CN109446188A

CN109446188A - Follow-up auditing method and robot system based on big data and deep learning

Info

Publication number: CN109446188A
Application number: CN201811207425.4A
Authority: CN
Inventors: 朱定局
Original assignee: Daguo Innovation Intelligent Technology Dongguan Co ltd
Current assignee: Daguo Innovation Intelligent Technology Dongguan Co ltd
Priority date: 2018-10-17
Filing date: 2018-10-17
Publication date: 2019-03-08
Anticipated expiration: 2038-10-17
Also published as: CN109446188B

Abstract

Follow-up auditing method and robot system based on big data and deep learning, it include: the auditing standard for obtaining pre-set categories, obtain the data of object to be audited, the corresponding data of the auditing standard are obtained from the data of the object, the corresponding data of the to be audited auditing standard of object described in when the corresponding data of the auditing standard of the object wait audit are audited with the last time are compared and judge whether to be changed: being then to go to follow-up auditing judgment step to execute；It is no, then it goes to duplication auditing result step and executes.The above method and system improve timeliness, objectivity, confidence level and the efficiency of audit by the follow-up auditing technology based on big data and deep learning, and reduce the cost of audit.

Description

Follow-up auditing method and robot system based on big data and deep learning

Technical field

The present invention relates to information technology fields, more particularly to a kind of follow-up auditing side based on big data and deep learning Method and robot system.

Background technique

Realize process of the present invention in, inventor discovery at least there are the following problems in the prior art: under the prior art into Row audit (such as financial audit, effectiveness audit etc.) is all disposably, just no longer to track after having audited, some objects (object includes enterprise, individual etc.) may pass through in audit, but after once passing through, just there is no the pressures of audit for object Power, and loosen the requirement to oneself, so that the requirement of audit may be no longer complies in a period of time after passing through of auditing, but simultaneously It will not influence auditing before as a result, such result that will lead to audit does not have timeliness, that is to say, that past audit As a result not necessarily meet the now practical situation of object, audit can thus be made to lose credible and meaning；And some objects After audit does not pass through, make great efforts to keep forging ahead, after making great efforts by the regular hour, although having had reached the standard of audit, Because audit failure before, object dare not slowly be applied auditing always again, in some instances it may even be possible to which audit fails and loses before because It goes confidence and no longer applies for audit always, can thus arrive the object for leading to have reached auditing standard because not applying again It audits and is constantly in the state that do not audit, so that the result that will lead to audit is not comprehensive；To sum up, existing audit technique It will lead to and no longer meet the object of auditing standard and still maintain the result that passes through of audit；And the object for having met auditing standard lacks It is unacceptable as a result, to can not grow with each passing hour to still maintain audit, the result audited loses timeliness, objectivity and can Letter property；In existing audit technique largely audit pass through the result is that permanent effective, then its disadvantage is just as described above very Obviously, minority also audits passing through the result is that in certain period of time (such as several years) effectively, needing to examine again after expired Meter, this aspect still can not solve drawbacks discussed above because described effective period of time be typically all it is long, Object can occur much to change in that time, it is most likely that become not meeting auditing standard from auditing standard is met, thus still The result of audit is so caused to lose timeliness, objectivity and credibility；Meanwhile on the other hand, the result passed through of auditing only exists In certain period of time effectively, need to audit again after expired, this aspect can waste the object passed through of having audited time and Manpower and material resources apply auditing again, while wasting time and the manpower and material resources of Internal Audit Supervisor's unit, because if having examined Data variation the auditing as a result, being so just not necessarily to application again without influence on audit for counting the object passed through, to lead Cause at high cost, the low efficiency of existing audit technique；The third aspect is also most important aspect, and the prior art is all in Internal Audit Supervisor The audit time of agencies dictate audits, and the object audited is caused not have pressure usually, do not making great efforts, when being in audit In evening, interim effort is also of no avail, results in the object that some are audited to audit by making a desperate move, makes and look for pass The illegal activities such as it is, plays tricks.

Therefore, the existing technology needs to be improved and developed.

Summary of the invention

Based on this, it is necessary to it is for the defects in the prior art or insufficient, provide based on big data and deep learning with Track auditing method and robot system are insufficient to solve the timeliness audited in the prior art, objectivity, confidence level, accuracy The shortcomings that at high cost, low efficiency.

In a first aspect, the embodiment of the present invention provides a kind of auditing method, which comprises

Auditing standard obtaining step, for obtaining the auditing standard of pre-set categories；

Object data obtaining step, for obtaining the data of object to be audited；The data of acquisition include third party's data, Data are more objective and feasible, it is possible to improve the objectivity and confidence level of audit；If data are not objective, audit also would not Accurately, because data are more objective, and then accuracy is also just improved；

The corresponding data acquisition step of standard, for obtaining the auditing standard from the data of the object to be audited Corresponding data；

Data variation detecting step, for by the corresponding data of the auditing standard and upper one of the object to be audited The corresponding data of the auditing standard of object to be audited described in when secondary audit are compared and judge whether to be changed: It is then to go to the execution of follow-up auditing judgment step；It is no, then it goes to duplication auditing result step and executes；

Follow-up auditing judgment step, for judging whether the corresponding data of the auditing standard meet the auditing standard；

Replicate auditing result step, for using the last auditing result of the object to be audited as described in wait audit Object this auditing result；

Tracing control step every preset time period or reaches default for obtaining preset time period or preset time point Time point just re-executes all steps in the auditing method.It audits no matter whether object to be audited is applied, can all determine Phase carries out follow-up auditing, so as to improve the timeliness of audit, while can also to audit more to meet the real-time condition of object, So that the result of audit is more credible and accurate.Meanwhile after audit is expired, weight is not necessarily to for the object that follow-up auditing passes through It is new to carry out manual audit, to reduce costs.

Preferably, the method also includes:

As a result feedback step for judging the variation of auditing result, and is sent to the object to be audited；

Warning step, for judging that it is obstructed that the object to be audited is audited in this and continuous several times follow-up auditing Whether the number crossed reaches threshold value: being then to carry out early warning.

Preferably, the object data obtaining step includes:

Data source obtaining step, for obtaining data source；

Object data retrieving step, for the data of the object to be audited to be retrieved and obtained from the data source；

The corresponding data acquisition step of the standard includes:

Data screening step, for filtering out the corresponding number of the auditing standard from the data of the object to be audited According to as the first data；

Data cleansing step, for extracting data corresponding with each single item standard from first data as described every Corresponding second data of one standard.

Preferably,

The result feedback step includes:

Relatively and judge that the object to be audited was examined with the object to be audited in last time in this result audited Whether the result of meter changes: being, then object to be audited described in the result notice by variation the knot of judgement this audit Whether fruit is that audit passes through:

It is then to notify that object to be audited has passed through audit or notice invites the object to be audited to walk application audit Process.Because this is judging automatically for system, and needs some administrative procedures in practical audit process, such as pair to be audited Some application materials of the unit stamp of elephant, so the audit of some classifications needs that the object to be audited to be notified to walk application The process of audit；

It is no, then notify the object to be audited to be rectified and improved and waited next follow-up auditing, it will not be by described default The result of the audit of classification feeds back to described to the audit target, and obtains incongruent each single item standard or overall standard is fed back to It is described to the audit target, and send described in information reminding to the audit target for incongruent each single item standard or overall mark Standard is rectified and improved；

The warning step: obtaining preset first number, and the judgement object to be audited is at this and continuously more Whether unacceptable number of auditing in secondary tracking audit has reached preset first number:

It is then to send information to notify the object to be audited will be by relevant investigation and punishment, breath of concurrently delivering letters is logical Know that relevant departments are investigated and punished to the object to be audited；

It is no, then judge that the object to be audited is audited unacceptable number in this and continuous several times follow-up auditing Whether have reached preset second number: being then to send the information notice object to be audited to carry out early warning, it is logical to send information Relevant investigation and punishment will be will receive after knowing that the object wait audit is rectified and improved and otherwise reaching first number.

