructing the financial index homonymy matrix based on the time sequence matrix.
5. The method according to claim 1, wherein the process of obtaining the financial fraud recognition model comprises:
obtaining a fraud sample and a non-fraud sample in a sample training set according to the total list of illegal information, wherein the fraud sample is financial index data of a marketing company with fraud, and the non-fraud sample is financial index data of a marketing company without fraud;
constructing a training homonymy matrix according to the sample training set;
and constructing a first model according to a plurality of training homonymy matrixes of a plurality of samples in the sample training set, wherein the first model is a two-way long-short-term memory network of a training attention mechanism.
6. The method of claim 5, wherein the sample training set further comprises a validation set, the method further comprising:
adopting the fraud sample in the verification set and the non-fraud sample in the verification set to verify the first model to obtain a verification result;
and if the verification result indicates that the accuracy of the first model is higher than a preset value, using the first model as the financial fraud identification model to obtain the classification result of the marketing company.
7. An apparatus for identifying financial fraud by a listing company, the apparatus comprising:
the acquisition unit is used for acquiring financial index data of a marketing company;
the processing unit is used for carrying out data processing on the financial index data to obtain processed financial index data;
the construction unit is used for constructing a financial index homonymy matrix according to the processed financial index data;
and the determining unit is used for determining the classification result of the marketing company according to the financial index homonymy matrix and the financial fraud recognition model, wherein the financial fraud recognition model is a model which is trained and used for classifying the input financial index homonymy matrix.
8. The method according to claim 7, wherein the acquisition unit is specifically configured to:
acquiring financial data of the marketing company according to a financial statement database;
acquiring index data of the marketing company according to a financial index analysis database;
and integrating the financial data and the index data to obtain the financial index data.
9. An electronic device comprising a memory and a processor for executing a program stored in the memory, running the method of any one of claims 1-6.
10. A computer readable storage medium, characterized in that the computer readable storage medium is for storing a computer program for executing the method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311023955.4A CN116993489A (en) | 2023-08-15 | 2023-08-15 | Method, device, equipment and medium for identifying financial fraud of marketing company |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311023955.4A CN116993489A (en) | 2023-08-15 | 2023-08-15 | Method, device, equipment and medium for identifying financial fraud of marketing company |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116993489A true CN116993489A (en) | 2023-11-03 |
Family
ID=88521231
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311023955.4A Pending CN116993489A (en) | 2023-08-15 | 2023-08-15 | Method, device, equipment and medium for identifying financial fraud of marketing company |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116993489A (en) |
-
2023
- 2023-08-15 CN CN202311023955.4A patent/CN116993489A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20180260891A1 (en) | Systems and methods for generating and using optimized ensemble models | |
| CN112017040B (en) | Credit scoring model training method, scoring system, equipment and medium | |
| Sauerwald et al. | Proxy advisors and shareholder dissent: A cross-country comparative study | |
| CN112102073A (en) | Credit risk control method and system, electronic device and readable storage medium | |
| Lokanan | Predicting money laundering using machine learning and artificial neural networks algorithms in banks | |
| CN111275338A (en) | Method, device, equipment and storage medium for judging enterprise fraud behaviors | |
| CN110930218A (en) | Method and device for identifying fraudulent customer and electronic equipment | |
| CN110782349A (en) | Model training method and system | |
| Khatun et al. | Earnings manipulation behavior in the banking industry of Bangladesh: the strategical implication of Beneish M-score model | |
| Suryani et al. | Fraud triangle perspective: Artificial neural network used in fraud analysis | |
| CN111709844A (en) | Insurance money laundering personnel detection method and device and computer readable storage medium | |
| CN112990989B (en) | Value prediction model input data generation method, device, equipment and medium | |
| CN114331735A (en) | Financial risk model training method and device and financial risk prediction method and device | |
| Marais et al. | Predicting financial statement manipulation in South Africa: A comparison of the Beneish and Dechow models | |
| CN112464051A (en) | Abnormal data detection method and device and computer readable storage medium | |
| CN110766547A (en) | Method, device, equipment and storage medium for determining credibility grade | |
| CN116629998A (en) | Automatic information counting method and device, electronic equipment and readable storage medium | |
| CN110389963A (en) | The recognition methods of channel effect, device, equipment and storage medium based on big data | |
| CN116189215A (en) | Automatic auditing method and device, electronic equipment and storage medium | |
| CN116993489A (en) | Method, device, equipment and medium for identifying financial fraud of marketing company | |
| CN114581209A (en) | Method, device and equipment for training financial analysis model and storage medium | |
| CN113240513A (en) | Method for determining user credit line and related device | |
| Terzi et al. | Comparison of financial distress prediction models: Evidence from turkey | |
| Njeru | Detection of Fraudulent Vehicle Insurance Claims Using Machine Learning | |
| Kusaya et al. | Insider abuse and fraud prediction for us banks: A comparison of machine learning approaches |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |