CN106599200A - Intelligent remote financial data anti-cheating system and anti-cheating method thereof - Google Patents
Intelligent remote financial data anti-cheating system and anti-cheating method thereof Download PDFInfo
- Publication number
- CN106599200A CN106599200A CN201611155908.5A CN201611155908A CN106599200A CN 106599200 A CN106599200 A CN 106599200A CN 201611155908 A CN201611155908 A CN 201611155908A CN 106599200 A CN106599200 A CN 106599200A
- Authority
- CN
- China
- Prior art keywords
- data
- unit
- financial
- cheating
- factor
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/25—Integrating or interfacing systems involving database management systems
- G06F16/258—Data format conversion from or to a database
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/285—Clustering or classification
-
- 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
- G06Q40/12—Accounting
- G06Q40/125—Finance or payroll
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Technology Law (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Quality & Reliability (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention discloses an intelligent remote financial data anti-cheating system. The intelligent remote financial data anti-cheating system comprises a data integration module, a database module and a processing analysis module; the data integration module is in communication connection with financial systems of various constituent companies; the financial system comprises a data storage unit; the data integration module comprises a data extraction unit, a data cleaning unit, a data conversion unit and a data loading unit; the data loading unit is in communication connection with the database module; the processing analysis module comprises an obtaining unit, an orthogonal factor analysis unit and an identification unit; one end of the obtaining unit is connected to the database module; the other end of the obtaining unit is connected to the orthogonal factor analysis unit; and the output end of the orthogonal factor analysis unit is connected to the identification unit. The invention further discloses an anti-cheating method of the intelligent remote financial data anti-cheating system. By means of the intelligent remote financial data anti-cheating system, the integration degree of the financial systems can be improved; and furthermore, remote financial monitoring of a group enterprise on various subsidiary companies can be realized.
Description
Technical field
The present invention relates to financial management field, specifically a kind of financial data intelligent remote anti-cheating system and its anti-cheating
Method.
Background technology
As the growth of enterprise develops, financial management and control becomes the important service of enterprise.Only guarantee the financial security of enterprise,
Just can ensure that enterprise completes business plan, and then reach the strategic objective of enterprise.The financial management and control of conglomerate is that group implements
One of most basic means of effective management and control.How conglomerate implements the financial management and control to subsidiary, is current group company face
The highly important problem faced.In recent years, conglomerate's Financial Information System construction achieves notable results, while have accumulated
Substantial amounts of business datum.But there is also following problem：Basic financial management software disunity, such as SAP, Oracle ERP, UFSOFT
NC, golden dish KIS etc.；The degree of integration of financial related system is not high, causes many information to circulate to get up to compare in each system
Slowly.For further hoisting power, in the urgent need to integrating to these data, to improve the science of decision-making.In addition, one
A little company there may be financial cheating situation, manipulate accounting profit, it is intended to by whitewashing financial statement to obtain great number income
With good financial position of the enterprise, basic financial management software disunity, also increase conglomerate and subsidiary is practised fraud
The difficulty of monitoring.
The content of the invention
It is an object of the invention to provide a kind of financial data intelligent remote anti-cheating system and its anti-cheating method, to solve
The problem proposed in certainly above-mentioned background technology.
For achieving the above object, the present invention provides following technical scheme：
A kind of financial data intelligent remote anti-cheating system, including data integration module, DBM, Treatment Analysis mould
Block；The data integration module communication is connected to the financial system of each subsidiary；The financial system includes Data Enter list
Unit, information registration unit and data storage cell, Data Enter unit and information registration unit are connected to data storage cell；
The data integration module includes data extracting unit, data cleansing unit, Date Conversion Unit and data load units, described
The collection terminal of data extracting unit is connected to data storage cell, and the outfan of data extracting unit is connected to data cleansing list
Unit, the financial data that data acquisition unit is used in acquired data storage unit, and the financial data of acquisition is sent to data
Cleaning unit, the other end of data cleansing unit are connected to data load units, data load units through Date Conversion Unit
Communication is connected to DBM, and data cleansing unit is screened to the financial data for obtaining, and by screening after financial number
According to sending to Date Conversion Unit, the financial data from different financial systems is converted into consolidation form by Date Conversion Unit
Financial data, and send to data load units, data load units preserve the financial data after conversion to DBM；
The Treatment Analysis module includes acquiring unit, orthogonal factor analytic unit and recognition unit, acquiring unit one end connection
To DBM, the other end is connected to orthogonal factor analytic unit, and acquiring unit is stored in DBM for obtaining
Subsidiary to be monitored multinomial financial data, and the multinomial financial data for obtaining is sent to orthogonal factor analytic unit, just
The outfan of factorial analyses unit is handed over to be connected to recognition unit, orthogonal factor analytic unit is for acquired multinomial financial number
According to orthogonal factor analysis is carried out, obtain multiple the first orthogonal common factors, formed using the plurality of first common factor as
The first eigenvector of component, and the first eigenvector is sent to into recognition unit, recognition unit utilizes default simplicity
Bayes classifier the financial cheating situation according to first eigenvector identification subsidiary to be monitored.
As further scheme of the invention：The multinomial financial data of the subsidiary to be monitored acquired in the acquiring unit
Including assets liquidity ratio, sales dollar ratio, net profit on sales rate, assets net profit margin, net assets income ratio, operating profit ratio,
Total assets cash recovery rate, stock turnover rate, accounts receivable turnover, turnover rate of fixed assets and interest cover ratio.