Preferably,

The follow-up auditing judgment step includes:

Substandard obtaining step, for obtaining each single item standard and overall standard in the auditing standard；

Corresponding data extraction step, for extracting each single item standard corresponding described second from first data Data；

Corresponding data changes detecting step, for by each single item standard corresponding second of the object to be audited Corresponding second data of the to be audited each single item standard of object described in when data are audited with the last time are compared and sentence It is disconnected whether to be changed: to be then to execute the corresponding preset model obtaining step of each single item standard or each single item standard judgement step Suddenly；It is no, then it jumps to corresponding data extraction step or the corresponding preset model obtaining step of overall standard executes；

The corresponding preset model obtaining step of each single item standard, for obtaining the corresponding default mould of each single item standard Type；

The corresponding third data generation step of each single item standard, for according to each single item standard corresponding described second Data and the corresponding preset model of each single item standard, are calculated the corresponding third data of each single item standard；

Each single item standard judgment step, for according to the corresponding third data of each standard and preset range and default Range, judges whether the object to be audited meets each single item standard；

The corresponding preset model obtaining step of overall standard, for obtaining the corresponding preset model of the overall standard；

Overall standard judgment step, for according to the corresponding third data of each single item standard, the overall standard pair The preset model and preset range answered, judge whether the object to be audited meets the overall standard；

Comprehensive descision step, for judging whether the object to be audited meets each single item described in the auditing standard Standard and the overall standard.

Preferably,

The corresponding preset model obtaining step of each single item standard includes:

The corresponding deep learning model initialization step of each single item standard, it is corresponding for initializing each single item standard Deep learning model is as the first deep learning model；

The corresponding historical data obtaining step of each single item standard, for obtaining each single item standard from history big data Second data and third data of the corresponding every an object for having carried out audit；

Second deep learning model generation step, for by each single item standard it is corresponding carried out audit it is each Input data of second data of object as the first deep learning model, to the first deep learning model into The unsupervised training of row, obtained the first deep learning model is as the second deep learning model；

Third deep learning model generation step, for by each single item standard it is corresponding carried out audit it is each Input data and output of second data and the third data of object respectively as the second deep learning model Data carry out Training to the second deep learning model, and obtained the second deep learning model is as third Deep learning model；

The corresponding preset model setting steps of each single item standard, for using the third deep learning model as described every The corresponding preset model of one standard；

The corresponding preset model obtaining step of the overall standard includes:

The corresponding deep learning model initialization step of overall standard, for initializing the corresponding depth of the overall standard Learning model, the obtained deep learning model is as the 4th deep learning model；

The corresponding historical data obtaining step of overall standard, for having carried out audit described in the acquisition from history big data Every an object the auditing standard in the corresponding third data of each single item standard set and the overall standard it is corresponding Third data；

5th deep learning model generation step, for having carried out each single item standard in the auditing standard is corresponding Input data of the set of the third data of every an object of audit as the deep learning model, it is deep to the described 4th It spends learning model and carries out unsupervised training, obtained the 4th deep learning model is as the 5th deep learning model；

6th deep learning model generation step, for by it is described carried out audit every an object the audit mark The set of the corresponding third data of each single item standard and the corresponding third data of the overall standard are made respectively in standard For the input data and output data of the 5th deep learning model, supervision instruction has been carried out to the 5th deep learning model Practice, obtained the 5th deep learning model is as the 6th deep learning model；

The corresponding preset model setting steps of overall standard, for using the 6th deep learning model as the totality The corresponding preset model of standard.

Preferably,

The each single item standard judgment step includes:

The corresponding preset range obtaining step of each single item standard, for obtaining the corresponding default model of each single item standard It encloses；

The corresponding third data judgment step of each single item standard, for determining it is described whether the object to be audited meets Each single item standard；

Step is reminded in the corresponding variation of each single item standard, for judge the object to be audited whether meet it is described every Variation in terms of one standard, and change information is sent to the object to be audited；

The overall standard judgment step includes:

The corresponding third data generation step of overall standard, for according to the corresponding third number of each single item standard According to the preset model corresponding with the overall standard, the corresponding third data of the overall standard are calculated；

The corresponding preset range obtaining step of overall standard, for obtaining the corresponding preset range of the overall standard；

The corresponding third data judgment step of overall standard, for judging it is described total whether the object to be audited meets Body standard.

Second aspect, the embodiment of the present invention provide a kind of auditing system, the system comprises:

Auditing standard obtains module, for obtaining the auditing standard of pre-set categories；

Object data obtains module, for obtaining the data of object to be audited；

The corresponding data acquisition module of standard, for obtaining the auditing standard from the data of the object to be audited Corresponding data；

Data variation detection module, for by the corresponding data of the auditing standard and upper one of the object to be audited The corresponding data of the auditing standard of object to be audited described in when secondary audit are compared and judge whether to be changed: It is then to go to the execution of follow-up auditing judgment module；It is no, then it goes to duplication auditing result module and executes；

Follow-up auditing judgment module, for judging whether the corresponding data of the auditing standard meet the auditing standard；

Replicate auditing result module, for using the last auditing result of the object to be audited as described in wait audit Object this auditing result；

As a result feedback module for judging the variation of auditing result, and is sent to the object to be audited；

Warning module, for judging that it is obstructed that the object to be audited is audited in this and continuous several times follow-up auditing Whether the number crossed reaches threshold value: being then to carry out early warning；

Tracing control module every preset time period or reaches default for obtaining preset time period or preset time point Time point just re-executes all modules in the auditing system.

Preferably,

The object data obtains module

Data source obtains module, for obtaining data source；

Object data retrieving module, for the data of the object to be audited to be retrieved and obtained from the data source；

The corresponding data acquisition module of the standard includes:

Data screening module, for filtering out the corresponding number of the auditing standard from the data of the object to be audited According to as the first data；

Data cleansing module, for extracting data corresponding with each single item standard from first data as described every Corresponding second data of one standard；

The follow-up auditing judgment module includes:

Substandard obtains module, for obtaining each single item standard and overall standard in the auditing standard；

Corresponding data extraction module, for extracting each single item standard corresponding described second from first data Data；

Corresponding data change detection module, for by each single item standard corresponding second of the object to be audited Corresponding second data of the to be audited each single item standard of object described in when data are audited with the last time are compared and sentence It is disconnected whether to be changed: to be then to jump to the corresponding preset model of each single item standard to obtain module or the judgement of each single item standard Module executes；It is no, then it jumps to corresponding data extraction module or the corresponding preset model of overall standard obtains module and executes；

The corresponding preset model of each single item standard obtains module, for obtaining the corresponding default mould of each single item standard Type；

The corresponding third data generation module of each single item standard, for according to each single item standard corresponding described second Data and the corresponding preset model of each single item standard, are calculated the corresponding third data of each single item standard；

Each single item standard judgment module, for according to the corresponding third data of each standard and preset range and default Range, judges whether the object to be audited meets each single item standard；

The corresponding preset model of overall standard obtains module, for obtaining the corresponding preset model of the overall standard；

Overall standard judgment module, for according to the corresponding third data of each single item standard, the overall standard pair The preset model and preset range answered, judge whether the object to be audited meets the overall standard；

Comprehensive judgment module, for judging whether the object to be audited meets each single item described in the auditing standard Standard and the overall standard.

The third aspect, the embodiment of the present invention provide a kind of robot system, are respectively configured in the robot just like second The described in any item auditing systems of aspect.

The embodiment of the present invention has the advantage that includes: with beneficial effect

1, the embodiment of the present invention periodically or non-periodically or in real time automatically track to the enterprise for having carried out audit careful Meter, is able to detect that whether object also is compliant with the requirement of audit within the time after audit, so as to update the knot of audit Fruit, so that the result of audit has timeliness, when the result of past audit does not meet the now practical situation of object, to audit Result be updated.And audited under the prior art (such as new high-tech enterprise audits, the talent audits etc.) it is all disposable , object may be no longer complies with the requirement of audit within a period of time after audit passes through, but audit before will not influence As a result, such result that will lead to audit does not have timeliness, that is to say, that the result that past audits not necessarily meets pair As present practical situation, audit can thus be made to lose credible and meaning.

