CN109409717B

CN109409717B - Active auditing method based on big data and deep learning and robot system

Info

Publication number: CN109409717B
Application number: CN201811207163.1A
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: 2021-06-25
Anticipated expiration: 2038-10-17
Also published as: CN109409717A

Abstract

The active auditing method and the robot system based on big data and deep learning comprise the following steps: the method comprises the steps of obtaining an audit standard of a preset category, obtaining data of an object which does not apply for audit, obtaining the data corresponding to the audit standard from the data of the object, judging whether the data corresponding to the audit standard meet the audit standard or not, and feeding back the result of audit to the object which does not apply for audit and does not pass the audit. The method and the system improve the timeliness of auditing and the timeliness of problem solving through the active auditing technology based on big data and deep learning, and avoid audit omission.

Description

Active auditing method based on big data and deep learning and robot system

Technical Field

The invention relates to the technical field of information, in particular to an active auditing method and a robot system based on big data and deep learning.

Background

In the process of implementing the invention, the inventor finds that at least the following problems exist in the prior art: in the prior art, auditing is carried out by an object (the object comprises an enterprise and the like) to apply for auditing (such as high-new enterprise auditing), and if the object does not apply for auditing, even if the object does not meet the auditing standard, the object cannot find problems through auditing; the subject will not apply for an audit in several cases: (1) the flow or standard of the audit by the object is not clear, and how to apply for the audit is not known: (2) the object subjectively overestimates the object, considers that the object can meet the auditing standard, and does not need to audit; (3) the object does not know the existence of the audit of the preset category at all; some subjects do not apply for an audit until long after they have not met the audit standard (e.g., 5 years ago, but 5 years later). No matter which kind of circumstances, all can lead to a very serious problem, that is "the hole is not mended, and the hole is taken five to the first time", if through early audit early discovery problem and in time rectification, then the problem will be solved, otherwise the small problem will become big problem long after long, finally reach unable rectification, illegal degree, even audit has come out, for the time late. Meanwhile, another situation exists, namely some objects clearly know that the objects can not reach the auditing standard, but intentionally escape the auditing and do not apply for the auditing, so that the missed fishes are formed, and the loss is caused to the country and the society. Therefore, the auditing in the prior art has the defects of untimely auditing, incapability of finding problems and missing auditing and the like.

Accordingly, the prior art is yet to be improved and developed.

Disclosure of Invention

Based on the above, it is necessary to provide an active auditing method and a robot system based on big data and deep learning aiming at the defects or shortcomings in the prior art, so as to solve the defects that auditing is not timely, problems cannot be found timely, and auditing is missed in the prior art.

In a first aspect, an embodiment of the present invention provides an auditing method, where the method includes:

an audit standard obtaining step, which is used for obtaining the audit standard of the preset category;

an object data acquisition step, which is used for acquiring the data of the object which does not apply for auditing; and data acquisition and audit are carried out on the object which does not apply for audit, and the acquired data can come from a third party, so that even if the object which does not apply for audit does not provide data, audit cannot be avoided.

A data acquisition step corresponding to a standard, which is used for acquiring data corresponding to the audit standard from the data of the object which does not apply for audit;

an audit judging step, which is used for judging whether the data corresponding to the audit standard meets the audit standard: if yes, the object which does not apply for auditing audits passes auditing; if not, the object which does not apply for auditing does not pass the auditing; and actively auditing the object which does not apply for auditing so that the object which does not apply for auditing can not escape auditing.

And actively feeding back the result of the audit to the object which does not apply for the audit and does not pass the audit. And the result of the audit is fed back to the object which does not pass the audit and does not apply for the audit, so that the object which does not apply for the audit can find the problem in time, the problem can be solved in time, and the timeliness of the audit is improved.

Preferably, the first and second electrodes are formed of a metal,

the object data acquiring step includes:

a data source obtaining step for obtaining a data source;

an object data retrieval step, which is used for retrieving and acquiring the data of the object which does not apply for auditing from the data source;

the data acquisition step corresponding to the standard comprises the following steps:

a data screening step, which is used for screening out data corresponding to the auditing standard from the data of the object which does not apply for auditing as first data;

and a data cleaning step, which is used for extracting data corresponding to each standard from the first data as second data corresponding to each standard.

Preferably, the first and second electrodes are formed of a metal,

the active feedback step comprises:

judging whether the audit result of the preset category passes the audit: and if not, sending the result of the audit failure to the object which does not apply for the audit, obtaining each non-conforming standard or overall standard and feeding back to the object which does not apply for the audit, and sending information to remind the object which does not apply for the audit to modify each non-conforming standard or overall standard.

Preferably, the first and second electrodes are formed of a metal,

the auditing judgment step comprises:

a sub-standard obtaining step, which is used for obtaining each standard and the overall standard in the auditing standard;

a corresponding data extracting step, configured to extract the second data corresponding to each criterion from the first data;

a preset model obtaining step corresponding to each standard, which is used for obtaining a preset model corresponding to each standard;

a third data generation step corresponding to each standard, wherein the third data generation step is used for calculating to obtain third data corresponding to each standard according to the second data corresponding to each standard and the preset model corresponding to each standard;

each standard judging step is used for judging whether the object which does not apply for auditing meets each standard or not according to third data and a preset range corresponding to each standard;

a preset model obtaining step corresponding to the overall standard, which is used for obtaining a preset model corresponding to the overall standard;

a total standard judging step, configured to judge whether the object for which audit is not applied conforms to the total standard according to the third data corresponding to each standard, the preset model corresponding to the total standard, and the preset range;

and a comprehensive judgment step for judging whether the object which does not apply for auditing meets each standard and the overall standard in the auditing standard.