As further scheme of the invention：Multinomial finance of the orthogonal factor analytic unit to subsidiary to be monitored
Data carry out orthogonal factor analysis, tire out in choosing the first all orthogonal common factor obtained by orthogonal factor analysis
The profit factor, the cash factor, stable factor and the flow constant that product variance contribution ratio exceedes predetermined threshold value is multiple as what is obtained
The first orthogonal common factor, the profit factor is by net profit on sales rate, assets net profit margin, net assets income ratio and operating profit
Rate reflects by sales dollar ratio, total assets cash recovery rate, the cash factor reflects that the stable factor is by accounts receivable
The reflection of turnover rate, turnover rate of fixed assets and interest cover ratio, the flow constant is by assets liquidity ratio and inventory turnover
Rate reflects.
As further scheme of the invention：The recognition unit is used to calculate each in the first eigenvector
First independent probability of first common factor under cheating class, and described the is calculated according to multiple first independent probabilities for being obtained
One characteristic vector belongs to the first posterior probability of the cheating class；Calculate in the first eigenvector each first it is public because
Second independent probability of the son under non-cheating class, and according to multiple second independent probabilities for being obtained calculate the fisrt feature to
Amount belongs to the second posterior probability of the non-cheating class；When first posterior probability is more than second posterior probability, sentence
The financial cheating situation of fixed subsidiary to be monitored is cheating；When first posterior probability is less than second posterior probability,
Judge the financial cheating situation of subsidiary to be monitored as non-cheating.
As further scheme of the invention：The Treatment Analysis module also includes the default unit of grader, grader
Default unit is used to determine the default Naive Bayes Classifier.
As further scheme of the invention：During the default unit of the grader obtains Duo Jia subsidiaries, every whole family is public
The multinomial history financial data and corresponding history finance cheating information of department；For described per whole family company, to acquired
Multinomial history financial data carry out orthogonal factor analysis and obtain multiple the second orthogonal common factors, formed with the plurality of the
Second feature vector of two common factors as component, and formation includes that the corresponding history finance of the second feature vector sum are made
The sample data of disadvantage information；When the multiple sample datas satisfactions first for being formed are pre-conditioned by the multiple sample numbers for being formed
According to one group of training sample data and one group of test sample data are divided into, wherein described first pre-conditioned multiple samples by being formed
In all second feature vectors that notebook data includes, every kind of second common factor meets Gauss distribution；Using one group of training sample
Notebook data is trained and obtains housebroken Naive Bayes Classifier for the Naive Bayes Classifier for recognizing finance cheating, is utilized
Housebroken Naive Bayes Classifier described in one group of test sample data detection, and it is default to meet second in assay
The housebroken Naive Bayes Classifier is set as into the default Naive Bayes Classifier during condition.
As further scheme of the invention：Also include alarm module, the alarm module communication is connected to process point
Analysis module.
As further scheme of the invention：Also include mobile terminal, the alarm module communication is connected to mobile whole
End.
The anti-cheating method of described financial data intelligent remote anti-cheating system, including it is as follows：
1) data integration module gathers the financial data of the financial system of each subsidiary by data acquisition unit, and will
The financial data of acquisition is sent to data cleansing unit；
2) data cleansing unit is screened to the financial data for obtaining, and the financial data after screening is sent to data
Converting unit；
3) financial data from different financial systems is converted into Date Conversion Unit the financial data of consolidation form, and
Send to data load units；
4) data load units preserve the financial data after conversion to DBM；
5) Treatment Analysis module obtains the multinomial of the subsidiary to be monitored being stored in DBM by acquiring unit
Financial data, and the multinomial financial data for obtaining is sent to orthogonal factor analytic unit；
6) orthogonal factor analytic unit carries out orthogonal factor analysis to acquired multinomial financial data, obtains multiple orthogonal
The first common factor, form first eigenvector using the plurality of first common factor as component, and by described first
Characteristic vector is sent to recognition unit；
7) recognition unit utilizes default Naive Bayes Classifier and according to first eigenvector identification son public affairs to be monitored
The financial cheating situation of department, if there is no cheating situation, return to step 5), the multinomial finance to next subsidiary to be monitored
Data are processed, and otherwise, carry out step 8)；
8) Treatment Analysis module carries out alert process by alarm module.
Compared with prior art, the invention has the beneficial effects as follows：
1st, the financial data from different financial systems can be converted into by the financial data intelligent remote anti-cheating system
The financial data of consolidation form, and be stored in data base, so that subsequent analysis are used, the degree of integration of financial system is improved,
It is easy to the quick circulation in each system of many information.
2nd, the financial data intelligent remote anti-cheating system, can realize conglomerate to the long-range of each subsidiary of subordinate
Financial monitoring, the financial system of each subsidiary of intelligent decision whether there is cheating, be conducive to the finance pipe of conglomerate
Reason.
Description of the drawings
Fig. 1 is the structural representation of financial data intelligent remote anti-cheating system and its anti-cheating method；
Fig. 2 is the alarm module schematic diagram of financial data intelligent remote anti-cheating system and its anti-cheating method；
Fig. 3 is the mobile terminal structure schematic diagram of financial data intelligent remote anti-cheating system and its anti-cheating method.
In figure：1- financial systems, 11- Data Enter units, 12- information registration units, 13- data storage cells, 2- numbers
According to integration module, 21- data extracting units, 22- data cleansing units, 23- Date Conversion Units, 24- data load units, 3-
DBM, 4- Treatment Analysis modules, 41- graders preset unit, 42- acquiring units, 43- orthogonal factor analytic units,
44- recognition units, 5- alarm modules, 6- mobile terminals.
Specific embodiment
Technical scheme is described in more detail with reference to specific embodiment.