2, the embodiment of the present invention also carries out follow-up auditing to unacceptable object of auditing, and examines once finding that the object meets The object can be actively invited to apply auditing again or directly have already been through audit by the object after the standard of meter, from And make the result of audit more comprehensively.And the object having in the prior art makes great efforts to keep forging ahead, by certain after audit does not pass through Time make great efforts after, although having had reached the standard of audit because before audit failure, object slowly dare not always Application audit again, in some instances it may even be possible to lose the confidence and no longer application audit always because of audit failure before, can thus arrive Lead to the object for having reached auditing standard and be constantly in the state that do not audit because of auditing without application again, to can lead Cause the result of audit not comprehensive.

3, the embodiment of the present invention ability when the object for having passed through audit is no longer complies with the standard of audit in development process Notice object rectified and improved, this aspect can remind enterprise automatically to improve, on the other hand also save object time and Manpower and material resources, as long as because object meets always the standard of audit, it is not necessary to apply again during follow-up auditing Audit, the result that last time audit passes through is also just still effective, without needing to apply again to examine again after a period of time Meter, to also reduce and reduce the workload of Internal Audit Supervisor department.And the result passed through of auditing in the prior art is only certain Period in effectively, need to audit again after expired, this aspect can waste time and the manpower of the object passed through of having audited Material resources apply auditing again, while wasting time and the manpower and material resources of Internal Audit Supervisor's unit, because if it is logical to have audited The data variation for the object crossed is without influence on audit as a result, so just It is not necessary to apply auditing again, so as to cause existing There are at high cost, the low efficiency of audit technique.

4, the embodiment of the present invention is usually constantly carrying out follow-up auditing, come constantly remind certain standards not over or It is overall to be rectified and improved not over the object of audit, so as to avoid object finally cannot by the tragedy situation of audit, so Follow-up auditing is can finally to be can be improved by audit by the sense of access and safety of the audit target in help by the audit target Sense, a possibility that reduction by the audit target to look for relationship to occur with equal illegal activities of playing tricks by audit.And the prior art All it is to audit in the audit time of Internal Audit Supervisor's agencies dictate, the object audited is caused not have pressure usually, do not making great efforts, It is late in audit, it is interim make great efforts it is also of no avail, result in the object that some are audited to audit by quickly and It takes a risk, makes and the illegal activities such as look for relationship, play tricks.

5, the result of follow-up auditing of the embodiment of the present invention can be with the audit of assisted audit expert:

(1) embodiment of the present invention carries out intellectual analysis based on big data combination auditing standard to judge whether object can be examined Be calculated as pre-set categories and carry out auxiliary expert auditing, the workload that audit expert audits can be reduced, improve audit expert into The efficiency of row audit.

(2) embodiment of the present invention is automatically generated for the default mould of audit using deep learning technology based on history big data Type can be further improved the intelligence and accuracy of audit.

(3) embodiment of the present invention can filter out the data for meeting pre-set categories auditing standard by auditing method and system And the speed that audit expert audits can be improved for expert reference of auditing in object in this way, reduces audit audit expert and carries out The workload of audit.

(4) embodiment of the present invention can filter out the number for not meeting pre-set categories auditing standard by auditing method and system According to and object, for audit expert reference, can make in this way audit expert it is stringenter to ineligible data and object Audit, to improve the accuracy rate of audit.

Follow-up auditing method and robot system provided in an embodiment of the present invention based on big data and deep learning, packet It includes: obtaining the auditing standard of pre-set categories, obtain the data of object to be audited, described examine is obtained from the data of the object The corresponding data of meter standard, described in when the corresponding data of the auditing standard of the object wait audit and last audit The corresponding data of the auditing standard of object to be audited are compared and judge whether to be changed: be then go to Track audit judgement step executes；It is no, then it goes to duplication auditing result step and executes.The embodiment of the present invention can be in real time according to right Timely update as real data audit as a result, and reminding object rectified and improved or invited object to apply auditing again.It is above-mentioned Method and system by the follow-up auditing technology based on big data and deep learning, improve the timeliness of audit, objectivity, can Reliability and efficiency, and reduce the cost of audit.

Detailed description of the invention

Fig. 1 is the flow chart for the auditing method that the embodiment of the present invention 1 provides；

Fig. 2 is the flow chart for the follow-up auditing judgment step that the embodiment of the present invention 4 provides；

Fig. 3 is the flow chart for the corresponding preset model obtaining step of each single item standard that the embodiment of the present invention 5 provides；

Fig. 4 is the flow chart for the corresponding preset model obtaining step of overall standard that the embodiment of the present invention 5 provides；

Fig. 5 is the functional block diagram for the auditing system that the embodiment of the present invention 7 provides；

Fig. 6 is the functional block diagram for the follow-up auditing judgment module that the embodiment of the present invention 10 provides；

Fig. 7 is that the corresponding preset model of each single item standard that the embodiment of the present invention 11 provides obtains the principle frame of module Figure；

Fig. 8 is that the corresponding preset model of overall standard that the embodiment of the present invention 11 provides obtains the functional block diagram of module.

Specific embodiment

Below with reference to embodiment of the present invention, technical solution in the embodiment of the present invention is described in detail.

(1) the various combinations that the method in various embodiments of the present invention includes the following steps:

Auditing standard obtaining step S100: the auditing standard of pre-set categories is obtained.

The auditing standard of pre-set categories is exactly used standard when carrying out the audit of pre-set categories to an object, such as To be financially audited to Enterprise Object, project auditing, effectiveness audit, innovation level audit, project auditing, system-based audit, etc. Deng.Described pre-set categories such as finance, benefit etc., for example financial audit of the audit of pre-set categories, the audit of benefit etc..

Object data obtaining step S200: the data of object to be audited are obtained.What object to be audited referred to be exactly to The object of track audit.Described object such as enterprise, individual etc..Such as the funds for the enterprise that needs to audit use whether close rule, again Such as need to audit personal funds using whether close rule, for another example audit one Enterprise Innovation Capability whether reach new high-tech enterprise Preparation etc..Object to be audited includes the object audit and audit passes through, audit and audited not By object, can also include do not carried out audit object.Object to be audited can be the object of application audit, can also To be the object for not applying for audit.When treating the object progress follow-up auditing of audit, application is gone to audit without object to be audited； Regardless of whether object to be audited is applied auditing, the object of audit can be all treated at regular intervals with regard to carrying out follow-up auditing.

Object data obtaining step S200 includes data source obtaining step S210, object data retrieving step S220.

Data source obtaining step S210: data source is obtained.The data source includes data, the third party's offer that object provides Data.The data that object provides refer to the data that the object provides.The data that third party provides include government department, industry association The object data of the departments such as meeting, Department of Intellectual Property storage.The presentation mode of data source includes data retrieval and acquisition interface.Pass through The interface can automatically retrieval and acquisition related data by computer program.Data source is generally online data source, passes through Internet can remotely obtain the data in online data source.The data source may include the multiple numbers for being distributed in different departments According to source.

Object data retrieving step S220: the data of the object are retrieved and obtained from data source.Because in data source It include the data and other data of many objects, if entirety obtains the network transmission for retrieving meeting overspending again Between, so needing first to retrieve the data of the object, then by the data acquisition of the object to locally.When there is multiple data sources When, the data of the object are retrieved from multiple data sources respectively, it is then locally downloading respectively.

The corresponding data acquisition step S300 of standard: the auditing standard is obtained from the data of the object to be audited Corresponding data.

The corresponding data acquisition step S300 of standard includes data screening step S310, data cleansing step S320.

Data screening step S310: according to the auditing standard, the object pair is filtered out from the data of the object The first data answered.The data of the object include various data, and wherein data are different establishes a capital and pre-set categories auditing standard It is related, so needing to retrieve data related with pre-set categories auditing standard from the data of the object, as described right As corresponding first data.Pre-set categories auditing standard has multinomial, and the corresponding data of all standard may be in different data sources In, such as financial data, in the data source that the administration for industry and commerce provides, intellectual property data provides in data source in Department of Intellectual Property, Quality detecting data is in the data source that quality testing department provides, and secure data is in the data source that public security department provides, so more preferably Mode is to obtain the default corresponding relationship in data source between data category and all standard, obtain from the data of the object The corresponding data of each data source are taken, are retrieved from the corresponding data of each data source corresponding with each data source The relevant data of standard.Such as the corresponding standard of data source that Intellectual Property Department provides is intellectual property standard, from knowledge Include the payment data of the intellectual property of this object in the data source that property right department provides, request for data, accept data, authorization Data, wherein the intellectual property of the data of " request for data accepts data, authorization data " these three classifications and pre-set categories audit Standard is related, then three data categories described in the data source of Intellectual Property Department's offer and the knowledge of pre-set categories audit produce Corresponding relationship is established between token standard, as the preset corresponding relationship.

