CN110136010A - The method, apparatus and computer equipment of risk case are judged based on neural network - Google Patents
The method, apparatus and computer equipment of risk case are judged based on neural network Download PDFInfo
- Publication number
- CN110136010A CN110136010A CN201910314537.8A CN201910314537A CN110136010A CN 110136010 A CN110136010 A CN 110136010A CN 201910314537 A CN201910314537 A CN 201910314537A CN 110136010 A CN110136010 A CN 110136010A
- Authority
- CN
- China
- Prior art keywords
- risk
- target
- information
- record
- event
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/08—Insurance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
This application discloses a kind of method, apparatus and computer equipment that risk case is judged based on neural network, and wherein method includes: the record information of reading event；Receive the target risk type that user terminal selects in multiple risk classifications；Obtain target information corresponding with the target risk type in the corresponding target risk judgment models of the target risk type and the record information；The target information is input in the target risk judgment models；Risk case coefficient is calculated；If the risk case coefficient is more than default first risk threshold value, determine the event for risk case.The application reads the record information of event automatically, and is calculated the corresponding data recorded in information according to the model of user's selection, objectively user is helped to calculate the risk system of outgoing event, saves the time of user.
Description
Technical field
This application involves field of artificial intelligence is arrived, one kind is especially related to based on neural network and judges risk case
Method, apparatus, computer equipment and storage medium.
Background technique
Vehicle insurance full name is automobile insurance, is referred to motor vehicles people due to caused by natural calamity or contingency
A kind of business insurance of body injures and deaths or the negative liability to pay compensation of property loss.User carries out using automobile as insurance subject to insurance company
It insures, when the case where insurance clause agreement occurs in automobile, insurance company can send prospecting personnel carry out whether checking audit
The case where meeting agreement, automobile is then sent to specified maintenance factory's maintenance or user, and being about to automobile sends to specified dimension certainly
Factory's maintenance is repaired, and undertakes whole or undertakes part because agreement situation occurs loss caused by insurer.
Being in danger for some vehicle insurances is due to insurer, prospecting personnel, maintenance factory, determines there is no either stretch of fact
Damage person etc. is related to carrying out insurance fraud between the one or more people between Claims Resolution process " cooperation ", gains insurance company by cheating to reach
Property.This kind of of insurance company's generation is defined as risk case by the case of insurance fraud.The Claims Resolution link of vehicle insurance is more, such as report a case to the security authorities,
It surveys, setting loss, quotation, guidance, adjustment etc., the staff and insurer of each link or the maintenance cooperated with insurance company
There may be to insurance company's insurance fraud for factory, setting loss mechanism etc..
Summary of the invention
The main purpose of the application is to provide a kind of method, apparatus and computer that risk case is judged based on neural network
Equipment, it is intended to solve the problems, such as staff oneself according to event to determine whether being risk case and needing the plenty of time.
In order to achieve the above-mentioned object of the invention, the application proposes a kind of method that risk case is judged based on neural network, packet
It includes:
The record information of reading event；
Receive the target risk type that user terminal selects in multiple risk classifications；
Obtain in the corresponding target risk judgment models of the target risk type and the record information with the mesh
Mark the corresponding target information of risk classifications；
The target information is input in the target risk judgment models；
The model coefficient of the target risk judgment models output is received, and risk is calculated according to the model coefficient
Event coefficient, the model coefficient refer to the data that target risk judgment models are obtained according to the associated information calculation of event；
If the risk case coefficient is more than preset the first risk threshold value corresponding with the target risk type, sentence
The fixed event is risk case.
Further, before the step target information being input in the target risk judgment models, packet
It includes:
Acquire the record information of multiple events, the multiple event include the risk case of multiple same risk classifications with
And multiple non-risk cases；
The record information input of the multiple event is trained to convolutional neural networks model, obtains the risk class
The risk judgment model of type.
Further, described to be trained the record information input of the multiple event to convolutional neural networks model
Step, comprising:
Read putting on record the date in the record information；
According to put on record date and the preset computation rule, the corresponding flow time rank of each process flow is calculated
Section；
The record date for reading the record data in record information respectively, according to the record date of each record data and respectively
The record data are added the label in corresponding flow time stage by the flow time stage respectively；
According to the training type that user selects, the note of the label in the trained type corresponding flow time stage will be had
Record data are input to convolutional neural networks model and are trained.
Further, the model coefficient for receiving the target risk judgment models output, and according to the model system
The step of risk case coefficient is calculated in number, comprising:
Read party's information in the record information；
Count the object event quantity of the event in preset time period with party's information；
According to the object event quantity, target weight system is determined in the corresponding relationship of event number and weight coefficient
Number；
The model coefficient for receiving the target risk judgment models output, by the target weight coefficient multiplied by the model
Coefficient obtains the risk case coefficient.
Further, the acquisition corresponding target risk judgment models of target risk type and the record
In information the step of target information corresponding with the target risk type before, comprising:
Read the target amount for which loss settled value in the record information；
The mapping table of preset amount for which loss settled threshold value and risk classifications is called, is obtained and the target risk type pair
The target amount for which loss settled threshold value answered；
Judge whether the target amount for which loss settled value is greater than the target amount for which loss settled threshold value；
If so, executing the step: obtaining the corresponding target risk judgment models of the target risk type, Yi Jisuo
State target information corresponding with the target risk type in record information.