Refer to Fig. 1, a kind of financial data intelligent remote anti-cheating system, including data integration module 2, DBM
3rd, Treatment Analysis module 4；The communication of data integration module 2 is connected to the financial system 1 of each subsidiary, the financial system
1 is SPA, Oracle ERP or UFSOFT NC etc., and the financial system 1 includes Data Enter unit 11,12 and of information registration unit
Data storage cell 13, Data Enter unit 11 and information registration unit 12 are connected to data storage cell 13, financial staff
By Data Enter unit 11 by financial information and the typing of single transaction journal to financial system 1, and pass through data storage list
Unit 13 is stored, and the paper money counter or POS of information registration unit 12 carry out financial number statistics in financial access procedure, and
Statistical result feeding data storage cell 13 is stored；The data integration module 2 includes data extracting unit 21, data
Cleaning unit 22, Date Conversion Unit 23 and data load units 24, the collection terminal of the data extracting unit 21 are connected to number
According to memory element 13, the outfan of data extracting unit 21 is connected to data cleansing unit 22, and data acquisition unit 21 is used to adopt
Financial data in collection data storage cell 13, and the financial data of acquisition is sent to data cleansing unit 22, data cleansing
The other end of unit 22 is connected to data load units 24 through Date Conversion Unit 23, and the communication of data load units 24 is connected to
DBM 3, the financial datas of the acquisition of data cleansing unit 22 pairs are screened, and by the financial data after screening send to
Date Conversion Unit 23, the purpose of data cleansing unit 22 are to ensure the quality of data in DBM 3, are embodied in
Correctness, integrity, concordance, completeness, it is ageing and can availability, by collection obtain financial data in, do not meet will
The data asked are weeded out, and predominantly subject managing detailed catalogue arranges lack of standardization or subject and is not set using auxiliary accounting on request
Etc. data, the financial data from different financial systems 1 is converted into Date Conversion Unit 23 financial data of consolidation form, and
Send to data load units 24, data load units 24 preserve the financial data after conversion to DBM 3, for rear
Continuous analysis is used；
The Treatment Analysis module 4 includes acquiring unit 42, orthogonal factor analytic unit 43 and recognition unit 44, described to obtain
Take 42 one end of unit and be connected to DBM 3, the other end is connected to orthogonal factor analytic unit 43, and acquiring unit 42 is used to obtain
Go bail for the multinomial financial data of the subsidiary to be monitored existed in DBM 3, and the multinomial financial data for obtaining is sent
To orthogonal factor analytic unit 43, the outfan of orthogonal factor analytic unit 43 is connected to recognition unit 44, orthogonal factor analysis
Unit 43 obtains multiple the first orthogonal common factors for carrying out orthogonal factor analysis to acquired multinomial financial data,
The first eigenvector using the plurality of first common factor as component is formed, and the first eigenvector is sent to into knowledge
Other unit 44, recognition unit 44 utilize default Naive Bayes Classifier and according to first eigenvector identification son public affairs to be monitored
The financial cheating situation of department；
The multinomial financial data of the subsidiary to be monitored acquired in the acquiring unit 42 includes assets liquidity ratio, sale
Cash ratio, net profit on sales rate, assets net profit margin, net assets income ratio, operating profit ratio, total assets cash recovery rate, stock
Turnover rate, accounts receivable turnover, turnover rate of fixed assets and interest cover ratio；
The orthogonal factor analytic unit 43 carries out orthogonal factor analysis to the multinomial financial data of subsidiary to be monitored, choosing
In taking the first all orthogonal common factor obtained by orthogonal factor analysis, cumulative proportion in ANOVA exceedes default threshold
The profit factor of value, the cash factor, stable factor and flow constant are as the first multiple orthogonal common factor for obtaining, described
The profit factor is mainly by the reflection of net profit on sales rate, assets net profit margin, net assets income ratio and operating profit ratio, the cash factor
It is main to reflect that the stable factor is mainly by accounts receivable turnover, fixation by sales dollar ratio, total assets cash recovery rate
Asset turnover and interest cover ratio reflection, the flow constant are mainly reflected by assets liquidity ratio and stock turnover rate；
The recognition unit 44 is only for calculating each first common factor in the first eigenvector under cheating class first
Vertical probability, and belong to the first of the cheating class according to the multiple first independent probabilities calculating first eigenvector for being obtained
Posterior probability；Second independent probability of each first common factor in the first eigenvector under non-cheating class is calculated,
And the second posteriority that the first eigenvector belongs to the non-cheating class is calculated according to multiple second independent probabilities for being obtained
Probability；When first posterior probability is more than second posterior probability, the financial cheating situation of subsidiary to be monitored is judged
For cheating；When first posterior probability is less than second posterior probability, the finance cheating shape of subsidiary to be monitored is judged
Condition is non-cheating；
The Treatment Analysis module 4 also includes that the default unit 41 of grader is used to determine the default naive Bayesian point
Class device；Specifically, the default unit 41 of grader can be used for：Obtain the multinomial history wealth of every whole family company in Duo Jia subsidiaries
Business data and corresponding history finance cheating information；For described per whole family company, to acquired multinomial history finance
Data carry out orthogonal factor analysis and obtain multiple the second orthogonal common factors, formed using the plurality of second common factor as
The second feature vector of component, and form the sample number for including the corresponding history finance cheating information of the second feature vector sum
According to；The multiple sample datas for being formed are divided into into one group of training when the multiple sample datas satisfactions first for being formed are pre-conditioned
Sample data and one group of test sample data, wherein the described first pre-conditioned institute included by the multiple sample datas for being formed
In having second feature vector, every kind of second common factor meets Gauss distribution；Utilizing one group of training sample data to train is used for
The Naive Bayes Classifier of identification finance cheating obtains housebroken Naive Bayes Classifier, using one group of test specimens
Notebook data checks the housebroken Naive Bayes Classifier, and when assay satisfaction second is pre-conditioned by the Jing
The Naive Bayes Classifier of training is set as the default Naive Bayes Classifier；
The financial data intelligent remote anti-cheating system, also including alarm module 5, the communication connection of the alarm module 5
To Treatment Analysis module 4, when Treatment Analysis module 4 judges to have cheating situation, reported to the police by alarm module 5, so as to phase
Close staff to be processed in time；The financial data intelligent remote anti-cheating system, also including mobile terminal 6, the report
The alert communication of module 5 is connected to mobile terminal 6, and mobile terminal is mobile phone or computer, and mobile terminal 6 is equipped to corresponding work people
Warning information is sent into mobile terminal 6 by member, alarm module 5 by wireless network, knows police in time in order to staff
Notify and breath processed.