Data cleansing step S320: corresponding second data of each single item standard are extracted from first data.By first Data corresponding with each single item standard are as the second data in data.Whether judge corresponding second data of each single item standard In the presence of: it is no, then the prompting message for lacking corresponding second data of each single item standard is sent to user；It is, then described in judgement Whether corresponding second data of each single item standard unique: it is no, then judge corresponding multiple second data of each single item standard it Between it is whether consistent: it is no, then retain highest second data of confidence level in the multiple second data, delete the multiple second number Other second data in.Corresponding second data of each single item standard extracted from first data are described first Data corresponding with each single item standard in data.

Highest second data of confidence level in the multiple second data of reservation described in data cleansing step S320, are deleted The step of other second data includes corresponding data source obtaining step S321, confidence level obtaining step in the multiple second data S322, confidence level selecting step S323.

Corresponding data source obtaining step S321: the corresponding data of each second data in the multiple second data are obtained Source.Because the first data are to obtain from data source, and the second data are extracted from the first data, it is possible to obtain To the corresponding data source of each second data.

Confidence level obtaining step S322: the confidence level of the corresponding data source of each second data is obtained.It is described credible Degree can be preset.Such as the confidence level of the data source of public security department is 100%, the confidence level of the data source of the administration for industry and commerce is 99%, the confidence level of the data source of Intellectual Property Department is 98%, and the confidence level for the data source that object itself provides is 80%.In advance The mode that the confidence level is first arranged includes being configured by expert to confidence level, further includes automatically generating the confidence level.

The step of automatically generating confidence level described in confidence level obtaining step S322 includes: by the confidence level of all data sources It is initialized as initial value, such as 50%.First data of each object are obtained from history big data and have cleaned to obtain The second data.The confidence level of the corresponding data source of the second data cleaned is increased into preset value, by described the The confidence level of data source corresponding with other consistent second data of the second data cleaned increases default in one data Value, such as 0.1%, it will be corresponding with other inconsistent second data of the second data cleaned in first data The confidence level of data source reduces preset value, such as 0.05%.When the confidence level of the corresponding data source of second data reaches When 100%, then preset value is not further added by；When the confidence level of the corresponding data source of second data is reduced to 0%, then no longer Build preset value.Increased this makes it possible to make the confidence level of different data sources whether correct according to the second data of history Subtract, to form the different confidence levels of different data sources.Second data cleaned refer to passing through manual type Or correct second data of other modes confirmation.As it can be seen that multiple pre- before the confidence level of unclear each data source If classification is audited, artificial mode is needed to carry out the cleaning of data, then can be analyzed according to these historical datas Obtain the confidence level of each data source.

Confidence level selecting step S323: it is chosen from the confidence level of the corresponding data source of each second data highest Confidence level.Retain corresponding second data of highest confidence level described in the multiple second data, deletes the multiple second number Other second data other than corresponding second data of the highest confidence level described in.This makes it possible in multiple second data phases Mutually when conflict, retain most believable data, and other conflicting data are deleted.

Data variation detecting step S400: judge whether the object to be audited was audited before this audit:

It is (the case where being audited), then by the corresponding data of the auditing standard and upper one of the object to be audited The corresponding data of the auditing standard of object to be audited described in when secondary audit are compared and judge whether to be changed: It is then to go to follow-up auditing judgment step S500 execution；It is no, then it goes to duplication auditing result step S600 and executes；

No (the case where not audited) then goes to follow-up auditing judgment step S500 execution.If described to be audited Object was never audited before this audit, then this audit is exactly to audit for the first time, the audit of last time is not present, from It so can not be just compared with the audit of last time, so if this audit is that the object to be audited is examined for the first time Meter, this step then go to follow-up auditing judgment step S500 execution.

Follow-up auditing judgment step S500: judge whether the corresponding data of the auditing standard meet the auditing standard: It is then to judge the audit that the object to be audited passes through pre-set categories；It is no, then it is pre- to judge that the object to be audited does not pass through If the audit of classification.

Audit judgement step S500 includes substandard obtaining step S510, corresponding data extraction step S520, corresponding data Change detecting step S530, the corresponding preset model obtaining step S540 of each single item standard, the corresponding third number of each single item standard According to generation step S550, each single item standard judgment step S560, the corresponding preset model obtaining step S570 of overall standard, totality Standard judgment step S580, comprehensive descision step S590.

Substandard obtaining step S510: each single item standard and overall standard in the auditing standard are obtained.

Corresponding data extraction step S520: corresponding second number of each single item standard is extracted from first data According to.

Corresponding data changes detecting step S530: by each single item standard corresponding second of the object to be audited Corresponding second data of the to be audited each single item standard of object described in when data are audited with the last time are compared and sentence It is disconnected whether to be changed:

It is then to judge that corresponding second data of each single item standard whether there is: is then to jump to S540 to continue to hold Row；It is no, then it sets empty for the corresponding third data of each single item standard, then branches to S560 and continue to execute.

It is no, then the result work for whether meeting each single item standard when the last audit of the object wait audit obtained For the object to be audited this whether meet each single item standard as a result, then do not continue to execute S540, S550,S560.The result whether last time of the object to be audited meets each single item standard includes meeting, not being inconsistent It closes.

The corresponding preset model obtaining step S540 of each single item standard: the corresponding default mould of each single item standard is obtained Type.The preset model includes formula or algorithm or deep learning model.

When the preset model is deep learning model, the corresponding preset model obtaining step S540 packet of each single item standard Include the corresponding deep learning model initialization step S541 of each single item standard, the corresponding historical data obtaining step of each single item standard S542, the second deep learning model generation step S543, third deep learning model generation step S544, each single item standard are corresponding Predetermined deep learning model setting steps S545.

The corresponding deep learning model initialization step S541 of each single item standard: it is corresponding to initialize each single item standard The input format of the deep learning model is set corresponding second data of each single item standard by deep learning model Format sets the output format of the deep learning model to the format of the corresponding third data of each single item standard, leads to The deep learning model for initializing and obtaining is crossed as the first deep learning model.

The corresponding historical data obtaining step S542 of each single item standard: each single item standard is obtained from history big data The second data and third data of the corresponding every an object for having carried out audit.History big data refers to a large amount of historical data Or have accumulated the data of long period.Object to be audited, including carried out audit pass through object, carried out audit but It audits unsanctioned object.Wherein, third data can be each single item standard it is corresponding scoring evaluation result or other It can reflect the numerical value of the degree of each single item standard described in second data fit.Degree can be a percentage, such as 0% to 100%, 0% indicates not meeting completely, and 100% indicates to comply fully with.

Second deep learning model generation step S543: by each single item standard it is corresponding carried out audit it is each Input data of second data of object as the first deep learning model carries out nothing to the first deep learning model Supervised training, the first deep learning model obtained by unsupervised learning is as the second deep learning model.

Third deep learning model generation step S544: by each single item standard it is corresponding carried out audit it is each The second data and third data of object respectively as the second deep learning model input data and output data, to institute It states the second deep learning model and carries out Training.By the corresponding every an object for having carried out audit of each single item standard The second data and third data respectively as the second deep learning model input data and output data, refer to by Second data of the corresponding every an object for having carried out audit of each single item standard are as the second deep learning model Input data, using each single item standard it is corresponding carried out audit every an object third data as described in The output data of second deep learning model, the second deep learning model obtained by supervised learning is as third depth Spend learning model.

The corresponding preset model setting steps S545 of each single item standard: using the third deep learning model as described every The corresponding preset model of one standard.