The application also provides a kind of device that risk case is judged based on neural network, comprising:
First read module, for reading the record information of event；
Receiving module, the target risk type selected in multiple risk classifications for receiving user terminal；
First obtains module, for obtaining corresponding target risk judgment models of the target risk type and described
Record target information corresponding with the target risk type in information；
Input module, for the target information to be input in the target risk judgment models；
Computing module, for receiving the model coefficient of the target risk judgment models output, and according to the model system
Risk case coefficient is calculated in number, and the model coefficient refers to target risk judgment models according to the associated information calculation of event
The data obtained；
Determination module, if being more than preset with the target risk type corresponding first for the risk case coefficient
Risk threshold value then determines the event for risk case.
Further, the device that risk case is judged based on neural network further include:
Acquisition module, for acquiring the record information of multiple events, the multiple event includes multiple same risk classes
The risk case of type and multiple non-risk cases；
Training module, for the record information input of the multiple event to be trained to convolutional neural networks model,
Obtain the risk judgment model of the risk classifications.
Further, the training module includes:
Date unit is read, for reading putting on record the date in the record information；
Calculation stages unit calculates each process flow for date and the preset computation rule of putting on record according to
The corresponding flow time stage；
Adding unit, for reading the record date of the record data in record information respectively, according to each record data
The record data are added the label in corresponding flow time stage by record date and each flow time stage respectively；
Training unit, the training type for being selected according to user will have the corresponding flow time of the trained type
The record data of the label in stage are input to convolutional neural networks model and are trained.
The application also provides a kind of computer equipment, including memory and processor, and the memory is stored with computer
The step of program, the processor realizes any of the above-described the method when executing the computer program.
The application also provides a kind of computer readable storage medium, is stored thereon with computer program, the computer journey
The step of method described in any of the above embodiments is realized when sequence is executed by processor.
The method, apparatus and computer equipment that risk case is judged based on neural network of the application, reads event automatically
Record information, and according to user selection model by record information in corresponding data calculate, objectively help use
Family calculates the risk system of outgoing event, is conducive to that user is helped quickly to determine whether the event is risk case, saves user
Time, improve the determination rate of accuracy of risk case.
Detailed description of the invention
Fig. 1 is the flow diagram of the method that risk case is judged based on neural network of one embodiment of the application；
Fig. 2 is the structural schematic block diagram of the device that risk case is judged based on neural network of one embodiment of the application；
Fig. 3 is the structural schematic block diagram of the computer equipment of one embodiment of the application.
The embodiments will be further described with reference to the accompanying drawings for realization, functional characteristics and the advantage of the application purpose.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not
For limiting the application.
Referring to Fig.1, the embodiment of the present application provides a kind of method that risk case is judged based on neural network, comprising steps of
S1, the record information for reading event；
S2, the target risk type that user terminal selects in multiple risk classifications is received；
S3, obtain in the corresponding target risk judgment models of target risk type and the record information with institute
State the corresponding target information of target risk type；
S4, the target information is input in the target risk judgment models；
S5, the model coefficient for receiving the target risk judgment models output, and be calculated according to the model coefficient
Risk case coefficient, the model coefficient refer to the number that target risk judgment models are obtained according to the associated information calculation of event
According to；
If S6, the risk case coefficient are more than preset the first risk threshold value corresponding with the target risk type,
Then determine the event for risk case.
As described in above-mentioned steps S1, in the concrete application scene of an insurance company, when insurer is public in an insurance
After department insures, that is, policy information is formed, which includes insurer's information, subject matter insured (automobile) information, Claims Resolution letter
Breath etc., wherein insurer can also be referred to as party, and corresponding insurer's information is also party's information.When the vapour of insurer
After accident has occurred in vehicle, that is, it is in danger, the reason of insurer reports a case to the security authorities to insurance company, and then insurance company is according to accident etc.
Relevant information is also recorded, and insurance information is formed.Insurance, that is, the event, entire policy information and insurance information are to belong to
In record information.The record information of each event includes specific number of being in danger, place of being in danger, the time of being in danger, be in danger people
Name, target of insuring information, situation of being in danger description information and the information of process tracking personnel, the information for reconnoitring personnel etc. and thing
Whole information of part and post-processing of being in danger.Wherein, the number of being in danger of each insurance company is unique.Staff exists
During all information of each event are uploaded onto the server.When staff needs to judge whether an event is risk case
When, the Case Number of incoming event, server finds the Case Number of the event according to the Case Number, reads the event
Whole record information.Alternatively, issuing some information relevant to event with user orientation server to carry out filtering out part thing
Part after then determination will determine whether the event of risk case, is then forwarded to server, and server reads the thing that user selectes
The record information of part.
As described in above-mentioned steps S2, risk case can be summarized as several classes, in one embodiment, according to risk source packet
Insurer, prospecting employee, repair shop's three classes are included, risk source is insurer indicates it is that insurer makes false event information to guarantor
Premium is gained by cheating by dangerous company；Risk source is that prospecting employee indicates to be insurer and prospecting employee's cooperative manufacture falseness record to insurance
Premium is gained by cheating by company；Risk source be repair shop indicate repair shop false maintenance record is done to the vehicle that is in danger with exaggerate maintenance the amount of money with
Maintenance cost is prepared to take to insurance company.Judgment mode corresponding to the risk case of different risk sources is not exactly the same, therefore
Using different types of risk judgment model.After server has read above-mentioned record information, all risk classifications are loaded out, with
A risk classifications are selected for user, as target risk type, i.e. user needs to determine whether the event has above-mentioned target
The corresponding risk of risk classifications.User sends the event for needing to determine to server when which risk classifications is judgement event be
Risk classifications, server receive user terminal issue target risk type.