Preferably, the first and second electrodes are formed of a metal,

the preset model acquisition step corresponding to each standard comprises the following steps:

a deep learning model initialization step corresponding to each standard, which is used for initializing the deep learning model corresponding to each standard as a first deep learning model;

a historical data acquisition step corresponding to each standard, which is used for acquiring the second data and the third data of each object which is subjected to auditing and corresponds to each standard from historical big data;

a second deep learning model generation step, configured to use the second data of each object, which is subjected to auditing and corresponds to each standard, as input data of the first deep learning model, perform unsupervised training on the first deep learning model, and use the obtained first deep learning model as a second deep learning model;

a third deep learning model generation step, configured to respectively use the second data and the third data of each object, which is subjected to auditing and corresponds to each standard, as input data and output data of the second deep learning model, perform supervised training on the second deep learning model, and use the obtained second deep learning model as a third deep learning model;

a preset model setting step corresponding to each standard, which is used for taking the third deep learning model as a preset model corresponding to each standard;

the step of obtaining the preset model corresponding to the overall standard comprises the following steps:

a deep learning model initialization step corresponding to an overall standard, which is used for initializing the deep learning model corresponding to the overall standard, and taking the obtained deep learning model as a fourth deep learning model;

a historical data acquisition step corresponding to an overall standard, which is used for acquiring a set of third data corresponding to each standard in the auditing standard of each object which is audited and third data corresponding to the overall standard from historical big data;

a fifth deep learning model generation step, configured to use the set of the third data of each object that has undergone audit and corresponds to each standard in the audit standards as input data of the deep learning model, perform unsupervised training on the fourth deep learning model, and use the obtained fourth deep learning model as a fifth deep learning model;

a sixth deep learning model generation step of taking the set of the third data corresponding to each standard in the audit standard of each object that has been audited and the third data corresponding to the overall standard as input data and output data of the fifth deep learning model, respectively, performing supervised training on the fifth deep learning model, and taking the obtained fifth deep learning model as a sixth deep learning model;

and a preset model setting step corresponding to the overall standard, wherein the preset model setting step is used for taking the sixth deep learning model as a preset model corresponding to the overall standard.

Preferably, the first and second electrodes are formed of a metal,

each standard judging step comprises the following steps:

a preset range acquisition step corresponding to each standard, which is used for acquiring the preset range corresponding to each standard;

a third data judgment step corresponding to each standard, which is used for judging whether the object which does not apply for auditing meets each standard;

the overall standard judging step comprises the following steps:

a third data generation step corresponding to the overall standard, configured to calculate third data corresponding to the overall standard according to the third data corresponding to each standard and the preset model corresponding to the overall standard;

a preset range acquisition step corresponding to the overall standard, which is used for acquiring the preset range corresponding to the overall standard;

and a third data judgment step corresponding to the overall standard, which is used for judging whether the object which does not apply for auditing meets the overall standard.

In a second aspect, an embodiment of the present invention provides an auditing system, where the system includes:

the audit standard acquisition module is used for acquiring the audit standard of a preset category;

the object data acquisition module is used for acquiring data of an object which does not apply for auditing;

the data acquisition module corresponding to the standard is used for acquiring data corresponding to the audit standard from the data of the object which does not apply for the audit;

the audit judging module is used for judging whether the data corresponding to the audit standard meets the audit standard: if yes, the object which does not apply for auditing audits passes auditing; if not, the object which does not apply for auditing does not pass the auditing;

and the active feedback module is used for feeding back the auditing result to the object which does not apply for auditing and does not pass the auditing.

Preferably, the first and second electrodes are formed of a metal,

the object data acquisition module includes:

the data source acquisition module is used for acquiring a data source;

the object data retrieval module is used for retrieving and acquiring the data of the object which does not apply for auditing from the data source;

the data acquisition module corresponding to the standard comprises:

the data screening module is used for screening out data corresponding to the audit standard from the data of the object which does not apply for audit as first data;

and the data cleaning module is used for extracting data corresponding to each standard from the first data as second data corresponding to each standard.

The active feedback module comprises:

Preferably, the first and second electrodes are formed of a metal,

the audit judgment module comprises:

the sub-standard acquisition module is used for acquiring each standard and the overall standard in the audit standard;

a corresponding data extraction module, configured to extract the second data corresponding to each criterion from the first data;

the preset model acquisition module corresponding to each standard is used for acquiring a preset model corresponding to each standard;

the third data generation module corresponding to each standard is used for calculating to obtain third data corresponding to each standard according to the second data corresponding to each standard and the preset model corresponding to each standard;

each standard judging module is used for judging whether the object which does not apply for auditing meets each standard or not according to the third data and the preset range corresponding to each standard;

the preset model acquisition module corresponding to the overall standard is used for acquiring a preset model corresponding to the overall standard;

the overall standard judging module is used for judging whether the object which does not apply for auditing meets the overall standard or not according to the third data corresponding to each standard, the preset model corresponding to the overall standard and the preset range;

and the comprehensive judgment module is used for judging whether the object which does not apply for auditing meets each standard and the overall standard in the auditing standard.

In a third aspect, an embodiment of the present invention provides a robot system, where the auditing systems of the second aspect are respectively configured in the robots.

The embodiment of the invention has the advantages and beneficial effects that:

1. the embodiment of the invention actively audits the object, can actively audit even if the object does not apply for audit, can timely find the object which does not reach the audit standard, and informs the object which does not reach the audit standard to carry out rectification, thereby enabling the object which does not reach the audit standard to be audited in time, further improving the timeliness of audit, further finding the problem as early as possible, solving the problem as early as possible, preventing and eliminating the problem in sprouting, and further enabling the object to benefit through audit. However, in the prior art, auditing (for example, financial auditing) is applied by an object (the object includes an enterprise and the like) which does not apply for auditing, if the object does not apply for auditing, even if the object does not reach the auditing standard, the object does not have a chance to find problems through auditing, no matter which situation, a very serious problem can be caused, namely' small hole is not supplemented, big hole is five, if the problem is found early through auditing in the morning and is rectified in time, the problem can be solved, otherwise, the small problem is accumulated in the morning and becomes a big problem, finally, the irreconcilable and illegal degree is reached, even if the auditing is brought out, the time is late.

2. The embodiment of the invention actively audits the object, and can also actively audit even if the object does not apply for auditing, so that the object which deliberately evades auditing can not become a missing fish, and further the loss of the object to the country and the society is reduced. Some objects in the prior art clearly know that the objects can not reach the auditing standard, but intentionally escape the auditing and do not apply for the auditing, so that the missed fishes are formed, and the loss is caused to the country and the society. Therefore, the auditing in the prior art has the defects of untimely auditing, incapability of finding problems and missing auditing and the like.