The corresponding third data generation step S550 of each single item standard: according to corresponding second data of each single item standard The corresponding third data of each single item standard are calculated in preset model corresponding with each single item standard.In computer It is upper to execute the corresponding preset model of each single item standard, using corresponding second data of each single item standard as described each The input of the corresponding preset model of item standard, the output being calculated is as the corresponding third data of each single item standard.It is excellent Selection of land, using corresponding second data of each single item standard as the corresponding third deep learning mould of each single item standard The input of type, the output for the third deep learning model being calculated is as the corresponding third number of each single item standard According to.Wherein, third data can be the corresponding scoring of each single item standard or evaluation result or other can reflect described the The numerical value of the degree of each single item standard described in two data fits.

Each single item standard judgment step S560: according to the corresponding third data of each standard and preset range, judgement Whether the object to be audited meets each single item standard.

Each single item standard judgment step S560 includes the corresponding preset range obtaining step S561 of each single item standard, each single item Step S563 is reminded in the corresponding variation of the corresponding third data judgment step S562 of standard, each single item standard.

The corresponding preset range obtaining step S561 of each single item standard: the corresponding default model of each single item standard is obtained It encloses.The corresponding preset range of different standards is different, has plenty of the standard of hardness, then has fixed range, and some standards are not It is the standard of hardness, then range is set as infinite to just infinite from bearing.If a standard is not the standard of hardness, this standard Corresponding result is all in the preset range.However, whether a standard is that rigid standard all can be to described wait audit Whether object can be had an impact by audit, because can be had an impact to final TOP SCORES, and the corresponding totality of TOP SCORES Standard General can all have a range, be greater than 80 points.

The corresponding third data judgment step S562 of each single item standard: judge the corresponding third data of each single item standard Whether it is empty:

It is then to judge whether the corresponding preset range of each single item standard is infinite to just infinite from bearing: is then to determine The object to be audited meets each single item standard；It is no, then determine that the object to be audited does not meet each single item Standard；

It is no: then to judge the corresponding third data of each single item standard whether in the corresponding default model of each single item standard In enclosing: being then to determine that the object to be audited meets each single item standard；It is no, then determine the object to be audited not Meet each single item standard.

Step S563 is reminded in the corresponding variation of each single item standard: judging that the object to be audited is before this audit It is no to be audited: to be, then variation of the judgement object to be audited in terms of whether meeting each single item standard, and will become Change information and is sent to the object to be audited.

If the object to be audited never was audited before this audit, this audit is exactly for the first time Audit, the audit of last time is not present, can not just be compared naturally with the audit of last time, so if this audit is described Object to be audited is audited for the first time, this step is just not necessarily to continue to execute.

If audited before the object to be audited, relatively and judge the object to be audited at this It is described whether the result and the object wait audit for whether meeting each single item standard when audit meet when auditing last time Whether the result of each single item standard changes: be, then it will be described in the change notification for the result for whether meeting each single item standard Object to be audited, whether the result for determining whether to meet each single item standard when this described audit is that audit passes through: it is no, Then notify that the object to be audited is rectified and improved referring to each single item standard.Whether the result of each single item standard is met Including the result for meeting the result of each single item standard, not meeting each single item standard.Such as the object to be audited Whether meet in this audit each single item standard the result is that the object to be audited meets each single item standard, And the object wait audit last time audit when whether meet each single item standard the result is that the object to be audited The each single item standard is not met, then shows to be changed；Or whether the object wait audit accords in this audit Close each single item standard the result is that the object to be audited does not meet each single item standard, and pair to be audited As whether meet when auditing last time each single item standard the result is that the object to be audited meets each single item mark Standard then shows to be changed.

The corresponding preset model obtaining step S570 of overall standard: the corresponding preset model of the overall standard is obtained.Institute Stating preset model includes formula or algorithm or deep learning model.Because being had been detected by data variation detecting step S400 Variation just executes S500, so that S570 can be just executed, because data are changed, then certainly existing a certain item or a few items Corresponding second data of standard are changed, and are changed to inevitably result in third data, to will affect overall mark Quasi- judgement as a result, so whether no matter whether the object to be audited meets the judging result of each single item standard occur Variation, is all possible to changed, institute to the judging result whether object to be audited meets the overall standard Once then whether being met the object to be audited described with having had been detected by variation in data variation detecting step S400 The judgment step of overall standard is essential.

When the preset model is deep learning model, the corresponding preset model obtaining step S570 of overall standard includes The corresponding deep learning model initialization step S571 of overall standard, the corresponding historical data obtaining step S572 of overall standard, 5th deep learning model generation step S573, the 6th deep learning model generation step S574, overall standard are corresponding default Model setting steps S575:

The corresponding deep learning model initialization step S571 of overall standard: the corresponding depth of the overall standard is initialized The input format of the deep learning model is set the corresponding third of each single item standard in the auditing standard by learning model The output format of the deep learning model is set the corresponding third data of the overall standard by the format of the set of data Format, by initializing the obtained deep learning model as the 4th deep learning model.

The corresponding historical data obtaining step S572 of overall standard: audit had been carried out described in obtaining from history big data Every an object the auditing standard in the corresponding third data of each single item standard set and the overall standard it is corresponding Third data.Wherein, the corresponding third data of each single item standard can be the corresponding scoring of each single item standard or evaluation knot Fruit or other can reflect the numerical value of the degree of each single item standard described in second data fit.The corresponding third of overall standard Data can be the corresponding scoring of the overall standard or evaluation result or other to can reflect each single item standard corresponding The numerical value of the degree of overall standard described in third data fit.

5th deep learning model generation step S573: it had carried out each single item standard in the auditing standard is corresponding Input data of the set of the third data of every an object of audit as the deep learning model, to the 4th depth It practises model and carries out unsupervised training, the 4th deep learning model obtained by unsupervised learning is as the 5th deep learning Model.

6th deep learning model generation step S574: by the audit mark of the every an object for having carried out audit The set of the corresponding third data of each single item standard and the corresponding third data of the overall standard are respectively as described in standard The input data and output data of five deep learning models carry out Training to the 5th deep learning model.By institute State the set and the totality of the third data of the corresponding every an object for having carried out audit of each single item standard in auditing standard For the corresponding third data of standard respectively as the input data and output data of the 5th deep learning model, referring to will be every Input of the third data of the corresponding every an object for having carried out audit of one standard as the 5th deep learning model Data, using the third data of the corresponding every an object for having carried out audit of the overall standard as the 5th depth The output data of learning model, the 5th deep learning model obtained by supervised learning is as the 6th deep learning mould Type.

The corresponding preset model setting steps S575 of overall standard: using the 6th deep learning model as the totality The corresponding preset model of standard.

Overall standard judgment step S580: according to the corresponding third data of each single item standard, the overall standard pair The preset model and preset range answered, judge whether the object to be audited meets the overall standard.

Overall standard judgment step S580 includes the corresponding third data generation step S581 of overall standard, overall standard pair The corresponding third data judgment step S583 of preset range obtaining step S582, overall standard answered.

The corresponding third data generation step S581 of overall standard: according to the corresponding third data of each single item standard and The corresponding preset model of the overall standard, is calculated the corresponding third data of the overall standard.It executes on computers The corresponding preset model of each single item standard, using the corresponding third data of each single item standard as the overall standard pair The input for the preset model answered, the output being calculated is as the corresponding third data of the overall standard.It preferably, will be described Input of the corresponding third data of each single item standard as the corresponding 6th deep learning model of the overall standard calculates The output of obtained the 6th deep learning model is as the corresponding third data of the overall standard.Wherein, each single item mark Quasi- corresponding third data can be the corresponding scoring of each single item standard or evaluation result or other can reflect described the The numerical value of the degree of each single item standard described in two data fits.The corresponding third data of overall standard can be the overall standard It is corresponding scoring evaluation result or other can reflect and totally marked described in the corresponding third data fit of each single item standard The numerical value of quasi- degree.

The corresponding preset range obtaining step S582 of overall standard: the corresponding preset range of the overall standard is obtained.Always The corresponding TOP SCORES of body standard generally can all have a range, be greater than 80 points.

The corresponding third data judgment step S583 of overall standard: whether judge the corresponding third data of the overall standard In the corresponding preset range of the overall standard: being then to judge that the object to be audited meets the overall standard；It is no, Then judge that the object to be audited does not meet the overall standard.