It is preferred as shown in Fig. 2 the alarm module 5 has a surface-mounted integrated circuit, described surface-mounted integrated circuit be provided with
Cheating information is carried out storing standby tune by the central processing unit 52 of the connection of Treatment Analysis module 4, memorizer 53, the memorizer；It is described
Central processing unit 52 is connected to alarm 1, alarm 2 51, communication module 55, SMS transmission module 56, automatic language
Sound dial module 57；Wherein, the communication module 55 is GPRS or 4G or WLAN, alarm 1 are sound and light alarm
Device, alarm 2 51 are abnormal smells from the patient alarm；Memorizer 53 is that 100T portable hard drives being capable of data storage for a long time；When Treatment Analysis mould
Block 4 is sent to cheating signal after central processing unit 52, control SMS transmission module 56, the automatic speech dialing of central processing unit 52
The preset SMS prompt text of cheating is sent to mobile terminal 6 by communication module 55 and dials reserved mobile end by module 57
End number, while controlling alarm 1, alarm 2 51 carries out the warning of acousto-optic and abnormal smells from the patient form.
As shown in figure 3, mobile terminal 6 includes being provided with signal receiving module 62 and connecting with which in housing 61. housing 61
The processor 63 for connecing, the processor 63 are connected with flashing light chip 64；61 upper end of the housing is provided with multiple and flashing light chip 64
The LED 65 of connection.The flashing light chip 64 is 8 sections of 4 tunnel, and 8 sections of flashing modes are circulated successively.
Naive Bayes Classifier is once introduced below.It is a kind of conditional probability to the Naive Bayes Classifier
Model.When being classified using Naive Bayes Classifier, only meet " class conditional independence " and require to be a characteristic attribute
It is worth to the impact for giving class independently of other characteristic attribute values, can just realizes preferable classifying quality.In order to tackle simple pattra leaves
Requirement of this grader to class conditional independence, analyzes (setting up Factor analysis) using orthogonal factor and processes selected change
Amount, to obtain multiple common factors for meeting class conditional independence as the characteristic attribute value for being classified.Orthogonal factor
Analysis is the one kind in factor analyses, and the essence of factor analyses is：Go to retouch with random quantity that is potential but can not observing
State the covariance relationship between many variables.Factor analyses are based primarily upon following proposition：It is assumed that the dependency between variable can be used
Variable is grouped, that is, is assumed that be height correlation between all variables in a specific group, and from different groups
In variable but have relatively small dependency.Thus, take out single latent factor to characterize each group variable.Above proposition
Ensure that between each monofactor for taking out it is class conditional sampling.
Factor analyses are related dependences inside research variable, and some are had intricate relation
Variable is attributed to a kind of multivariate statistical analysis method of a few multi-stress.Set up Factor Analysis Model to be described as follows：
For the observation random vector X (X for having P composition1, X2..., XP), mean vector E (X)=0 and covariance matrix
Cov (X)=∑.As represented by following formula, x-ray depends on the common factor F that several can not be observed1, F2...,
FmWith p additional specific factor e1, e2..., ep。
X1=l11F1+l12F2+…+l1mFm+e1
X2=l21F1+l22F2+…+l2mFm+e2
……
Xp=lp1F1+lp2F2+…+lpmFm+ep。
Wherein coefficient lijThe load of referred to as i-th variable in j-th factor, L is Factor load-matrix.
Make X=(X1, X2..., Xp) ', F=(F1, F2..., Fm) ', e=(e1, e2..., ep) ', is if meet following conditions：F
With e independences；Y is diagonal matrix,
Then meet orthogonality, now the orthogonal model of m common factor can be expressed from the next：
Wherein eiFor i-th specific factor, FiI-th common factor, lijI-th variable is on j-th common factor
Load.
The course of work of Naive Bayes Classifier is briefly described below.Suppose there is t class C1, C2..., Ct, for given
Unknown data sample X ', grader by predict X ' belong to the class with highest posterior probability (under condition X ').That is, plain
Unknown sample X ' is distributed to class C by plain Bayes classifieriNecessary and sufficient condition be expressed from the next：
P(Ci/ X ') ＞ P (Cj/ X '), i ≠ j, t=i, j
Wherein P (Ci/ X ') maximum class CiReferred to as maximum a posteriori assumes.Following formula is had according to Bayes theorem：
It is given that there is perhaps multiattribute data set, calculate P (X '/Ci) expense may be very big.For reduce calculate P (X '/
Ci) expense, the simple hypothesis of class conditional sampling can be done.The class label of given sample, it is assumed that property value condition of reciprocity independence,
There is no dependence i.e. between attribute, therefore obtain following formula：
Probability P (X1’/Ci), P (X2’/Ci) ..., P (Xs’/Ci) can be by training sample valuation.