3. The embodiment of the invention also has the following advantages and beneficial effects in active audit:

(1) the embodiment of the invention can be used for full-automatic audit of the object in the active audit process or assisting the expert in evaluating and carrying out semi-automatic audit of the object, thereby improving the automation, the intellectualization and the audit efficiency of the active audit.

(2) Various data of the object can be involved in the process of the preset type audit, for example, financial data, intellectual property data, product data, safety data, quality data and the like of the enterprise object during the process of the high and new enterprise audit, the data can be acquired by third parties such as an industrial and commercial department, an intellectual property bureau, a tax department, a public security department, a quality inspection department and the like, but the data of the third parties are not fully utilized to improve the credibility of the audit during the real-time preset type audit. The data source utilized by the embodiment of the invention comprises big data acquired from a third party, which is more credible than data provided by the object.

(3) The embodiment of the invention judges whether the object can be audited to the preset category by intelligently analyzing based on the big data and the audit standard so as to assist the auditing of the expert, thereby reducing the workload of auditing by the auditing expert and improving the efficiency of auditing by the auditing expert.

(4) The embodiment of the invention adopts a deep learning technology to automatically generate the preset model for auditing based on historical big data, and can further improve the intelligence and accuracy of auditing.

(5) According to the embodiment of the invention, the data and the objects which meet the preset type audit standard can be screened out through the audit method and the audit system for reference of the audit expert, so that the audit speed of the audit expert can be improved, and the workload of the audit expert for auditing is reduced.

(6) According to the embodiment of the invention, the data and the object which do not accord with the preset type audit standard can be screened out through the audit method and the audit system to be referred by the audit expert, so that the audit expert can carry out more strict audit on the data and the object which do not accord with the conditions, and the audit accuracy is improved.

The embodiment of the invention provides an active auditing method and a robot system based on big data and deep learning, which comprises the following steps: the method comprises the steps of obtaining an audit standard of a preset category, obtaining data of an object which does not apply for audit, obtaining the data corresponding to the audit standard from the data of the object, judging whether the data corresponding to the audit standard meet the audit standard or not, and feeding back the result of audit to the object which does not apply for audit and does not pass the audit. The method and the system improve the timeliness of auditing and the timeliness of problem solving through the active auditing technology based on big data and deep learning, and avoid audit omission.

Drawings

FIG. 1 is a flow chart of an auditing method provided by embodiment 1 of the present invention;

FIG. 2 is a flowchart of audit determination steps provided in embodiment 4 of the present invention;

fig. 3 is a flowchart of a preset model obtaining step corresponding to each criterion according to embodiment 5 of the present invention;

fig. 4 is a flowchart of a preset model obtaining step corresponding to the overall standard provided in embodiment 5 of the present invention;

FIG. 5 is a functional block diagram of an audit system provided by embodiment 7 of the present invention;

FIG. 6 is a functional block diagram of an audit determination module provided in embodiment 10 of the present invention;

fig. 7 is a schematic block diagram of a preset model obtaining module corresponding to each standard according to embodiment 11 of the present invention;

fig. 8 is a schematic block diagram of a preset model obtaining module corresponding to the overall standard according to embodiment 11 of the present invention.

Detailed Description

The technical solutions in the examples of the present invention are described in detail below with reference to the embodiments of the present invention.

The methods in various embodiments of the present invention include various combinations of the following steps:

an audit standard acquisition step S100: and acquiring the audit standard of a preset category.

The preset type audit standard is a standard used for carrying out preset type audit on an object, such as financial audit, project audit, benefit audit, innovation level audit, project audit, system audit and the like on an enterprise object. The preset categories are financial, benefit and the like, and the auditing of the preset categories is financial auditing, benefit auditing and the like.

Object data acquisition step S200: and acquiring data of the object which does not apply for auditing. Such as a business, an individual, etc. For example, it is necessary to audit whether the financial use of an enterprise is in compliance, whether an individual financial use is in compliance, whether an enterprise innovation capability is in preparation for a new enterprise, and the like.

The object data acquisition step S200 includes a data source acquisition step S210 and an object data retrieval step S220.

Data source acquisition step S210: a data source is acquired. The data source comprises data provided by an object and data provided by a third party. Object provided data refers to data provided by the object. The data provided by the third party comprises object data stored by government departments, industry associations, intellectual property offices and the like. The data source is provided by a data retrieval and acquisition interface. The relevant data can be retrieved and acquired automatically by the computer program through the interface. The data source is generally an online data source, and data in the online data source can be remotely acquired through the internet. The data source may include a plurality of data sources distributed in different departments.

Object data retrieval step S220: data for the object is retrieved and acquired from a data source. Because the data source contains many object data and other data, if the data source is obtained as a whole and then retrieved, it will take too much network transmission time, so the object data needs to be retrieved first, and then the object data needs to be obtained locally. When there are multiple data sources, the data of the object is retrieved from the multiple data sources respectively and then downloaded to the local.

Data acquisition step S300 corresponding to the standard: and acquiring data corresponding to the audit standard from the data of the object.

The data acquiring step S300 corresponding to the standard includes a data filtering step S310 and a data cleaning step S320.

Data filtering step S310: and screening out first data corresponding to the object from the data of the object according to the auditing standard. The data of the object includes various data, wherein the data is not necessarily related to the preset category audit standard, so that the data related to the preset category audit standard needs to be retrieved from the data of the object as the first data corresponding to the object. The preset type audit standard has a plurality of items, data corresponding to each item of standard may be in different data sources, for example, financial data is in a data source provided by an industrial and commercial department, intellectual property data is in a data source provided by an intellectual property office, quality inspection data is in a data source provided by a quality inspection department, and safety data is in a data source provided by a public security department. For example, the data source provided by the intellectual property department corresponds to an intellectual property standard, the data source provided by the intellectual property department includes payment data, application data, acceptance data and authorization data of the intellectual property of the object, wherein the three categories of data of the application data, the acceptance data and the authorization data are related to the intellectual property standard audited in a preset category, and then the corresponding relationship is established between the three categories of data and the intellectual property standard audited in the preset category in the data source provided by the intellectual property department as the preset corresponding relationship.