Comprehensive descision step S590: judge whether the object to be audited meets each single item standard in the auditing standard And overall standard: being, then determines audit of the object to be audited by pre-set categories, that is, auditing the result is that described pending The result of this audit of the object of meter is that audit passes through；It is no, then determine that the object to be audited does not pass through examining for pre-set categories Meter, that is, it is auditing the result is that the result of this audit of the object to be audited is that audit does not pass through.Meet the auditing standard Middle each single item standard and overall standard refer to meeting simultaneously each single item standard and overall standard in the auditing standard.If there is A certain item standard or overall standard are not met, then judge that the object to be audited does not pass through the audit of pre-set categories.

Replicate auditing result step S600: using the last auditing result of the object to be audited as described wait audit Object this auditing result.Auditing result includes that audit passes through, auditing does not pass through.Last time refer to this before it is upper Primary audit, last time audit may be to audit for the first time, it is also possible to the audit of traditional approach, it is also possible to follow-up auditing.

As a result feedback step S700: judge whether the object to be audited was audited before this audit: being, then Judge the variation of auditing result, and is sent to the object to be audited；It is no, then by this result audited be sent to it is described to The object of audit.

If the object to be audited never was audited before this audit, this audit is exactly for the first time Audit, the audit of last time is not present, can not just be compared naturally with the auditing result of last time, so if this audit is The object to be audited is audited for the first time, this step only needs the result by this audit to be sent to pair to be audited As.

If audited before the object to be audited, relatively and judge the object to be audited at this Whether the result of audit changes with the object to be audited in the result that last time audits: being, then by the result notice of variation Whether the result of the object to be audited, judgement this audit is that audit passes through:

It is (the case where last time audit does not pass through, this audit passes through), then object to be audited is notified to pass through audit Or notice invites the object to be audited to walk the process that application is audited.Because this is judging automatically for system, and is actually audited Need some administrative procedures in the process, for example, object to be audited unit stamp some application materials, so some classifications Audit need that the object to be audited to be notified to walk the process of application audit；

No (last time audit passes through, this unsanctioned situation of auditing), then notify that the object to be audited is rectified and improved And next follow-up auditing is waited, it is fed back to described to the audit target, and obtained by the result of the audit of the pre-set categories by not It takes incongruent each single item standard or overall standard to feed back to described to the audit target, and sends described in information reminding wait audit pair As being rectified and improved for incongruent each single item standard or overall standard.

Such as this audit the result is that the object to be audited by the audits of pre-set categories, and the knot of last time audit Fruit is the audit that the object to be audited does not pass through pre-set categories, then shows to be changed, and is obstructed from the audit of last time It crosses and becomes this audit and pass through；Or this audit the result is that the object to be audited do not pass through pre-set categories examine Meter, and last time audit the result is that the object to be audited by the audit of pre-set categories, then shows to be changed, be from The audit of last time is not passed through by becoming this audit.It is described whether to meet each single item standard every to the audit target One standard judgment step S560 judged, it is possible to be obtained.Such as one share 6 standards, wherein the 1st, 3 is not inconsistent It closes, then gives the 1st, 3 standard feedback described to the audit target.It is described whether to meet overall standard total to the audit target The disconnected step S580 of body standard judged, it is possible to be obtained.

Warning step S800: it is obstructed to judge that the object to be audited is audited in this and continuous several times follow-up auditing Whether the number crossed reaches threshold value: being then to carry out early warning.Preset first number is obtained, judges that the object to be audited exists Whether unacceptable number of auditing in this and continuous several times follow-up auditing has reached preset first number: being then to send The information notice object to be audited will by relevant investigation and punishment, breath of concurrently delivering letters notify relevant departments to it is described to The object of audit is investigated and is punished；No, then the judgement object to be audited is in this and continuous several times follow-up auditing Whether the middle unacceptable number of audit has reached preset second number: being then to send the information notice object to be audited Early warning is carried out, information is sent and notifies that the object to be audited is rectified and improved, correlation will be will receive after otherwise reaching first number Investigation and punishment.Wherein, first number is greater than second number.The object to be audited this and continuous several times track Unacceptable number of auditing in audit refers to that this audit does not pass through and audits in the continuous several times follow-up auditing before this Unacceptable number, such as: this audit does not pass through, last time audit does not pass through, the audit of upper last time does not pass through, upper audit of upper last time Pass through, then the unacceptable number of continuous audit is 3.It must be spaced continuously, between unacceptable each time of statistics.

Tracing control step S900: obtaining preset time period or preset time point, every preset time period or reaches default Time point just re-executes all steps in the auditing method.

Above-mentioned steps can execute in the big data platforms such as Spark, to accelerate the speed of big data processing.

(2) the various combinations that the system in various embodiments of the present invention comprises the following modules:

Auditing standard obtains module 100 and audits standard obtaining step S100.

It executes object data and obtains the execution of module 200 object data obtaining step S200.

It includes that data source obtains module 210, object data retrieving module 220 that object data, which obtains module 200,.

Data source obtains module 210 and executes data source obtaining step S210.

Object data retrieving module 220 executes object data retrieving step S220.

The corresponding data acquisition module 300 of standard executes the corresponding data acquisition step S300 of standard.

The corresponding data acquisition module 300 of standard includes data screening module 310, data cleansing module 320.

Data screening module 310 executes data screening step S310.

Data cleansing module 320 executes data cleansing step S320.

Data cleansing module 320 includes that corresponding data source obtains module 321, confidence level obtains module 322, confidence level is chosen Module 323.

Corresponding data source obtains module 321 and executes corresponding data source obtaining step S321.

Confidence level obtains module 322 and executes confidence level obtaining step S322.

Confidence level chooses module 323 and executes confidence level selecting step S323.

Data variation detection module 400 executes data variation detecting step S400.

Follow-up auditing judgment module 500 executes follow-up auditing judgment step S500.

Follow-up auditing judgment module 500 includes that substandard obtains module 510, corresponding data extraction module 520, corresponding data The corresponding preset model of change detection module 530, each single item standard obtains module 540, the corresponding third data of each single item standard The corresponding preset model of generation module 550, each single item standard judgment module 560, overall standard obtains module 570, overall standard Judgment module 580, comprehensive judgment module 590.

Substandard obtains module 510 and executes substandard obtaining step S510.

Corresponding data extraction module 520 executes corresponding data extraction step S520.

Corresponding data change detection module 530 executes corresponding data and changes detecting step S530.

The corresponding preset model of each single item standard obtains module 540 and executes the corresponding preset model acquisition step of each single item standard Rapid S540.

The corresponding preset model of each single item standard obtains module 540 including at the beginning of the corresponding deep learning model of each single item standard The corresponding historical data of beginningization module 541, each single item standard obtain module 542, the second deep learning model generation module 543, Third deep learning model generation module 544, the corresponding predetermined deep learning model setup module 545 of each single item standard.

The corresponding deep learning model initialization module 541 of each single item standard executes the corresponding deep learning of each single item standard Model initialization step S541.

The corresponding historical data of each single item standard obtains module 542 and executes the corresponding historical data acquisition step of each single item standard Rapid S542.

Second deep learning model generation module 543 executes the second deep learning model generation step S543.

Third deep learning model generation module 544 executes third deep learning model generation step S544.

The corresponding predetermined deep learning model setup module 545 of each single item standard executes the corresponding default mould of each single item standard Type setting steps S545.

The corresponding third data generation module 550 of each single item standard executes the corresponding third data of each single item standard and generates step Rapid S550.

Each single item standard judgment module 560 executes each single item standard judgment step S560.

Each single item standard judgment module 560 includes that the corresponding preset range of each single item standard obtains module 561, each single item mark Quasi- corresponding third data judgment module 562, the corresponding variation reminding module 563 of each single item standard.

The corresponding preset range of each single item standard obtains module 561 and executes the corresponding preset range acquisition step of each single item standard Rapid S561.

The corresponding third data judgment module 562 of each single item standard executes the corresponding third data judgement step of each single item standard Rapid S562.

The corresponding variation reminding module 563 of each single item standard executes the corresponding variation of each single item standard and reminds step S563.