As described in above-mentioned steps S3, after server has received the target risk type of user terminal sending, in preset wind
Target risk judgment models corresponding with above-mentioned target risk type are searched in dangerous judgment models database.Risk judgment pattern number
According to have in library it is multiple it is trained after risk judgment model, in one embodiment, have insurer's risk judgment model, survey
Examine multiple risk judgment models such as employee's risk judgment model, repair shop's risk judgment model.Meanwhile each risk classifications are corresponding
Decision factor be it is different, i.e., it is corresponding record information in information be not identical.Risk classifications with it is corresponding
The corresponding relationship of record information type is also to obtain after staff rule of thumb arranges to process and relevant relationship.Service
Device finds target information corresponding with target risk type in record information according to the corresponding relationship.
As described in above-mentioned steps S4, after server has selected target risk judgment models, then target information is input to
In target risk judgment models, target risk judgment models obtain model coefficient, then based on model coefficient and in advance by calculating
If rule, risk case coefficient is calculated.Above-mentioned multiple risk judgment models are using identical neural network model
What training obtained, except that, according to different risk classifications, obtaining the different numbers in record information when training
According to being trained.Recording in information includes all information relevant to Claims Resolution, is analyzed according to flow nodes, entire to settle a claim
Process include report a case to the security authorities process, survey process, checking process, setting loss process, maintenance flow, compensate process, in a record information
It include information involved by above-mentioned whole processes.For example, will have when being trained to insurer's risk judgment model
The process of reporting a case to the security authorities participated in insurer in the record information of insurer's risk is surveyed information involved in process and is called out,
As the sample of the training model, multiple samples are so acquired, sample set is formed, the data in sample set is then input to mind
Through in network model, neural network model optimizes to obtain the spy of insurer's risk automatically according to the data in above-mentioned sample set
Coefficient is levied, even if the neural network model can be used for judging recording whether information has insurer's risk.In subsequent judgement one
When whether a record information has insurer's risk, only by the corresponding data of process of reporting a case to the security authorities of the record information and stream need to be surveyed
The corresponding data of journey are input in the neural network model, and then neural network model calculates whether the data include above-mentioned instruction
The characteristic coefficient got, if it is, determining the record information, there are insurer's risks.
As described in above-mentioned steps S5, the calculating process of monitoring server target risk judgment models, when target risk judges
After model calculates, server receives the model coefficient of model output.Then server is by model coefficient according to preset meter
It calculates rule and carries out calculation processing, obtain risk case coefficient.Risk case coefficient indicates that the event is above-mentioned target risk type
Risk case a possibility that, risk case coefficient is higher, illustrates a possibility that event is the risk case of the risk classifications
It is bigger.
As described in above-mentioned steps S6, the first risk threshold value is, different type corresponding with above-mentioned target risk judgment models
The risk threshold value of risk judgment model be not identical.The risk case of each type respectively has a corresponding risk threshold
Value.Above-mentioned risk case coefficient the first risk threshold value corresponding with the target risk type is compared by server, if wind
Dangerous event coefficient is more than above-mentioned first risk threshold value, illustrates that event is the risk case of the risk classifications.
In one embodiment, it is above-mentioned by the target information be input to the step in the target risk judgment models it
Before, comprising:
The record information of S401, the multiple events of acquisition, the multiple event includes the risk of multiple same risk classifications
Event and multiple non-risk cases；
S402, the record information input of the multiple event is trained to convolutional neural networks model, is obtained described
The risk judgment model of risk classifications.
In the present embodiment, the record information of multiple events is acquired by staff, and is marked to each event
Whether note is risk case, wherein above-mentioned event is the event of the same risk classifications, to establish the wind of the risk classifications
Dangerous judgment models.Server read work personnel acquisition multiple events record information and it is corresponding whether be risk
The label of event, and be input in convolutional neural networks model the learning training that exercises supervision, i.e., to convolutional neural networks model
Parameter optimization obtains the risk judgment model of above-mentioned risk classifications.
In one embodiment, the record information input described above by the multiple event is to convolutional neural networks model
The step of being trained, comprising:
S412, putting on record the date in the record information is read；
S422, put on record according to date and preset computation rule, when calculating the corresponding process of each process flow
Between the stage；
S432, the record date for recording the record data in information is read respectively, according to the record date of each record data
And each flow time stage, the record data are added to the label in corresponding flow time stage respectively；
S442, the training type selected according to user will have the mark in the trained type corresponding flow time stage
The record data of label are input to convolutional neural networks model and are trained.
In the present embodiment, recording in information includes many data, and whole related datas with event are to belong to record
In information.Wherein, it including puts on record the date, the date of putting on record refers to the date that event occurs and registers.It is recorded on record information
In the first row, server reads this and puts on record the date.Then preset computation rule is called, computation rule includes each process flow
Corresponding time phase, i.e., the corresponding time span of each process in processing event procedure, such as requirements of process one day of reporting a case to the security authorities,
Survey requirements of process two days, checking process need one day, setting loss requirements of process three days, maintenance flow need week etc..Clothes
Device be engaged according to above-mentioned computation rule, the date that will put on record plus corresponding number of days, obtains the time phase of each process, it can
When recorded by the data in record information, so that it may determine that the data are to belong to the data of which flow stages.So
Afterwards by the whole record data recorded in information according to record date, which determination be in flow time stage, then in addition
Corresponding flow time phase tag.When user selects one of risk classifications to be trained, the server calls risk
The corresponding data requirements of type obtains the data type for training the risk classifications to need, that is, which flow stages is needed to need
Data are recorded, the data of the label comprising the flow time stage in record information are then input to convolutional neural networks model
In, it is trained.Finally obtain the corresponding risk judgment model of the risk classifications.