Data cleaning step S320: and extracting second data corresponding to each standard from the first data. And taking the data corresponding to each standard in the first data as second data. Judging whether second data corresponding to each standard exists or not: if not, sending reminding information of the second data corresponding to each item of standard which is lacked to the user; if yes, judging whether the second data corresponding to each standard is unique: if not, judging whether the plurality of second data corresponding to each standard are consistent: and if not, retaining the second data with the highest credibility in the plurality of second data, and deleting other second data in the plurality of second data. The second data corresponding to the each criterion extracted from the first data is the data corresponding to the each criterion in the first data.

In the data cleaning step S320, the step of retaining the second data with the highest reliability in the plurality of second data, and the step of deleting other second data in the plurality of second data include a corresponding data source obtaining step S321, a reliability obtaining step S322, and a reliability selecting step S323.

Corresponding data source acquisition step S321: and acquiring a data source corresponding to each second data in the plurality of second data. Because the first data is obtained from the data source and the second data is extracted from the first data, the data source corresponding to each second data can be obtained.

Confidence level acquisition step S322: and acquiring the credibility of the data source corresponding to each second data. The confidence level may be preset. For example, the credibility of the data source of the police department is 100%, the credibility of the data source of the industrial and commercial department is 99%, the credibility of the data source of the intellectual property department is 98%, and the credibility of the data source provided by the object itself is 80%. The method for presetting the credibility comprises the steps of setting the credibility by an expert and automatically generating the credibility.

The step of automatically generating the reliability in the reliability obtaining step S322 includes: the trustworthiness of all data sources is initialized to an initial value, e.g. 50%. And acquiring the first data and the cleaned second data of each object from the historical big data. Increasing the credibility of the data source corresponding to the cleaned second data by a preset value, increasing the credibility of the data source corresponding to other second data in the first data, which is consistent with the cleaned second data, by a preset value, for example, 0.1%, and reducing the credibility of the data source corresponding to other second data in the first data, which is inconsistent with the cleaned second data, by a preset value, for example, 0.05%. When the credibility of the data source corresponding to the second data reaches 100%, no preset value is added; and when the credibility of the data source corresponding to the second data is reduced to 0%, no preset value is built. Therefore, the credibility of different data sources can be increased or decreased according to whether the historical second data is correct or not, and different credibility of different data sources is formed. The second data obtained by cleaning refers to correct second data confirmed by manual mode or other modes. Therefore, when the preset category is checked for many times before the credibility of each data source is not clear, the data needs to be cleaned manually, and then the credibility of each data source can be analyzed and obtained according to the historical data.

Reliability selecting step S323: and selecting the highest credibility from the credibility of the data source corresponding to each second data. And reserving second data corresponding to the highest reliability in the plurality of second data, and deleting other second data except the second data corresponding to the highest reliability in the plurality of second data. This makes it possible to retain the most reliable data and delete the other conflicting data when a plurality of second data conflicts with each other.

An audit judgment step S400: judging whether the data corresponding to the audit standard meets the audit standard: if yes, judging that the object which does not apply for auditing passes the auditing of the preset category; and if not, judging that the object which does not apply for auditing does not pass the auditing of the preset category.

The audit judgment step S400 includes a sub-standard acquisition step S410, a corresponding data extraction step S420, a preset model acquisition step S430 corresponding to each standard, a judgment step S440 for each standard, a preset model acquisition step S450 corresponding to an overall standard, an overall standard judgment step S460, and a comprehensive judgment step S470.

Sub-standard acquisition step S410: and acquiring each standard and the overall standard in the audit standard.

Corresponding data extraction step S420: and extracting second data corresponding to each standard from the first data. Judging whether second data corresponding to each standard exists or not: if yes, jumping to S430 to continue execution; otherwise, setting the third data corresponding to each standard to be null, and then jumping to S450 to continue execution.

A preset model obtaining step S430 corresponding to each standard: and acquiring a preset model corresponding to each standard. The preset model comprises a formula or an algorithm or a deep learning model.

When the preset model is a deep learning model, the preset model obtaining step S430 corresponding to each standard includes a deep learning model initializing step S431 corresponding to each standard, a historical data obtaining step S432 corresponding to each standard, a second deep learning model generating step S433, a third deep learning model generating step S434, and a preset deep learning model setting step S435 corresponding to each standard.

Initializing the deep learning model corresponding to each standard, setting the input format of the deep learning model as the format of the second data corresponding to each standard, setting the output format of the deep learning model as the format of the third data corresponding to each standard, and taking the deep learning model obtained through initialization as the first deep learning model.

A historical data acquisition step S432 corresponding to each criterion: and acquiring second data and third data of each object which is subjected to auditing and corresponds to each standard from historical big data. Historical big data refers to a large amount of historical data or data accumulated for a long time. The audited objects comprise objects which have been audited to pass and objects which have been audited but have not passed. The third data may be a score or an evaluation result corresponding to each criterion or other numerical values reflecting the degree of the second data meeting each criterion.

And a second deep learning model generation step S433, namely, taking second data of each object which is subjected to auditing and corresponds to each standard as input data of the first deep learning model, carrying out unsupervised training on the first deep learning model, and taking the first deep learning model obtained through unsupervised training as the second deep learning model.

And a third deep learning model generation step S434, using the second data and the third data of each object which is subjected to auditing and corresponds to each standard as input data and output data of the second deep learning model respectively, and performing supervised training on the second deep learning model. And respectively taking the second data and the third data of each object which is subjected to auditing and corresponds to each standard as input data and output data of the second deep learning model, namely taking the second data of each object which is subjected to auditing and corresponds to each standard as the input data of the second deep learning model, taking the third data of each object which is subjected to auditing and corresponds to each standard as the output data of the second deep learning model, and taking the second deep learning model obtained through supervised training as the third deep learning model.

A preset model setting step S435 corresponding to each standard: and taking the third deep learning model as a preset model corresponding to each standard.