P (X') is constant to all classes.Ratio of classification Ci in sample set is P (Ci)=Si/ S (wherein SiIt is class CiIn
Number of training, S be choose total sample number).Each class is belonged to the unknown data sample X ' that this obtains being expressed from the next
Other probability：
If given attribute character is continuous, it is m to often assume that successive value obeys average, and variance is the Gauss point of s
, then there is following formula in cloth (i.e. normal distribution)：
The anti-cheating method of the financial data intelligent remote anti-cheating system, step are as follows：
1) data integration module 2 gathers the financial data of the financial system of each subsidiary by data acquisition unit 21,
And the financial data of acquisition is sent to data cleansing unit 22；
2) financial data of 22 pairs of acquisitions of data cleansing unit is screened, and the financial data after screening is sent to number
According to converting unit 23；
3) financial data from different financial systems 1 is converted into Date Conversion Unit 23 the financial number of consolidation form
According to, and send to data load units 24；
4) data load units 24 preserve the financial data after conversion to DBM 3；
5) Treatment Analysis module 4 obtains the subsidiary to be monitored being stored in DBM 3 by acquiring unit 42
Multinomial financial data, and the multinomial financial data for obtaining is sent to orthogonal factor analytic unit 43；
The multinomial financial data of acquired subsidiary to be monitored can include：Assets liquidity ratio X1, sales dollar ratio
Rate X2, net profit on sales rate X3, assets net profit margin X4, net assets income ratio X5, operating profit ratio X6, total assets cash recovery rate X7、
Stock turnover rate X8, accounts receivable turnover X9, turnover rate of fixed assets X10With interest cover ratio X11.Specifically, asset stream
Dynamic ratio X1Represent that total of current asset/current liability adds up to, sales dollar ratio X2Represent business activities net cash flow/master
Battalion's health service revenue, net profit on sales rate X3Represent net profit/income from sales, assets net profit margin X4Represent that net profit/average assets are total
Volume, net assets income ratio X5Represent net profit/average net assets, operating profit ratio X6Operating profit/whole health service revenue is represented,
Total assets cash recovery rate X7Represent cash net amount/average total assets, stock turnover rate X8Expression cost of marketing/averagely deposit
Goods remaining sum, accounts receivable turnover X9The current sale net income of expression/[(more than initial account receivable remaining sum+end of term account receivable
Volume)/2], turnover rate of fixed assets X10Represent income from sales/average net fixed assets, interest cover ratio X11Represent (profit
Total value+interest expense)/interest expense.
6) the multinomial financial data acquired in orthogonal factor analytic unit 43 pairs carries out orthogonal factor analysis, obtain it is multiple just
The first common factor handed over, forms first eigenvector using the plurality of first common factor as component, and by described the
One characteristic vector is sent to recognition unit 44；
By to producing liquidity ratio X1, sales dollar ratio X2, net profit on sales rate X3, assets net profit margin X4, Net Gains of Asset
Rate X5, operating profit ratio X6, total assets cash recovery rate X7, stock turnover rate X8, accounts receivable turnover X9, fixed assets turnover
Rate X10With interest cover ratio X11, each variable carry out KMO and Bartlett test, KMO statistic of test be 0.713, therefore
Think that the partial correlation between above-mentioned each variable can receive.The Sig of Bartlett sphericity test statistics<0.01, therefore recognize
To there is significant dependency between above-mentioned each variable.Therefore, above-mentioned each variable in the present embodiment be adapted to do it is orthogonal because
Son analysis.
Carry out orthogonal factor analysis to above-mentioned each variable, choose and obtain all orthogonal is analyzed by the orthogonal factor
The first common factor in cumulative proportion in ANOVA exceed the profit factor of predetermined threshold value, the cash factor, stable factor and flowing
, used as the first multiple orthogonal common factor for obtaining, the profit factor is mainly by net profit on sales rate X for the factor3, assets net profit
Rate X4, net assets income ratio X5, operating profit ratio X6Reflection, the cash factor is mainly by sales dollar ratio X2, total assets it is existing
Gold recovery X7Reflection, the stable factor is mainly by accounts receivable turnover X9, turnover rate of fixed assets X10With interest guarantee
Multiple X11Reflection, the flow constant is mainly by product liquidity ratio X1, stock turnover rate X8Reflection.
Above-mentioned predetermined threshold value can be set by the user.Four resulting the first orthogonal common factors are phases each other
It is mutually independent, therefore the class conditional independence assumption of Naive Bayes Classification can be eliminated, you can with Naive Bayes Classifier
Classification and Identification is carried out based on four resulting the first orthogonal common factors.F1、F2、F3And F4Represent respectively the profit because
Son, the cash factor, stable factor and flow constant, obtain first eigenvector F=(F1, F2, F3, F4)
M orthogonal common factor F1 is determined using orthogonal factor analysis, F2 ..., Fm are tried to achieve
Individual common factor F1, F2..., Fs(s=4 herein) can fully react the integral level of original variable very much.Each is treated
The data sample of monitoring subsidiary has a s dimension attribute feature vector, Xs '={ f1, f2..., fsRepresent, fi(i=1,2 ..., s)
Represent to attribute Fi(i=1,2 ..., tolerance s).