The corresponding preset model of overall standard obtains module 570 and executes the corresponding preset model obtaining step of overall standard S570。

It includes the corresponding deep learning model initialization of overall standard that the corresponding preset model of overall standard, which obtains module 570, It is deep that the corresponding historical data of module 571, overall standard obtains module 572, the 5th deep learning model generation module the 573, the 6th Spend learning model generation module 574, the corresponding preset model setup module 575 of overall standard:

The corresponding deep learning model initialization module 571 of overall standard executes the corresponding deep learning model of overall standard Initialization step S571.

The corresponding historical data of overall standard obtains module 572 and executes the corresponding historical data obtaining step of overall standard S572。

5th deep learning model generation module 573 executes the 5th deep learning model generation step S573.

6th deep learning model generation module 574 executes the 6th deep learning model generation step S574.

The corresponding preset model setup module 575 of overall standard executes the corresponding preset model setting steps of overall standard S575。

Overall standard judgment module 580 executes overall standard judgment step S580.

Overall standard judgment module 580 is corresponding including the corresponding third data generation module 581 of overall standard, overall standard Preset range obtain module 582, the corresponding third data judgment module 583 of overall standard.

The corresponding third data generation module 581 of overall standard executes the corresponding third data generation step of overall standard S581。

The corresponding preset range of overall standard obtains module 582 and executes the corresponding preset range obtaining step of overall standard S582。

The corresponding third data judgment module 583 of overall standard executes the corresponding third data judgment step of overall standard S583。

Comprehensive judgment module 590 executes comprehensive descision step S590.

It replicates auditing result module 600 and executes duplication auditing result step S600.

As a result 700 implementing result feedback step S700 of feedback module.

Warning module 800 executes warning step S800.

Tracing control module 900 executes tracing control step S900.

It obtains preset time period or preset time point, every preset time period or reaches preset time point and just re-execute institute State all modules in auditing system.

Above-mentioned module can dispose in the big data platforms such as Spark, to accelerate the speed of big data processing.

(3) several embodiments of the invention

Embodiment 1 provides a kind of auditing method, and the auditing method includes auditing standard obtaining step S100, object data The corresponding data acquisition step S300 of obtaining step S200, standard, data variation detecting step S400, follow-up auditing judgment step S500, duplication auditing result step S600, result feedback step S700, warning step S800, tracing control step S900, such as scheme Shown in 1.

Embodiment 2 provides a kind of auditing method, each step including method described in embodiment 1；Wherein, object data obtains Taking step S200 includes data source obtaining step S210, object data retrieving S220, the corresponding data acquisition step S300 of standard Including data screening step S310, data cleansing step S320.

Embodiment 3 provides a kind of auditing method, each step including method described in embodiment 2；Wherein, data cleansing walks Rapid S320 includes corresponding data source obtaining step S321, confidence level obtaining step S322, confidence level selecting step S323.

Embodiment 4 provides a kind of auditing method, each step including method described in embodiment 2；Wherein, audit judgement walks Rapid S500 includes substandard obtaining step S510, corresponding data extraction step S520, corresponding data variation detecting step S530, every It is the corresponding preset model obtaining step S540 of one standard, the corresponding third data generation step S550 of each single item standard, each It is standard judgment step S560, the corresponding preset model obtaining step S570 of overall standard, overall standard judgment step S580, comprehensive Judgment step S590 is closed, as shown in Figure 2.

Embodiment 5 provides a kind of auditing method, each step including method described in embodiment 4；Wherein, each single item standard Corresponding preset model obtaining step S540 includes the corresponding deep learning model initialization step S541 of each single item standard, each The corresponding historical data obtaining step S542 of item standard, the second deep learning model generation step S543, third deep learning mould The corresponding predetermined deep learning model setting steps S545 of type generation step S544, each single item standard, as shown in Figure 3；Overall mark Quasi- corresponding preset model obtaining step S570 includes the corresponding deep learning model initialization step S571 of overall standard, totality The corresponding historical data obtaining step S572 of standard, the 5th deep learning model generation step S573, the 6th deep learning model The corresponding preset model setting steps S575 of generation step S574, overall standard, as shown in Figure 4.

Embodiment 6 provides a kind of auditing method, each step including method described in embodiment 4；Wherein, each single item standard Judgment step S560 includes the corresponding preset range obtaining step S561 of each single item standard, the corresponding third data of each single item standard Judgment step S562；Overall standard judgment step S580 includes the corresponding third data generation step S581 of overall standard, totality The corresponding preset range obtaining step S582 of standard, the corresponding third data judgment step S583 of overall standard.

Embodiment 7 provides a kind of auditing system, and the auditing system includes that auditing standard obtains module 100, object data Obtain module 200, the corresponding data acquisition module 300 of standard, data variation detection module 400, follow-up auditing judgment module 500, auditing result module 600, result feedback module 700, warning module 800, tracing control module 900 are replicated, such as Fig. 5 institute Show.

Embodiment 8 provides a kind of auditing system, each step including system described in embodiment 7；Wherein, object data obtains Modulus block 200 includes that data source obtains module 210, object data retrieving S220, and the corresponding data acquisition module 300 of standard includes Data screening module 310, data cleansing module 320.

Embodiment 9 provides a kind of auditing system, each step including system described in embodiment 8；Wherein, data cleansing mould Block 320 includes that corresponding data source obtains module 321, confidence level obtains module 322, confidence level chooses module 323.

Embodiment 10 provides a kind of auditing system, each step including system described in embodiment 8；Wherein, follow-up auditing Judgment module 500 include substandard obtain module 510, corresponding data extraction module 520, corresponding data change detection module 530, The corresponding preset model of each single item standard obtains module 540, the corresponding third data generation module 550 of each single item standard, each Item standard judgment module 560, the corresponding preset model of overall standard obtain module 570, overall standard judgment module 580, synthesis Judgment module 590, as shown in Figure 6.

Embodiment 11 provides a kind of auditing system, each step including system described in embodiment 10；Wherein, each single item mark Quasi- corresponding preset model obtain module 540 include the corresponding deep learning model initialization module 541 of each single item standard, it is each The corresponding historical data of item standard obtains module 542, the second deep learning model generation module 543, third deep learning model The corresponding predetermined deep learning model setup module 545 of generation module 544, each single item standard, as shown in Figure 7；Overall standard pair It includes the corresponding deep learning model initialization module 571 of overall standard, overall standard pair that the preset model answered, which obtains module 570, The historical data answered obtains module 572, the 5th deep learning model generation module 573, the 6th deep learning model generation module 574, the corresponding preset model setup module 575 of overall standard, as shown in Figure 8.

Embodiment 12 provides a kind of auditing system, each step including system described in embodiment 10；Wherein, each single item mark Quasi- judgment module 560 includes that the corresponding preset range of each single item standard obtains module 561, the corresponding third data of each single item standard Judgment module 562；Overall standard judgment module 580 includes the corresponding third data generation module 581 of overall standard, overall standard Corresponding preset range obtains module 582, the corresponding third data judgment module 583 of overall standard.

Embodiment 13 provides a kind of robot system, is each configured in the robot such as embodiment 7 to embodiment 12 The auditing system.

Method and system in the various embodiments described above can be in computer, server, Cloud Server, supercomputer, machine Device people, embedded device, electronic equipment etc. are upper to be executed and disposes.

The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention Protect range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.

Claims

1. a kind of auditing method, which is characterized in that the described method includes:

Object data obtaining step, for obtaining the data of object to be audited；

The corresponding data acquisition step of standard, it is corresponding for obtaining the auditing standard from the data of the object to be audited Data；

Data variation detecting step, for examining the corresponding data of the auditing standard of the object to be audited with the last time The corresponding data of the auditing standard of object to be audited described in timing are compared and judge whether to be changed: be, Then go to the execution of follow-up auditing judgment step；It is no, then it goes to duplication auditing result step and executes；

Auditing result step is replicated, for using the last auditing result of the object to be audited as pair to be audited This auditing result of elephant；

Tracing control step every preset time period or reaches preset time for obtaining preset time period or preset time point Point just re-executes all steps in the auditing method.

2. auditing method according to claim 1, which is characterized in that the method also includes:

Warning step, for judging that it is unacceptable that the object to be audited is audited in this and continuous several times follow-up auditing Whether number reaches threshold value: being then to carry out early warning.