And a third data generation step S440 corresponding to each standard, wherein the third data corresponding to each standard is obtained through calculation according to the second data corresponding to each standard and the preset model corresponding to each standard. Executing the preset model corresponding to each standard on a computer, taking the second data corresponding to each standard as the input of the preset model corresponding to each standard, and taking the calculated output as the third data corresponding to each standard. Preferably, the second data corresponding to each criterion is used as the input of the third deep learning model corresponding to each criterion, and the calculated output of the third deep learning model is used as the third data corresponding to each criterion. The third data may be a score or an evaluation result corresponding to each criterion or other numerical values reflecting the degree of the second data meeting each criterion.

And each standard judging step S450, judging whether the object which does not apply for auditing meets each standard or not according to the third data and the preset range corresponding to each standard.

Each criterion determining step S450 includes a preset range acquiring step S451 corresponding to each criterion, and a third data determining step S452 corresponding to each criterion.

A preset range acquiring step S451 corresponding to each standard, acquiring a preset range corresponding to each standard. The preset ranges corresponding to different standards are different, some standards are hard standards, and have fixed ranges, and some standards are not hard standards, and the ranges are set from negative infinity to positive infinity. If a criterion is not a hard criterion, the result corresponding to this criterion is within the preset range. However, whether a standard is a hard standard or not will affect whether the subject who does not apply for auditing can pass the auditing because the final overall score will be affected, and the overall standard corresponding to the overall score will generally have a range, such as more than 80 points.

A third data judgment step S452 corresponding to each criterion is to judge whether the third data corresponding to each criterion is empty:

if yes, judging whether the preset range corresponding to each standard is from negative infinity to positive infinity: if yes, judging that the object which does not apply for auditing meets each standard; if not, judging that the object which does not apply for auditing does not accord with each standard;

otherwise: judging whether the third data corresponding to each standard is in a preset range corresponding to each standard: if yes, judging that the object which does not apply for auditing meets each standard; and if not, judging that the object which does not apply for auditing does not meet each standard.

A preset model obtaining step S460 corresponding to the overall standard: and acquiring a preset model corresponding to the overall standard. The preset model comprises a formula or an algorithm or a deep learning model.

When the preset model is a deep learning model, the step S460 of obtaining the preset model corresponding to the overall standard includes a step S461 of initializing the deep learning model corresponding to the overall standard, a step S462 of obtaining historical data corresponding to the overall standard, a step S463 of generating a fifth deep learning model, a step S464 of generating a sixth deep learning model, and a step S465 of setting the preset model corresponding to the overall standard:

and a deep learning model initialization step S461 corresponding to the overall standard, wherein the deep learning model corresponding to the overall standard is initialized, the input format of the deep learning model is set to be the format of a set of third data corresponding to each standard in the audit standard, the output format of the deep learning model is set to be the format of the third data corresponding to the overall standard, and the deep learning model obtained through initialization is used as a fourth deep learning model.

A history data acquisition step S462 corresponding to the overall standard: and acquiring a set of third data corresponding to each standard in the auditing standards of each object which is audited and third data corresponding to the overall standard from historical big data. The third data corresponding to each criterion may be a score or an evaluation result corresponding to each criterion or other numerical values reflecting the degree of the second data meeting each criterion. The third data corresponding to the overall standard may be a score or evaluation result corresponding to the overall standard or other numerical value reflecting the degree to which the third data corresponding to each standard meets the overall standard.

And a fifth deep learning model generation step S463, which is to perform unsupervised training on the fourth deep learning model by using a set of third data of each object, which is subjected to audit and corresponds to each standard in the audit standards, as input data of the deep learning model, and to obtain the fourth deep learning model as the fifth deep learning model through unsupervised training.

And a sixth deep learning model generating step S464, namely respectively taking the third data set corresponding to each standard in the auditing standard of each audited object and the third data corresponding to the overall standard as input data and output data of the fifth deep learning model, and carrying out supervised training on the fifth deep learning model. And respectively taking a set of third data of each object which is subjected to auditing and corresponds to each standard in the auditing standards and third data corresponding to the overall standard as input data and output data of the fifth deep learning model, namely taking the third data of each object which is subjected to auditing and corresponds to each standard as the input data of the fifth deep learning model, taking the third data of each object which is subjected to auditing and corresponds to the overall standard as the output data of the fifth deep learning model, and taking the fifth deep learning model obtained through supervised training as a sixth deep learning model.

A step S465 of setting a preset model corresponding to the overall standard: and taking the sixth deep learning model as a preset model corresponding to the overall standard.

And an overall standard judging step S470, judging whether the object which does not apply for auditing meets the overall standard according to the third data corresponding to each standard, the preset model corresponding to the overall standard and the preset range.

The overall criterion judging step S470 includes a third data generating step S471 corresponding to the overall criterion, a preset range acquiring step S472 corresponding to the overall criterion, and a third data judging step S473 corresponding to the overall criterion.

And a third data generation step S471, in which third data corresponding to the overall standard is calculated according to the third data corresponding to each standard and the preset model corresponding to the overall standard to obtain the third data corresponding to the overall standard. And executing the preset model corresponding to each standard on a computer, taking the third data corresponding to each standard as the input of the preset model corresponding to the overall standard, and taking the calculated output as the third data corresponding to the overall standard. Preferably, the third data corresponding to each criterion is used as the input of the sixth deep learning model corresponding to the overall criterion, and the calculated output of the sixth deep learning model is used as the third data corresponding to the overall criterion. The third data corresponding to each criterion may be a score or an evaluation result corresponding to each criterion or other numerical values reflecting the degree of the second data meeting each criterion. The third data corresponding to the overall standard may be a score or evaluation result corresponding to the overall standard or other numerical value reflecting the degree to which the third data corresponding to each standard meets the overall standard.

A preset range obtaining step S472 corresponding to the overall standard, obtaining a preset range corresponding to the overall standard. The overall score for the overall criteria will generally have a range, e.g., greater than 80 points.