7) recognition unit 44 utilizes default Naive Bayes Classifier and recognizes son to be monitored according to first eigenvector
The financial cheating situation of company, if there is no cheating situation, return to step 5), the multinomial wealth to next subsidiary to be monitored
Business data are processed, and otherwise, carry out step 8)；
It is described to recognize that the son to be monitored is public using default Naive Bayes Classifier, according to the first eigenvector
The Financial fraud situation of department, can include：Each first common factor in the first eigenvector is calculated under malpractices class
The first independent probability, and calculate the first eigenvector according to multiple first independent probabilities for being obtained and belong to the malpractices
First posterior probability of class；Calculate each first common factor in the first eigenvector under non-malpractices class second only
Vertical probability, and calculate that the first eigenvector belongs to the non-malpractices class according to multiple second independent probabilities for being obtained the
Two posterior probability；When first posterior probability is more than second posterior probability, the wealth of the subsidiary to be monitored is judged
Business malpractices situation is malpractices；When first posterior probability is less than second posterior probability, judge that the son to be monitored is public
The Financial fraud situation of department is non-malpractices.
Specifically, it is identified in the following way.In the embodiment that the present invention is provided, malpractices class C is only existed1, it is non-
Malpractices class C2Two classifications, wherein malpractices class C1The Financial fraud situation of expression company is malpractices, non-malpractices class C2Expression company
Financial fraud situation is non-malpractices.Make above-mentioned first eigenvector F=(F1, F2, F3, F4) naive Bayesian point is carried out as X '
Class, i.e., with F1、F2、F3And F4Naive Bayes Classification is carried out respectively as characteristic attribute value.
When P (malpractices) ＞ P (non-malpractices), it is possible to determine that the Financial fraud situation of the subsidiary to be monitored is malpractices；
When P (malpractices) ＜ P (non-malpractices), it is possible to determine that the Financial fraud situation of the subsidiary to be monitored is non-malpractices；When P (dances
Disadvantage)=P (non-malpractices) when, then cannot judge.For above formula, P (Fk/C1) represent each first common factor in malpractices class C1Under
The first independent probability, P (malpractices)=P (C1/ F) represent that the first eigenvector belongs to malpractices class C1The first posteriority
Probability.P(Fk/C2) represent each first common factor in non-malpractices class C2Under the second independent probability, P (non-malpractices)=P (C2/
F) represent that the first eigenvector belongs to non-malpractices class C2The second posterior probability.
8) Treatment Analysis module 4 carries out alert process by alarm module 5.
The Naive Bayes Classifier by finance cheating condition feedback in Treatment Analysis module, lead to by the Treatment Analysis module
Cross telecommunication device with note or with phone in the form of cheating information is sent to into receiving terminal；
The alarm module points out form and abnormal smells from the patient signal prompt form to carry out alarm with sound and light signal；
The mobile terminal receives the report for police service after processing to signal after signal by flashing light chip controls mobile terminal
LED is circulated 8 sections of flash for prompting.
The step 2
Financial data from different financial systems can be converted into by the financial data intelligent remote anti-cheating system
The financial data of consolidation form, and be stored in data base, so that subsequent analysis are used, the degree of integration of financial system is improved,
It is easy to the quick circulation in each system of many information.The financial data intelligent remote anti-cheating system, can realize
Conglomerate is monitored to the Tele Financing of each subsidiary of subordinate, and the financial system of each subsidiary of intelligent decision is with the presence or absence of work
Disadvantage behavior, is conducive to the financial management of conglomerate.
Above the better embodiment of the present invention is explained in detail, but the present invention is not limited to above-mentioned embodiment party
Formula, in the ken that one skilled in the relevant art possesses, can be with the premise of without departing from present inventive concept
Various changes can be made.
Claims (9)
1. a kind of financial data intelligent remote anti-cheating system, it is characterised in that including data integration module (2), data base's mould
Block (3), Treatment Analysis module (4)；Data integration module (2) communication is connected to the financial system (1) of each subsidiary；Institute
Financial system (1) is stated including Data Enter unit (11), information registration unit (12) and data storage cell (13), Data Enter
Unit (11) and information registration unit (12) are connected to data storage cell (13)；The data integration module (2) is including number
According to extraction unit (21), data cleansing unit (22), Date Conversion Unit (23) and data load units (24), the data are carried
The collection terminal for taking unit (21) is connected to data storage cell (13), and it is clear that the outfan of data extracting unit (21) is connected to data
Unit (22) is washed, data acquisition unit (21) is for the financial data in acquired data storage unit (13), and the wealth that will be obtained
Business data is activation connects to data cleansing unit (22), the other end of data cleansing unit (22) through Date Conversion Unit (23)
To data load units (24), data load units (24) communication is connected to DBM (3), and data cleansing unit (22) is right
The financial data of acquisition is screened, and the financial data after screening is sent to Date Conversion Unit (23), data conversion list
First (23) financial data from different financial systems (1) is converted into the financial data of consolidation form, and is sent to data dress
Carrier unit (24), data load units (24) preserve the financial data after conversion to DBM (3)；The Treatment Analysis
Module (4) includes acquiring unit (42), orthogonal factor analytic unit (43) and recognition unit (44), the acquiring unit (42)
End is connected to DBM (3), and the other end is connected to orthogonal factor analytic unit (43), and acquiring unit (42) is protected for obtaining
There is the multinomial financial data of the subsidiary to be monitored in DBM (3), and by the multinomial financial data for obtaining send to
Orthogonal factor analytic unit (43), the outfan of orthogonal factor analytic unit (43) are connected to recognition unit (44), orthogonal factor
Analytic unit (43) obtains multiple orthogonal first public for carrying out orthogonal factor analysis to acquired multinomial financial data
The factor, forms the first eigenvector using the plurality of first common factor as component, and the first eigenvector is sent out
Recognition unit (44) is given, recognition unit (44) is recognized using default Naive Bayes Classifier and according to first eigenvector
The financial cheating situation of subsidiary to be monitored.