In one embodiment, the model coefficient of the above-mentioned reception target risk judgment models output, and according to described
The step of risk case coefficient is calculated in model coefficient, comprising:
S51, the party's information recorded in information is read；
The object event quantity of event with party's information in S52, statistics preset time period；
S53, according to the object event quantity, weight system is determined in the corresponding relationship of event number and weight coefficient
Number；
S54, the model coefficient for receiving the target risk judgment models output, by the target weight coefficient multiplied by described
Model coefficient obtains the risk case coefficient.
Include party's information in above-mentioned record information in the present embodiment, party's information include insurer name,
The information relevant to insurer such as gender, head portrait photo, identity card picture, identification card number.Server selects one of them to have
Then the information (such as identification card number) of unique identification is being taken as the party's information for representing insurer with party's information
The lookup insurer in business device carries out reporting a case to the security authorities within a preset period of time the quantity of the event that generates.The preset time period is by working
Personnel are arranged according to above-mentioned risk judgment model, are usually arranged 1 year, and corresponding event number is also with weight coefficient
Staff is arranged according to above-mentioned risk judgment model.After getting above-mentioned event number, above-mentioned event number is called out
With the corresponding relationship of weight coefficient, target weight coefficient is found.Then target weight coefficient is obtained multiplied by above-mentioned model coefficient
To risk case coefficient, risk case system is compared with the first risk threshold value then, this is determined according to comparison result
Whether event is risk case.
In one embodiment, the corresponding target risk judgment models of the above-mentioned acquisition target risk type, Yi Jisuo
Before the step of stating target information corresponding with the target risk type in record information, comprising:
S301, the target amount for which loss settled value recorded in information is read；
S302, the mapping table for calling preset amount for which loss settled threshold value and risk classifications are obtained and the risk classifications
Corresponding target amount for which loss settled threshold value；
S303, judge whether the target amount for which loss settled value is greater than the target amount for which loss settled threshold value, if so, executing institute
State step: obtain in the corresponding target risk judgment models of the target risk type and the record information with the mesh
Mark the corresponding target information of risk classifications.
In the present embodiment, recording has amount for which loss settled, that is, after the loss in insurance subject has occurred, the warp of generation in information
After Ji loss, insurance company needs to compensate to the amount of money of insurer.Amount for which loss settled is also a part belonged in record information.Clothes
Business device reads the record information of above-mentioned event, reads target amount for which loss settled value therein, while reading the target of user terminal input
Then risk classifications call the mapping table of staff preset risk classifications and amount for which loss settled threshold value, find and mesh
The corresponding target amount for which loss settled threshold value of risk classifications is marked, then by row ratio when target amount for which loss settled value and target amount for which loss settled threshold value
Compared with if target amount for which loss settled value is higher than target amount for which loss settled threshold value, it is likely that be that client wants higher by Claims Resolution acquisition
Interests, which is likely to be risk case, thus generates input instruction, for the specify information in the record information is defeated
Enter into target risk judgment models, further determines whether to be risk case.If above-mentioned target amount for which loss settled value compares target
Amount for which loss settled threshold value is low, and client is very small by the interests that the behavior of insurance fraud obtains, such behavior from legal principle for be
Unreasonable, insurance fraud need not be carried out, so that assert is the loss that insurance agreement really has occurred.It is managed by one target of setting
The judgement for paying for amount of money threshold value, can carry out simple data screening, to reduce the meter of target risk judgment models before calculating
Calculation amount can greatly improve the speed for judging multiple events in the case where whether judge multiple events is risk case.
In one embodiment, after the step of above-mentioned judgement event is risk case, comprising:
S7, the record information of the event is sent to the work people on list corresponding with the target risk type
Member, so that the staff handles the event.
In the present embodiment, the judgement of risk case is judged after being calculated by risk model, can not be certainly
Risk case needs staff further to verify.The staff of insurance company is more, and according to different risk classifications
And classify, the corresponding event of target risk type, which is sent to corresponding staff, to be handled, and the efficiency of processing is more
It is high more acurrate.The employee of managerial staff member is in advance by sole duty in the staff that handles same risk classifications and corresponding
It is that mode (mailbox) is included in onto a list.The server later period has determined the risk case of a target risk type
Afterwards, list corresponding with target risk type is called out, the mailbox then risk case being sent in list.
In one embodiment, it is above-mentioned the record information of the event is sent to it is corresponding with the target risk type
Before the step of staff on list, comprising:
S701, the record information of the event is encrypted.
Include the privacy information of client in the present embodiment, in event, if mail is intercepted by hacker, will lead to the letter of client
Breath leaks, and therefore, the record information of event is encrypted, is then sent to staff by mail again, can be further
Effectively prevent the information leakage of client.Record information is exported in the form of word document, is then added word document
Then word document is compressed and (can be compressed into rar format using winrar compressed software), then will by upper first password
Compressed rar file adds the second password for decompression, and first password and the second password be not identical, and playing preferably prevents
The effect of information leakage.
In one embodiment, it is above-mentioned the record information of the event is sent to it is corresponding with the target risk type
After the step of staff on list, comprising:
S8, the party recorded in information is obtained；
Other event informations of party described in S9, invoking server；
S10, other described event informations are sent to the staff.