A third data judgment step S473 corresponding to the overall standard, which is to judge whether the third data corresponding to the overall standard is in a preset range corresponding to the overall standard: if yes, judging that the object which does not apply for auditing meets the overall standard; and if not, judging that the object which does not apply for auditing does not accord with the overall standard.

A comprehensive judgment step S480: judging whether the object which does not apply for auditing meets each standard and the overall standard in the auditing standard: if yes, judging that the object which does not apply for auditing passes the auditing of a preset category, namely, judging that the object which does not apply for auditing passes the auditing as a result of the auditing; and if not, judging that the object which does not apply for auditing does not pass the auditing of the preset category, namely the result of the auditing is that the object which does not apply for auditing does not pass the auditing. Compliance with each of the audit standards and the overall standard refers to compliance with both of the audit standards and the overall standard. And if a certain standard or the total standard is not met, judging that the object which does not apply for auditing does not pass the auditing of the preset category.

An active feedback step S500: judging whether the audit result of the preset category passes the audit: if not (under the condition of not passing the audit), sending the result of not passing the audit to the object which does not apply for the audit, obtaining each standard or overall standard which does not conform to the audit, feeding back the standard or overall standard to the object which does not apply for the audit, and sending information to remind the object which does not apply for the audit of correcting each standard or overall standard which does not conform to the standard or overall standard.

If the object which does not apply for audit passes the audit, the audit shows that the object which does not apply for audit reaches the audit standard, so that the object which does not apply for audit does not need to be informed of the audit result, and the object which does not apply for audit does not need to be rectified, so that the object which does not apply for audit is not disturbed, and the effect of performing active audit service on the object which does not apply for audit silently is achieved. In the active auditing process, the object which does not apply for auditing does not need to pay any time, manpower and material resources to participate, even the object which does not apply for auditing in the whole process is unknown, and the object can be informed only under the condition that the auditing in the active auditing process does not pass. Therefore, the method can protect the driving of the object which does not apply for the audit, and the object which does not apply for the audit cannot be interfered.

The steps can be executed on a big data platform such as Spark and the like to accelerate the speed of big data processing.

(II) the system in various embodiments of the invention includes various combinations of the following modules:

the audit standard acquisition module 100 performs the audit standard acquisition step S100.

The execution object data acquisition module 200 executes the object data acquisition step S200.

The object data acquisition module 200 includes a data source acquisition module 210 and an object data retrieval module 220.

The data source acquisition module 210 performs a data source acquisition step S210.

The object data retrieving module 220 performs the object data retrieving step S220.

The data acquisition module 300 corresponding to the standard performs the data acquisition step S300 corresponding to the standard.

The standard corresponding data acquisition module 300 comprises a data screening module 310 and a data cleaning module 320.

The data filtering module 310 performs a data filtering step S310.

The data washing module 320 performs a data washing step S320.

The data cleaning module 320 includes a corresponding data source obtaining module 321, a reliability obtaining module 322, and a reliability selecting module 323.

The corresponding data source acquiring module 321 executes the corresponding data source acquiring step S321.

The reliability acquisition module 322 executes the reliability acquisition step S322.

The reliability selecting module 323 performs the reliability selecting step S323.

The audit determination module 400 performs the audit determination step S400.

The audit determination module 400 includes a sub-standard obtaining module 410, a corresponding data extracting module 420, a preset model obtaining module 430 corresponding to each standard, a judgment module 440 of each standard, a preset model obtaining module 450 corresponding to a total standard, a judgment module 460 of the total standard, and a comprehensive judgment module 470.

The sub-criterion acquisition module 410 performs the sub-criterion acquisition step S410.

The correspondence data extraction module 420 performs a correspondence data extraction step S420.

The preset model obtaining module 430 corresponding to each standard executes the preset model obtaining step S430 corresponding to each standard.

The preset model obtaining module 430 corresponding to each standard includes a deep learning model initializing module 431 corresponding to each standard, a historical data obtaining module 432 corresponding to each standard, a second deep learning model generating module 433, a third deep learning model generating module 434, and a preset deep learning model setting module 435 corresponding to each standard.

The deep learning model initialization module 431 corresponding to each criterion performs the deep learning model initialization step S431 corresponding to each criterion.

The historical data acquisition module 432 corresponding to each criterion performs the historical data acquisition step S432 corresponding to each criterion.

The second deep learning model generation module 433 executes a second deep learning model generation step S433.

The third deep learning model generation module 434 performs the third deep learning model generation step S434.

The preset deep learning model setting module 435 corresponding to each standard performs the preset model setting step S435 corresponding to each standard.

The third data generation module 440 corresponding to each item of standard performs the third data generation step S440 corresponding to each item of standard.

Each criterion judging module 450 performs each criterion judging step S450.

Each criterion judging module 450 includes a preset range acquiring module 451 corresponding to each criterion, and a third data judging module 452 corresponding to each criterion.

The preset range acquisition module 451 corresponding to each criterion performs the preset range acquisition step S451 corresponding to each criterion.

The third data determining module 452 corresponding to each criterion performs the third data determining step S452 corresponding to each criterion.

The preset model obtaining module 460 corresponding to the overall standard performs the preset model obtaining step S460 corresponding to the overall standard.

The preset model obtaining module 460 corresponding to the overall standard includes a deep learning model initializing module 461 corresponding to the overall standard, a historical data obtaining module 462 corresponding to the overall standard, a fifth deep learning model generating module 463, a sixth deep learning model generating module 464, and a preset model setting module 465 corresponding to the overall standard:

the ensemble-standard corresponding deep learning model initialization module 461 performs an ensemble-standard corresponding deep learning model initialization step S461.

The history data acquisition module 462 corresponding to the overall criterion performs the history data acquisition step S462 corresponding to the overall criterion.

The fifth deep learning model generation module 463 performs a fifth deep learning model generation step S463.

The sixth deep learning model generation module 464 performs a sixth deep learning model generation step S464.

The preset model setting module 465 corresponding to the overall standard executes the preset model setting step S465 corresponding to the overall standard.

The overall criterion judging module 470 performs the overall criterion judging step S470.

The overall criterion determining module 470 includes a third data generating module 471 corresponding to the overall criterion, a preset range obtaining module 472 corresponding to the overall criterion, and a third data determining module 473 corresponding to the overall criterion.