2. financial data intelligent remote anti-cheating system according to claim 1, it is characterised in that the acquiring unit
(42) the multinomial financial data of the subsidiary to be monitored acquired in includes assets liquidity ratio, sales dollar ratio, net profit on sales
Rate, assets net profit margin, net assets income ratio, operating profit ratio, total assets cash recovery rate, stock turnover rate, accounts receivable week
Rate of rotation, turnover rate of fixed assets and interest cover ratio.
3. financial data intelligent remote anti-cheating system according to claim 2, it is characterised in that the orthogonal factor point
Analysis unit (43) carries out orthogonal factor analysis to the multinomial financial data of subsidiary to be monitored, chooses by the orthogonal factor point
In the first all orthogonal common factor for obtaining of analysis cumulative proportion in ANOVA exceed the profit factor of predetermined threshold value, cash because
, used as the first multiple orthogonal common factor for obtaining, the profit factor is by net profit on sales for son, stable factor and flow constant
The reflection of rate, assets net profit margin, net assets income ratio and operating profit ratio, the cash factor is by sales dollar ratio, total assets
Cash recovery rate reflects that the stable factor is reflected by accounts receivable turnover, turnover rate of fixed assets and interest cover ratio,
The flow constant is reflected by assets liquidity ratio and stock turnover rate.
4. financial data intelligent remote anti-cheating system according to claim 3, it is characterised in that the recognition unit
(44) for calculating first independent probability of each first common factor in the first eigenvector under cheating class, and root
The first posterior probability that the first eigenvector belongs to the cheating class is calculated according to multiple first independent probabilities for being obtained；Meter
Calculate the second independent probability of each first common factor in the first eigenvector under non-cheating class, and according to being obtained
Multiple second independent probabilities calculate the second posterior probability that the first eigenvector belongs to the non-cheating class；When described
When one posterior probability is more than second posterior probability, judge the financial cheating situation of subsidiary to be monitored as cheating；When described
When first posterior probability is less than second posterior probability, judge the financial cheating situation of subsidiary to be monitored as non-cheating.
5. the financial data intelligent remote anti-cheating system according to claim 1 or 2 or 3 or 4, it is characterised in that described
Treatment Analysis module (4) also includes the default unit (41) of grader, and it is described default for determining that grader presets unit (41)
Naive Bayes Classifier.
6. financial data intelligent remote anti-cheating system according to claim 5, it is characterised in that the grader is preset
Unit (41) obtains the multinomial history financial data of every whole family company in Duo Jia subsidiaries and the finance cheating of corresponding history
Information；For described per whole family company, orthogonal factor analysis carried out to acquired multinomial history financial data and obtains multiple
The second orthogonal common factor, forms the second feature vector using the plurality of second common factor as component, and forms bag
Include the sample data of the corresponding history finance cheating information of the second feature vector sum；When the multiple sample datas for being formed expire
The multiple sample datas for being formed are divided into one group of training sample data and one group of test sample data when pre-conditioned by foot first,
In the wherein described first pre-conditioned all second feature vectors included by the multiple sample datas for being formed, every kind of second is public
Common factor meets Gauss distribution；One group of training sample data are utilized to train for recognizing the naive Bayesian point of finance cheating
Class device obtains housebroken Naive Bayes Classifier, using housebroken simplicity described in one group of test sample data detection
Bayes classifier, and the housebroken Naive Bayes Classifier is set when assay satisfaction second is pre-conditioned
For the default Naive Bayes Classifier.
7. financial data intelligent remote anti-cheating system according to claim 6, it is characterised in that also including alarm module
(5), alarm module (5) communication is connected to Treatment Analysis module (4).
8. financial data intelligent remote anti-cheating system according to claim 7, it is characterised in that also including mobile terminal
(6), alarm module (5) communication is connected to mobile terminal (6).
9. a kind of anti-cheating method of financial data intelligent remote anti-cheating system as claimed in claim 8, it is characterised in that
Including as follows：
Data integration module (2) gathers the financial data of the financial system of each subsidiary by data acquisition unit (21), and
The financial data of acquisition is sent to data cleansing unit (22)；
Data cleansing unit (22) is screened to the financial data for obtaining, and the financial data after screening is sent to data turn
Change unit (23)；
Financial data from different financial systems (1) is converted into Date Conversion Unit (23) financial data of consolidation form,
And send to data load units (24)；
Data load units (24) preserve the financial data after conversion to DBM (3)；
Treatment Analysis module (4) obtains the subsidiary to be monitored being stored in DBM (3) by acquiring unit (42)
Multinomial financial data, and the multinomial financial data for obtaining is sent to orthogonal factor analytic unit (43)；
Orthogonal factor analytic unit (43) carries out orthogonal factor analysis to acquired multinomial financial data, obtains multiple orthogonal
First common factor, forms the first eigenvector using the plurality of first common factor as component, and special by described first
Levy vector and be sent to recognition unit (44)；
Recognition unit (44) recognizes subsidiary to be monitored using default Naive Bayes Classifier and according to first eigenvector
Financial cheating situation, if there is no cheating situation, return to step 5), the multinomial financial number to next subsidiary to be monitored