In the present embodiment, party, that is, insurer, insurer is usually the buyer insured, when insurance subject is damaged
When, while being also that insurer obtains the people of corresponding interests.The specified location recorded in information is party's information, server
Read record information in designated position, acquire the information of insurer, the identification card number including insurer, then server with
The identification card number of the insurer scans for, and obtains other event informations relevant to insurer, then by other event informations
It is also sent to staff, it is further more acurrate convenient for staff to give staff's more data relevant to insurer
Above-mentioned risk case is handled.
In conclusion the method for judging risk case based on neural network of the application, the automatic record letter for reading event
Breath, and calculated the corresponding data recorded in information according to the model of user's selection, objectively user is helped to calculate
The risk system of event is conducive to that user is helped quickly to determine whether the event is risk case, saves the time of user, mention
The determination rate of accuracy of high risk event.
Referring to Fig. 2, a kind of device that risk case is judged based on neural network is also provided in the embodiment of the present application, comprising:
First read module 1, for reading the record information of event；
Receiving module 2, the target risk type selected in multiple risk classifications for receiving user terminal；
First obtains module 3, for obtaining corresponding target risk judgment models of the target risk type and described
Record target information corresponding with the target risk type in information；
Input module 4, for the target information to be input in the target risk judgment models；
Computing module 5, for receiving the model coefficient of the target risk judgment models output, and according to the model system
Risk case coefficient is calculated in number, and the model coefficient refers to target risk judgment models according to the associated information calculation of event
The data obtained；
Determination module 6, if being more than preset corresponding with the target risk type for the risk case coefficient
One risk threshold value then determines the event for risk case.
In the present embodiment, when insurer is after an insurance company insures, that is, policy information is formed, which includes
Party's information, subject matter insured (automobile) information, Claims Resolution information etc..After accident has occurred in the automobile of insurer, that is, it is in danger,
The reason of insurer reports a case to the security authorities to insurance company, and then insurance company is according to accident etc. relevant information also recorded,
Form insurance information.Insurance, that is, the event, entire policy information and insurance information are to belong to record information.Each event
Record information all includes specific number of being in danger, place of being in danger, the time of being in danger, the people's name that is in danger, target of insuring information, is in danger
Situation description information and the information of process tracking personnel, the information for reconnoitring personnel etc. and event and post-processing of being in danger it is complete
The information in portion.Wherein, the number of being in danger of each insurance company is unique.Staff each event all information on
It passes in server.When staff needs to judge whether an event is risk case, the Case Number of incoming event
One read module 1 is numbered according to the outgoing event, finds the number of being in danger of the event, reads whole record letters of the event
Breath.Alternatively, issuing some information relevant to event with user orientation server to carry out filtering out partial event, then determination will be sentenced
It is fixed whether be risk case event after, be then forwarded to the first read module 1, the first read module 1 reads the thing that user selectes
The record information of part.
Risk case can be summarized as several classes, include insurer, scout according to risk source in one embodiment
Work, repair shop's three classes, risk source are insurer indicates it is that insurer makes false event information to gain premium by cheating to insurance company；
Risk source is that prospecting employee indicates to be insurer and prospecting employee's cooperative manufacture falseness record to gain premium by cheating to insurance company；Wind
Dangerous source is that repair shop indicates that false maintenance record is done to the vehicle that is in danger to exaggerate the maintenance amount of money to be prepared to take to insurance company by repair shop
Maintenance cost.Judgment mode corresponding to the risk case of different risk sources is not exactly the same, therefore using different types of
Risk judgment model.After server has read above-mentioned record information, all risk classifications are loaded out, one for selection by the user
Risk classifications, as target risk type, i.e. user needs to determine whether the event has above-mentioned target risk type corresponding
Risk.User sends the risk classifications for needing the event determined to server when which risk classifications is judgement event be, receives
Module 2 receives the target risk type that user terminal issues.
After receiving module 2 has received the target risk type of user terminal sending, first obtains module 3 in preset risk
Target risk judgment models corresponding with above-mentioned target risk type are searched in judgment models database.Risk judgment model data
Have in library it is multiple it is trained after risk judgment model, in one embodiment, have insurer's risk judgment model, prospecting
Multiple risk judgment models such as employee's risk judgment model, repair shop's risk judgment model.Meanwhile each risk classifications are corresponding
Decision factor be it is different, i.e., it is corresponding record information in information be not identical.Risk classifications and corresponding note
The corresponding relationship of record information type is also to obtain after staff rule of thumb arranges to process and relevant relationship.First obtains
Modulus block 3 finds target information corresponding with target risk type in record information according to the corresponding relationship.
After first acquisition module 3 has selected target risk judgment models, then target information is input to mesh by input module 4
It marks in risk judgment model, target risk judgment models obtain model coefficient, then based on model coefficient and preset by calculating
Rule, risk case coefficient is calculated.Above-mentioned multiple risk judgment models are instructed using identical neural network model
It gets, except that, according to different risk classifications, obtaining the different data in record information when training
To be trained.Recording in information includes all information relevant to Claims Resolution, is analyzed according to flow nodes, entire Claims Resolution stream
Journey include report a case to the security authorities process, survey process, checking process, setting loss process, maintenance flow, compensate process, wrap in a record information
Information involved by above-mentioned whole processes is included.For example, will have throwing when being trained to insurer's risk judgment model
The process of reporting a case to the security authorities participated in insurer in the record information of guarantor's risk is surveyed information involved in process and is called out, and makees
For the sample of the training model, multiple samples are so acquired, sample set is formed, the data in sample set is then input to nerve
In network model, neural network model optimizes to obtain the feature of insurer's risk automatically according to the data in above-mentioned sample set
Coefficient, even if the neural network model can be used for judging recording whether information has insurer's risk.In subsequent judgement one
When whether record information has insurer's risk, only by the corresponding data of process of reporting a case to the security authorities of the record information and process need to be surveyed
Corresponding data are input in the neural network model, and then neural network model calculates whether the data include above-mentioned training
Obtained characteristic coefficient, if it is, determining the record information, there are insurer's risks.