3. auditing method according to claim 1, which is characterized in that

The object data obtaining step includes:

Data source obtaining step, for obtaining data source；

The corresponding data acquisition step of the standard includes:

Data screening step is made for filtering out the corresponding data of the auditing standard from the data of the object to be audited For the first data；

Data cleansing step, for extracting data corresponding with each single item standard from first data as each single item Corresponding second data of standard.

4. auditing method according to claim 2, which is characterized in that

The result feedback step includes:

Relatively and judge what the object to be audited was audited with the object to be audited in last time in this result audited As a result whether change: being, then object to be audited described in the result notice by variation, judgement this audit the result is that It is no to pass through for audit:

It is then to notify that object to be audited has passed through audit or notice invites the object to be audited to walk the stream that application is audited Journey；

It is no, then it notifies the object to be audited to be rectified and improved and waited next follow-up auditing, the pre-set categories will not be passed through Audit result feed back to it is described to the audit target, and obtain incongruent each single item standard or overall standard feed back to it is described To the audit target, and send described in information reminding to the audit target for incongruent each single item standard or overall standard into Row rectification；

The warning step: obtaining preset first number, the judgement object to be audited this and continuous several times with Whether unacceptable number of auditing in track audit has reached preset first number:

It is then to send information to notify the object to be audited will be by relevant investigation and punishment, breath of concurrently delivering letters notifies phase Pass department is investigated and is punished to the object to be audited；

It is no, then judge the object to be audited audit in this and continuous several times follow-up auditing unacceptable number whether It has reached preset second number: being then to send the information notice object to be audited to carry out early warning, send information and notify institute State the object wait audit rectified and improved otherwise reach first number after will will receive relevant investigation and punishment.

5. auditing method according to claim 3, which is characterized in that

The follow-up auditing judgment step includes:

Corresponding data extraction step, for extracting corresponding second number of each single item standard from first data According to；

Corresponding data changes detecting step, for by corresponding second data of each single item standard of the object to be audited Corresponding second data of the to be audited each single item standard of object described in when auditing with the last time are compared and judge It is no to be changed: to be then to execute the corresponding preset model obtaining step of each single item standard or each single item standard judgment step； It is no, then it jumps to corresponding data extraction step or the corresponding preset model obtaining step of overall standard executes；

The corresponding preset model obtaining step of each single item standard, for obtaining the corresponding preset model of each single item standard；

The corresponding third data generation step of each single item standard, for according to corresponding second data of each single item standard The corresponding third data of each single item standard are calculated in the preset model corresponding with each single item standard；

Each single item standard judgment step, for according to the corresponding third data of each standard and preset range and default model It encloses, judges whether the object to be audited meets each single item standard；

Overall standard judgment step, for corresponding according to the corresponding third data of each single item standard, the overall standard Preset model and preset range, judge whether the object to be audited meets the overall standard；

Comprehensive descision step, for judging whether the object to be audited meets each single item standard described in the auditing standard With the overall standard.

6. auditing method according to claim 5, which is characterized in that

The corresponding deep learning model initialization step of each single item standard, for initializing the corresponding depth of each single item standard Learning model is as the first deep learning model；

The corresponding historical data obtaining step of each single item standard, it is corresponding for obtaining each single item standard from history big data Carried out audit every an object second data and third data；

Second deep learning model generation step, for by each single item standard it is corresponding carried out audit every an object Input data of second data as the first deep learning model, nothing is carried out to the first deep learning model Supervised training, obtained the first deep learning model is as the second deep learning model；

Third deep learning model generation step, for by each single item standard it is corresponding carried out audit every an object Second data and the third data respectively as the second deep learning model input data and output data, Training is carried out to the second deep learning model, obtained the second deep learning model is as third depth Practise model；

The corresponding preset model setting steps of each single item standard, for using the third deep learning model as each single item The corresponding preset model of standard；

The corresponding preset model obtaining step of the overall standard includes:

The corresponding deep learning model initialization step of overall standard, for initializing the corresponding deep learning of the overall standard Model, the obtained deep learning model is as the 4th deep learning model；

The corresponding historical data obtaining step of overall standard, for having carried out the every of audit described in the acquisition from history big data The set and the corresponding third of the overall standard of the corresponding third data of each single item standard in the auditing standard of an object Data；

5th deep learning model generation step, for having carried out audit for each single item standard in the auditing standard is corresponding Every an object the third data input data of the set as the deep learning model, to the 4th depth It practises model and carries out unsupervised training, obtained the 4th deep learning model is as the 5th deep learning model；

6th deep learning model generation step, for that described will carry out in the auditing standard for the every an object audited The set and the corresponding third data of the overall standard of the corresponding third data of each single item standard are respectively as institute The input data and output data for stating the 5th deep learning model carry out Training to the 5th deep learning model, Obtained the 5th deep learning model is as the 6th deep learning model；

The corresponding preset model setting steps of overall standard, for using the 6th deep learning model as the overall standard Corresponding preset model.

7. auditing method according to claim 5, which is characterized in that

The each single item standard judgment step includes:

The corresponding preset range obtaining step of each single item standard, for obtaining the corresponding preset range of each single item standard；

The corresponding third data judgment step of each single item standard, for determining it is described each whether the object to be audited meets Item standard；

Step is reminded in the corresponding variation of each single item standard, for judging whether the object to be audited is meeting each single item Variation in terms of standard, and change information is sent to the object to be audited；

The overall standard judgment step includes:

The corresponding third data generation step of overall standard, for according to the corresponding third data of each single item standard and The corresponding preset model of the overall standard, is calculated the corresponding third data of the overall standard；

The corresponding third data judgment step of overall standard, for judging whether the object to be audited meets the overall mark It is quasi-.

8. a kind of auditing system, which is characterized in that the system comprises:

Object data obtains module, for obtaining the data of object to be audited；

The corresponding data acquisition module of standard, it is corresponding for obtaining the auditing standard from the data of the object to be audited Data；

Data variation detection module, for examining the corresponding data of the auditing standard of the object to be audited with the last time The corresponding data of the auditing standard of object to be audited described in timing are compared and judge whether to be changed: be, Then go to the execution of follow-up auditing judgment module；It is no, then it goes to duplication auditing result module and executes；

Auditing result module is replicated, for using the last auditing result of the object to be audited as pair to be audited This auditing result of elephant；

Warning module, for judging that it is unacceptable that the object to be audited is audited in this and continuous several times follow-up auditing Whether number reaches threshold value: being then to carry out early warning；

Tracing control module every preset time period or reaches preset time for obtaining preset time period or preset time point Point just re-executes all modules in the auditing system.

9. auditing system according to claim 8, which is characterized in that

The object data obtains module

Data source obtains module, for obtaining data source；

The corresponding data acquisition module of the standard includes:

Data screening module is made for filtering out the corresponding data of the auditing standard from the data of the object to be audited For the first data；

Data cleansing module, for extracting data corresponding with each single item standard from first data as each single item Corresponding second data of standard；

The follow-up auditing judgment module includes:

Corresponding data extraction module, for extracting corresponding second number of each single item standard from first data According to；

Corresponding data change detection module, for by corresponding second data of each single item standard of the object to be audited Corresponding second data of the to be audited each single item standard of object described in when auditing with the last time are compared and judge It is no to be changed: to be then to jump to the corresponding preset model of each single item standard to obtain module or each single item standard judgment module It executes；It is no, then it jumps to corresponding data extraction module or the corresponding preset model of overall standard obtains module and executes；

The corresponding preset model of each single item standard obtains module, for obtaining the corresponding preset model of each single item standard；

The corresponding third data generation module of each single item standard, for according to corresponding second data of each single item standard The corresponding third data of each single item standard are calculated in the preset model corresponding with each single item standard；

Each single item standard judgment module, for according to the corresponding third data of each standard and preset range and default model It encloses, judges whether the object to be audited meets each single item standard；

Overall standard judgment module, for corresponding according to the corresponding third data of each single item standard, the overall standard Preset model and preset range, judge whether the object to be audited meets the overall standard；

Comprehensive judgment module, for judging whether the object to be audited meets each single item standard described in the auditing standard With the overall standard.

10. a kind of robot system, which is characterized in that be respectively configured in the robot just like any one of claim 8-9 institute The auditing system stated.