The third data generation module 471 corresponding to the overall criterion executes the third data generation step S471 corresponding to the overall criterion.

The preset range obtaining module 472 corresponding to the overall standard executes the preset range obtaining step S472 corresponding to the overall standard.

The third data judgment module 473 corresponding to the overall criterion performs the third data judgment step S473 corresponding to the overall criterion.

The comprehensive decision module 480 performs a comprehensive decision step S480.

The active feedback module 500 performs an active feedback step S500.

The modules can be deployed on large data platforms such as Spark and the like to accelerate the speed of large data processing.

(III) several embodiments of the invention

Embodiment 1 provides an auditing method, which includes an auditing standard acquisition step S100, an object data acquisition step S200, a data acquisition step S300 corresponding to a standard, an auditing judgment step S400, and an active feedback step S500, as shown in fig. 1.

Embodiment 2 provides an auditing method comprising the steps of the method described in embodiment 1; the object data acquiring step S200 includes a data source acquiring step S210 and an object data retrieving step S220, and the data acquiring step S300 corresponding to the standard includes a data screening step S310 and a data cleaning step S320.

Embodiment 3 provides an auditing method comprising the steps of the method described in embodiment 2; the data cleaning step S320 includes a corresponding data source obtaining step S321, a reliability obtaining step S322, and a reliability selecting step S323.

Embodiment 4 provides an auditing method comprising the steps of the method described in embodiment 2; the audit judgment step S400 includes a sub-standard obtaining step S410, a corresponding data extracting step S420, a preset model obtaining step S430 corresponding to each standard, a third data generating step S440 corresponding to each standard, a criterion judging step S450, a preset model obtaining step S460 corresponding to the overall standard, an overall standard judging step S470, and a comprehensive judging step S480, as shown in fig. 2.

Embodiment 5 provides an auditing method comprising the steps of the method described in embodiment 4; the preset model obtaining step S430 corresponding to each standard includes a deep learning model initializing step S431 corresponding to each standard, a historical data obtaining step S432 corresponding to each standard, a second deep learning model generating step S433, a third deep learning model generating step S434, and a preset deep learning model setting step S435 corresponding to each standard, as shown in fig. 3; the preset model acquiring step S460 corresponding to the overall standard includes a deep learning model initializing step S461 corresponding to the overall standard, a historical data acquiring step S462 corresponding to the overall standard, a fifth deep learning model generating step S463, a sixth deep learning model generating step S464, and a preset model setting step S465 corresponding to the overall standard, as shown in fig. 4.

Embodiment 6 provides an auditing method comprising the steps of the method described in embodiment 4; each standard judging step S450 includes a preset range acquiring step S451 corresponding to each standard, and a third data judging step S452 corresponding to each standard; the overall criterion judging step S470 includes a third data generating step S471 corresponding to the overall criterion, a preset range acquiring step S472 corresponding to the overall criterion, and a third data judging step S473 corresponding to the overall criterion.

Embodiment 7 provides an auditing system, which includes an auditing standard obtaining module 100, an object data obtaining module 200, a data obtaining module 300 corresponding to a standard, an auditing judging module 400, and an active feedback module 500, as shown in fig. 5.

Embodiment 8 provides an auditing system comprising the steps of the system described in embodiment 7; the object data obtaining module 200 includes a data source obtaining module 210 and an object data retrieving S220, and the data obtaining module 300 corresponding to the standard includes a data screening module 310 and a data cleaning module 320.

Embodiment 9 provides an auditing system comprising the steps of the system described in embodiment 8; the data cleaning module 320 includes a corresponding data source obtaining module 321, a reliability obtaining module 322, and a reliability selecting module 323.

Embodiment 10 provides an auditing system comprising the steps of the system described in embodiment 8; the audit determination module 400 includes a sub-standard obtaining module 410, a corresponding data extracting module 420, a preset model obtaining module 430 corresponding to each standard, a third data generating module 440 corresponding to each standard, each standard determining module 450, a preset model obtaining module 460 corresponding to a total standard, a total standard determining module 470, and a comprehensive determining module 480, as shown in fig. 6.

Embodiment 11 provides an auditing system comprising the steps of the system described in embodiment 10; the preset model obtaining module 430 corresponding to each standard includes a deep learning model initializing module 431 corresponding to each standard, a historical data obtaining module 432 corresponding to each standard, a second deep learning model generating module 433, a third deep learning model generating module 434, and a preset deep learning model setting module 435 corresponding to each standard, as shown in fig. 7; the preset model obtaining module 460 corresponding to the overall standard includes a deep learning model initializing module 461 corresponding to the overall standard, a historical data obtaining module 462 corresponding to the overall standard, a fifth deep learning model generating module 463, a sixth deep learning model generating module 464, and a preset model setting module 465 corresponding to the overall standard, as shown in fig. 8.

Embodiment 12 provides an auditing system comprising the steps of the system described in embodiment 10; each standard judging module 450 comprises a preset range obtaining module 451 corresponding to each standard and a third data judging module 452 corresponding to each standard; the overall criterion determining module 470 includes a third data generating module 471 corresponding to the overall criterion, a preset range obtaining module 472 corresponding to the overall criterion, and a third data determining module 473 corresponding to the overall criterion.

Embodiment 13 provides a robot system in which the auditing systems as described in embodiments 7 to 12 are respectively arranged.

The methods and systems of the various embodiments described above may be performed and deployed on computers, servers, cloud servers, supercomputers, robots, embedded devices, electronic devices, and the like.