The calculating process of 5 monitoring objective risk judgment model of computing module, after target risk judgment models calculate,
Computing module 5 receives the model coefficient of model output.Then computing module 5 carries out model coefficient according to preset computation rule
Calculation processing obtains risk case coefficient.Risk case coefficient indicates that the event is the risk case of above-mentioned target risk type
A possibility that, risk case coefficient is higher, illustrates that a possibility that event is the risk case of the risk classifications is bigger.
First risk threshold value is, the wind of different types of risk judgment model corresponding with above-mentioned target risk judgment models
Dangerous threshold value is not identical.The risk case of each type respectively has a corresponding risk threshold value.Determination module 6 will be above-mentioned
Risk case coefficient the first risk threshold value corresponding with the target risk type is compared, if risk case coefficient is more than upper
The first risk threshold value is stated, 6 judgement event of determination module is the risk case of the risk classifications.
In one embodiment, the above-mentioned device that risk case is judged based on neural network further include:
Acquisition module, for acquiring the record information of multiple events, the multiple event includes multiple same risk classes
The risk case of type and multiple non-risk cases；
Training module, for the record information input of the multiple event to be trained to convolutional neural networks model,
Obtain the risk judgment model of the risk classifications.
In the present embodiment, the record information of multiple events is acquired by acquisition module, and is marked to each event
Whether note is risk case, wherein above-mentioned event is the event of the same risk classifications, to establish the wind of the risk classifications
Dangerous judgment models.Training module read work personnel acquisition multiple events record information and it is corresponding whether be wind
The label of dangerous event, and be input in convolutional neural networks model the learning training that exercises supervision, i.e., to convolutional neural networks model
Parameter optimization, obtain the risk judgment model of above-mentioned risk classifications.
In one embodiment, above-mentioned training module includes:
Date unit is read, for reading putting on record the date in the record information；
Calculation stages unit calculates each process flow for date and the preset computation rule of putting on record according to
The corresponding flow time stage；
Adding unit, for reading the record date of the record data in record information respectively, according to each record data
The record data are added the label in corresponding flow time stage by record date and each flow time stage respectively；
Training unit, the training type for being selected according to user will have the corresponding flow time of the trained type
The record data of the label in stage are input to convolutional neural networks model and are trained.
In the present embodiment, recording in information includes many data, and whole related datas with event are to belong to record
In information.Wherein, it including puts on record the date, the date of putting on record refers to the date that event occurs and registers.It is recorded on record information
In the first row, read date unit and read this and put on record the date.Then the preset computation rule of calculation stages cell call calculates
Rule includes the corresponding time phase of each process flow, i.e., the corresponding time span of each process, example in processing event procedure
Such as report a case to the security authorities requirements of process one day, survey requirements of process two days, checking process needs one day, setting loss requirements of process three days, maintenance stream
Journey needs week etc..According to above-mentioned computation rule, the date that will put on record plus corresponding number of days, obtains calculation stages unit
The time phase of each process, it can when recorded by the data in record information, so that it may determine that the data are
Belong to the data of which flow stages.Then by the whole record data recorded in information according to record date, which determination is
The flow time stage, then adding unit adds upper corresponding flow time phase tag.When user selects one of wind
When dangerous type is trained, the corresponding data requirements of the server calls risk classifications obtains training risk classifications needs
Data type, that is, the record data for needing which flow stages to need, it includes the process in information that then training unit, which will record,
The data of the label of time phase are input in convolutional neural networks model, are trained.It is corresponding to finally obtain the risk classifications
Risk judgment model.
In one embodiment, above-mentioned computing module 5 includes:
Reading unit, for reading party's information in the record information；
Statistic unit, for counting the object event quantity of the event in preset time period with party's information；
Determination unit is used for according to the object event quantity, in the corresponding relationship of event number and weight coefficient really
Determine weight coefficient；
Computing unit, for receiving the model coefficient of the target risk judgment models output, by the target weight system
Number obtains the risk case coefficient multiplied by the model coefficient.
Include party's information in above-mentioned record information in the present embodiment, party's information include insurer name,
The information relevant to insurer such as gender, head portrait photo, identity card picture, identification card number.The one of tool of reading unit selection
There is the information (such as identification card number) of unique identification, as the party's information for representing insurer, then statistic unit is with the relationship
The lookup insurer of people's information in the server carries out reporting a case to the security authorities the quantity of the event that generates within a preset period of time.The preset time
Section is to be arranged by staff according to above-mentioned risk judgment model, is usually arranged 1 year, corresponding event number and power
Weight coefficient is also that staff is arranged according to above-mentioned risk judgment model.After getting above-mentioned event number, determination unit
The corresponding relationship for calling out above-mentioned event number and weight coefficient finds target weight coefficient.Then computing unit is by target
Weight coefficient obtains risk case coefficient multiplied by above-mentioned model coefficient, then by risk case system and the first risk threshold value into
Row compares, and determines whether the event is risk case according to comparison result.