According to being processed, otherwise, carry out step 8)；
Treatment Analysis module (4) carries out alert process by alarm module (5).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611155908.5A CN106599200A (en) | 2016-12-14 | 2016-12-14 | Intelligent remote financial data anti-cheating system and anti-cheating method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611155908.5A CN106599200A (en) | 2016-12-14 | 2016-12-14 | Intelligent remote financial data anti-cheating system and anti-cheating method thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106599200A true CN106599200A (en) | 2017-04-26 |
Family
ID=58801331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611155908.5A Pending CN106599200A (en) | 2016-12-14 | 2016-12-14 | Intelligent remote financial data anti-cheating system and anti-cheating method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106599200A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109240163A (en) * | 2018-09-25 | 2019-01-18 | 南京信息工程大学 | Intelligent node and its control method for industrialization manufacture |
| CN109801097A (en) * | 2018-12-14 | 2019-05-24 | 深圳壹账通智能科技有限公司 | Analysis method, device, storage medium and the analytical equipment of operation data |
| CN110674131A (en) * | 2019-08-30 | 2020-01-10 | 深圳壹账通智能科技有限公司 | Financial statement data processing method and device, computer equipment and storage medium |
| CN110717078A (en) * | 2019-09-16 | 2020-01-21 | 武汉安诠加信息技术有限公司 | Beauty shop business data monitoring method, device, equipment and medium |
| CN111091456A (en) * | 2019-12-13 | 2020-05-01 | 天津中德应用技术大学 | Financial management system and method |
| CN111553597A (en) * | 2020-04-29 | 2020-08-18 | 支付宝(杭州)信息技术有限公司 | Method and device for carrying out financial fraud risk identification on enterprise |
| CN111681044A (en) * | 2020-05-28 | 2020-09-18 | 中国工商银行股份有限公司 | Method and device for processing point exchange cheating behaviors |
| CN117687764A (en) * | 2024-02-04 | 2024-03-12 | 南京九洲会计咨询有限公司 | Financial data intelligent accounting method and system based on SaaS platform |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678451A (en) * | 2016-01-04 | 2016-06-15 | 宁宇新 | Method and device for automatically identifying financial fraud on the basis of financial data |
-
2016
- 2016-12-14 CN CN201611155908.5A patent/CN106599200A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105678451A (en) * | 2016-01-04 | 2016-06-15 | 宁宇新 | Method and device for automatically identifying financial fraud on the basis of financial data |
Non-Patent Citations (1)
| Title |
|---|
| 夏之凉溪: "E财通集团财务综合监管信息系统方案简介", 《百度文库》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109240163A (en) * | 2018-09-25 | 2019-01-18 | 南京信息工程大学 | Intelligent node and its control method for industrialization manufacture |
| CN109240163B (en) * | 2018-09-25 | 2024-01-02 | 南京信息工程大学 | Intelligent node for industrial manufacturing and control method thereof |
| CN109801097A (en) * | 2018-12-14 | 2019-05-24 | 深圳壹账通智能科技有限公司 | Analysis method, device, storage medium and the analytical equipment of operation data |
| CN110674131A (en) * | 2019-08-30 | 2020-01-10 | 深圳壹账通智能科技有限公司 | Financial statement data processing method and device, computer equipment and storage medium |
| CN110717078A (en) * | 2019-09-16 | 2020-01-21 | 武汉安诠加信息技术有限公司 | Beauty shop business data monitoring method, device, equipment and medium |
| CN111091456A (en) * | 2019-12-13 | 2020-05-01 | 天津中德应用技术大学 | Financial management system and method |
| CN111553597A (en) * | 2020-04-29 | 2020-08-18 | 支付宝(杭州)信息技术有限公司 | Method and device for carrying out financial fraud risk identification on enterprise |
| CN111681044A (en) * | 2020-05-28 | 2020-09-18 | 中国工商银行股份有限公司 | Method and device for processing point exchange cheating behaviors |
| CN117687764A (en) * | 2024-02-04 | 2024-03-12 | 南京九洲会计咨询有限公司 | Financial data intelligent accounting method and system based on SaaS platform |
| CN117687764B (en) * | 2024-02-04 | 2024-04-30 | 南京九洲会计咨询有限公司 | Financial data intelligent accounting method and system based on SaaS platform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106599200A (en) | Intelligent remote financial data anti-cheating system and anti-cheating method thereof | |
| CN113240509B (en) | Loan risk assessment method based on multi-source data federal learning | |
| US7386506B2 (en) | Fraud score calculating program, method of calculating fraud score, and fraud score calculating system for credit cards | |
| US20210365963A1 (en) | Target customer identification method and device, electronic device and medium | |
| CN110659318B (en) | Big data-based policy pushing method, system and computer equipment | |
| CN111291900A (en) | Method and device for training risk recognition model | |
| Garrido et al. | A Robust profit measure for binary classification model evaluation | |
| CN110689437A (en) | Communication construction project financial risk prediction method based on random forest | |
| CN112487284A (en) | Bank customer portrait generation method, equipment, storage medium and device | |
| CN110490750A (en) | Data know method for distinguishing, system, electronic equipment and computer storage medium | |
| CN114372884A (en) | Risk identification method and risk identification device for transaction data | |
| CN112835910B (en) | Method and device for processing enterprise information and policy information | |
| Jurgovsky | Context-aware credit card fraud detection | |
| CN115713399B (en) | User credit evaluation system combined with third-party data source | |
| KR100686466B1 (en) | System and method for valuing loan portfolios using fuzzy clustering | |
| CN116029839B (en) | Automatic execution operating system for operation of service system | |
| CN118071482A (en) | Method for constructing retail credit risk prediction model and consumer credit business Scorebetad model | |
| Zhu | DeepValidity: deep convolutional neural networks for multi-factors forecasting | |
| CN116308590A (en) | Bill product pushing method, device and system | |
| CN114022712A (en) | User classification method and device, computer equipment and storage medium | |
| Thakur | NEURAL NETWORKS FOR DATA MINING | |
| Norrbin et al. | I (2) representations of US money demand | |
| CN118504929A (en) | Digital hotel management system | |
| CN118428980A (en) | Customer grouping method and device based on decision tree | |
| CN117764692A (en) | Method for predicting credit risk default probability |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170426 |