The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. An active auditing method is characterized in that in the active auditing process, enterprises which do not apply for auditing do not need to pay any time, manpower and material resources to participate, even enterprises which do not apply for auditing in the whole process are not informed, and the enterprises can receive notifications only under the condition that the enterprises do not pass in the active auditing process, and the method comprises the following steps:

an audit standard obtaining step, which is used for obtaining the audit standard of the preset category; an object data acquisition step, which is used for acquiring data of enterprises which do not apply for auditing from a data source; the data source comprises data provided by a third party;

a data acquisition step corresponding to a standard, which is used for acquiring data corresponding to the audit standard from the data of the enterprise which does not apply for audit; the data acquisition step corresponding to the standards comprises the steps of increasing and decreasing the credibility of different data sources according to whether the historical second data are correct or not to form different credibility of the different data sources, and acquiring the second data corresponding to each standard according to the credibility of the data sources;

an audit judging step, which is used for judging whether the data corresponding to the audit standard meets the audit standard: if so, the enterprise which does not apply for auditing audits passes the auditing, and the enterprise which does not apply for auditing is not notified of the auditing result; if not, the enterprise audit which does not apply for audit does not pass;

and actively feeding back the result of the audit to the enterprise which does not apply for the audit and does not pass the audit.

2. An auditing method according to claim 1,

the object data acquiring step includes:

a data source obtaining step for obtaining a data source;

the object data retrieval step is used for retrieving and acquiring the data of the enterprise which does not apply for auditing from the data source;

a data screening step, which is used for screening out data corresponding to the auditing standard from the data of the enterprise which does not apply for auditing as first data;

3. The auditing method of claim 1, wherein the proactive feedback step comprises:

judging whether the audit result of the preset category passes the audit: and if not, sending the result which does not pass the audit to the enterprise which does not apply for the audit, obtaining each non-conforming standard or total standard and feeding back the non-conforming standard or total standard to the enterprise which does not apply for the audit, and sending information to remind the enterprise which does not apply for the audit to carry out rectification on each non-conforming standard or total standard.

4. An auditing method according to claim 2,

the auditing judgment step comprises:

each standard judging step is used for judging whether the enterprises which do not apply for auditing meet each standard or not according to the third data and the preset range corresponding to each standard;

a total standard judging step, configured to judge whether the enterprise that does not apply for auditing meets the total standard according to the third data corresponding to each standard, the preset model corresponding to the total standard, and the preset range;

and a comprehensive judgment step for judging whether the enterprises which do not apply for auditing meet each standard and the overall standard in the auditing standard.

5. Auditing method according to claim 4,

a historical data acquisition step corresponding to each standard, which is used for acquiring the second data and the third data of each enterprise which is audited and corresponds to each standard from historical big data;

a second deep learning model generation step, configured to use the second data of each enterprise, which corresponds to each standard and has undergone auditing, as input data of the first deep learning model, perform unsupervised training on the first deep learning model, and use the obtained first deep learning model as a second deep learning model;

a third deep learning model generation step, configured to respectively use the second data and the third data of each enterprise, which is subjected to auditing and corresponds to each standard, as input data and output data of the second deep learning model, perform supervised training on the second deep learning model, and use the obtained second deep learning model as a third deep learning model;

a historical data acquisition step corresponding to an overall standard, which is used for acquiring a set of third data corresponding to each standard in the auditing standard of each enterprise which has undergone auditing and third data corresponding to the overall standard from historical big data;

a fifth deep learning model generation step, configured to use the set of the third data of each enterprise that has undergone auditing and corresponds to each standard in the auditing standards as input data of the deep learning model, perform unsupervised training on the fourth deep learning model, and use the obtained fourth deep learning model as a fifth deep learning model;

a sixth deep learning model generation step, configured to use the set of the third data corresponding to each standard in the audit standards of each enterprise that has undergone audit and the third data corresponding to the overall standard as input data and output data of the fifth deep learning model, respectively, perform supervised training on the fifth deep learning model, and use the obtained fifth deep learning model as a sixth deep learning model;

6. Auditing method according to claim 4,

each standard judging step comprises the following steps:

a third data judgment step corresponding to each standard, which is used for judging whether the enterprises which do not apply for auditing meet each standard or not;

the overall standard judging step comprises the following steps:

and a third data judgment step corresponding to the overall standard, which is used for judging whether the enterprise which does not apply for auditing meets the overall standard.

7. An active auditing system, characterized in that, in the active auditing process, the enterprise which does not apply for auditing does not need to pay any time and manpower and material resources to participate, even the enterprise which does not apply for auditing in the whole process is not informed, and the enterprise can receive notice only under the condition that the enterprise does not pass in the active auditing process, the system comprises:

the object data acquisition module is used for acquiring data of enterprises which do not apply for auditing from a data source; the data source comprises data provided by a third party;

the data acquisition module corresponding to the standard is used for acquiring data corresponding to the audit standard from the data of the enterprise which does not apply for the audit; the data acquisition module corresponding to the standard comprises a submodule for increasing and decreasing the credibility of different data sources according to whether the historical second data are correct or not to form different credibility of the different data sources and acquiring the second data corresponding to each standard according to the credibility of the data sources;

the audit judging module is used for judging whether the data corresponding to the audit standard meets the audit standard: if so, the enterprise which does not apply for auditing audits passes the auditing, and the enterprise which does not apply for auditing is not notified of the auditing result; if not, the enterprise audit which does not apply for audit does not pass;

and the active feedback module is used for feeding back the auditing result to the enterprise which does not apply for auditing and does not pass the auditing.

8. The audit system of claim 7,

the enterprise data acquisition module comprises:

the data source acquisition module is used for acquiring a data source;

the enterprise data retrieval module is used for retrieving and acquiring the data of the enterprise which does not apply for auditing from the data source;

the data acquisition module corresponding to the standard comprises:

the data screening module is used for screening out data corresponding to the audit standard from the data of the enterprise which does not apply for audit as first data;

the data cleaning module is used for extracting data corresponding to each standard from the first data to serve as second data corresponding to each standard;

the active feedback module comprises:

9. The audit system of claim 8,

the audit judgment module comprises:

each standard judging module is used for judging whether the enterprises which do not apply for auditing meet each standard or not according to the third data and the preset range corresponding to each standard;

the overall standard judging module is used for judging whether the enterprise which does not apply for auditing meets the overall standard or not according to the third data corresponding to each standard, the preset model corresponding to the overall standard and the preset range;

and the comprehensive judgment module is used for judging whether the enterprises which do not apply for auditing meet each standard and the overall standard in the auditing standard.

10. A robot system, characterized in that the robot is provided with an auditing system according to any one of claims 7-9.