In one embodiment, the above-mentioned device that risk case is judged based on neural network further include:
Second read module, for reading the target amount for which loss settled value in the record information；
First calling module, for calling the mapping table of preset amount for which loss settled threshold value and risk classifications, obtain with
The corresponding target amount for which loss settled threshold value of the risk classifications；
Judgment module, for judging whether the target amount for which loss settled value is greater than the target amount for which loss settled threshold value, if institute
Target amount for which loss settled value is stated greater than the target amount for which loss settled threshold value, then the first acquisition module 3 is called to obtain the target risk
Target letter corresponding with the target risk type in the corresponding target risk judgment models of type and the record information
Breath.
In the present embodiment, recording has amount for which loss settled, that is, after the loss in insurance subject has occurred, the warp of generation in information
After Ji loss, insurance company needs to compensate to the amount of money of insurer.Amount for which loss settled is also a part belonged in record information.The
Two read modules read the record information of above-mentioned event, read target amount for which loss settled value therein, while the first calling module is read
Then the target risk type for taking user terminal to input calls the preset risk classifications of staff corresponding with amount for which loss settled threshold value
Relation table, the first calling module find target amount for which loss settled threshold value corresponding with target risk type, and then judgment module will
Target amount for which loss settled value is when target amount for which loss settled threshold value compared with row, if target amount for which loss settled value is higher than target amount for which loss settled threshold
Value, it is likely that be that client wants to obtain higher interests by Claims Resolution, which is likely to be risk case, thus instruction module
Input instruction is generated, further to sentence for the specify information in the record information to be input in target risk judgment models
Whether disconnected is risk case.If above-mentioned target amount for which loss settled value is lower than target amount for which loss settled threshold value, the row that client passes through insurance fraud
For and obtain interests it is very small, such behavior from legal principle for be it is unreasonable, insurance fraud need not be carried out, to assert
It is the loss that insurance agreement really has occurred.By the way that the judgement of a target amount for which loss settled threshold value is arranged, can advance calculating
The simple data screening of row is judging whether multiple events are risk things to reduce the calculation amount of target risk judgment models
In the case where part, the speed for judging multiple events can be greatly improved.
In one embodiment, the above-mentioned device that risk case is judged based on neural network further include:
First sending module, for the record information of the event to be sent to name corresponding with the target risk type
Staff on list, so that the staff handles the event.
In the present embodiment, the judgement of risk case is judged after being calculated by risk model, can not be certainly
Risk case needs staff further to verify.The staff of insurance company is more, and according to different risk classifications
And classify, the corresponding event of target risk type is sent to corresponding staff and handled by the first sending module,
What is handled is more efficient more acurrate.The employee of managerial staff member is in advance by sole duty in the staff for handling same risk classifications
And corresponding contact method (mailbox) is included in onto a list.First sending module later period determined a target wind
After the risk case of dangerous type, list corresponding with target risk type is called out, the risk case is then sent to list
In mailbox.
In one embodiment, the above-mentioned device that risk case is judged based on neural network further include:
Encrypting module, for encrypting the record information of the event.
Include the privacy information of client in the present embodiment, in event, if mail is intercepted by hacker, will lead to the letter of client
Breath leaks, and therefore, encrypting module encrypts the record information of event, is then sent to staff by mail again, can
Further effectively to prevent the information leakage of client.Record information is exported in the form of word document, then by word text
Shelves are carried out plus first password, and then being compressed word document (can be compressed into rar lattice using winrar compressed software
Formula), then compressed rar file is added to the second password for being used for decompressing, first password and the second password be not identical, plays
Preferably prevent the effect of information leakage.
In one embodiment, the above-mentioned device that risk case is judged based on neural network further include:
Third obtains module, for obtaining the insurer in the record information；
Second calling module, other insurance informations for insurer described in invoking server；
Second sending module, for other described insurance informations to be sent to the staff.
In the present embodiment, insurer is usually the buyer insured, when insurance subject is damaged, while being also to insure
People obtains the people of corresponding interests.Specified location in record information is party's information, and third obtains module and reads record
Designated position in information acquires the information of insurer, and the identification card number including insurer, then the second calling module is with this
The identification card number of insurer scans for, and obtains other insurance informations relevant to insurer, then the second sending module by its
He is also sent to staff by insurance information, to give the more data relevant to insurer of staff, convenient for staff into
One step is more accurately handled above-mentioned risk case.
In conclusion the device for judging risk case based on neural network of the application, the automatic record letter for reading event
Breath, and calculated the corresponding data recorded in information according to the model of user's selection, objectively user is helped to calculate
The risk system of event is conducive to that user is helped quickly to determine whether the event is risk case, saves the time of user, mention
The determination rate of accuracy of high risk event.
Referring to Fig. 3, a kind of computer equipment is also provided in the embodiment of the present application, which can be server,
Its internal structure can be as shown in Figure 3.The computer equipment includes processor, the memory, network connected by system bus
Interface and database.Wherein, the processor of the Computer Design is for providing calculating and control ability.The computer equipment is deposited
Reservoir includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program
And database.The internal memory provides environment for the operation of operating system and computer program in non-volatile memory medium.It should
The database of computer equipment is for storing the data such as record information, party's information, risk judgment model.The computer equipment
Network interface be used to communicate with external terminal by network connection.To realize one when the computer program is executed by processor
The method that kind judges risk case based on neural network.
The step of above-mentioned processor executes the above-mentioned method that risk case is judged based on neural network: the record of event is read
Information；Receive the target risk type that user terminal selects in multiple risk classifications；It is corresponding to obtain the target risk type
Target information corresponding with the target risk type in target risk judgment models and the record information；By the mesh
Information input is marked into the target risk judgment models；The model coefficient of the target risk judgment models output is received, and
Risk case coefficient is calculated according to the model coefficient, the model coefficient refers to target risk judgment models according to event
The data that obtain of associated information calculation；If the risk case coefficient is more than preset corresponding with the target risk type
First risk threshold value then determines the event for risk case.
In one embodiment, above-mentioned processor execution is described is input to the target risk judgement for the target information
Before step in model, comprising: acquire the record information of multiple events, the multiple event includes multiple same risk classes
The risk case of type and multiple non-risk cases；By the record information input of the multiple event to convolutional neural networks model
It is trained, obtains the risk judgment model of the risk classifications.
On in one embodiment, above-mentioned processor executes the record information input by the multiple event to convolution
The step of neural network model is trained, comprising: read putting on record the date in the record information；It is put on record the date according to described
And preset computation rule, calculate each process flow corresponding flow time stage；The note in record information is read respectively
The record date for recording data, according to the record date of each record data and each flow time stage, by the record data point
The label in corresponding flow time stage is not added；It, will be corresponding with the trained type according to the training type that user selects
The record data of label in flow time stage be input to convolutional neural networks model and be trained.
In one embodiment, above-mentioned processor executes the model system for receiving the target risk judgment models output
Number, and the step of risk case coefficient is calculated according to the model coefficient, comprising: read the relationship in the record information
People's information；Count the object event quantity of the event in preset time period with party's information；According to the target thing
Number of packages amount determines target weight coefficient in the corresponding relationship of event number and weight coefficient；Receive the target risk judgement
The model coefficient of model output obtains the risk case coefficient by the target weight coefficient multiplied by the model coefficient.
In one embodiment, above-mentioned processor executes the corresponding target risk of the target risk type that obtains and sentences
In disconnected model and the record information the step of target information corresponding with the target risk type before, comprising: reading
Target amount for which loss settled value in the record information；The mapping table of preset amount for which loss settled threshold value and risk classifications is called,
Obtain target amount for which loss settled threshold value corresponding with the target risk type；Judge whether the target amount for which loss settled value is greater than institute
State target amount for which loss settled threshold value；If so, executing the step: obtaining the corresponding target risk judgement of the target risk type
Target information corresponding with the target risk type in model and the record information.
In one embodiment, after above-mentioned processor executes described the step of determining the event for risk case, packet
It includes: the record information of the event being sent to the staff on list corresponding with the target risk type, with toilet
Staff is stated to handle the event.
In one embodiment, above-mentioned processor, which executes, described is sent to the record information of the event and the target
Before the step of staff on the corresponding list of risk classifications, comprising: encrypt the record information of the event.
Event is in conclusion the computer equipment of the application reads the record information of event automatically, and is selected according to user
Model by record information in corresponding data calculate, objectively help user calculate outgoing event risk system, have
Conducive to helping user quickly to determine whether the event is risk case, the time of user is saved, the judgement of risk case is improved
Accuracy rate.
It will be understood by those skilled in the art that structure shown in Fig. 3, only part relevant to application scheme is tied
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme.
One embodiment of the application also provides a kind of computer readable storage medium, is stored thereon with computer program, calculates
Machine program realizes a kind of method that risk case is judged based on neural network when being executed by processor, specifically: read event
Record information；Receive the target risk type that user terminal selects in multiple risk classifications；Obtain the target risk type pair
Target information corresponding with the target risk type in the target risk judgment models and the record information answered；By institute
Target information is stated to be input in the target risk judgment models；Receive the model system of the target risk judgment models output
Number, and risk case coefficient is calculated according to the model coefficient, the model coefficient refers to target risk judgment models root
The data obtained according to the associated information calculation of event；If the risk case coefficient is more than the preset and target risk type
Corresponding first risk threshold value then determines the event for risk case.
In one embodiment, above-mentioned processor execution is described is input to the target risk judgement for the target information
Before step in model, comprising: acquire the record information of multiple events, the multiple event includes multiple same risk classes
The risk case of type and multiple non-risk cases；By the record information input of the multiple event to convolutional neural networks model
It is trained, obtains the risk judgment model of the risk classifications.
On in one embodiment, above-mentioned processor executes the record information input by the multiple event to convolution
The step of neural network model is trained, comprising: read putting on record the date in the record information；It is put on record the date according to described
And preset computation rule, calculate each process flow corresponding flow time stage；The note in record information is read respectively
The record date for recording data, according to the record date of each record data and each flow time stage, by the record data point
The label in corresponding flow time stage is not added；It, will be corresponding with the trained type according to the training type that user selects
The record data of label in flow time stage be input to convolutional neural networks model and be trained.
In one embodiment, above-mentioned processor executes the model system for receiving the target risk judgment models output
Number, and the step of risk case coefficient is calculated according to the model coefficient, comprising: read the relationship in the record information
People's information；Count the object event quantity of the event in preset time period with party's information；According to the target thing
Number of packages amount determines target weight coefficient in the corresponding relationship of event number and weight coefficient；Receive the target risk judgement
The model coefficient of model output obtains the risk case coefficient by the target weight coefficient multiplied by the model coefficient.
In one embodiment, above-mentioned processor executes the corresponding target risk of the target risk type that obtains and sentences
In disconnected model and the record information the step of target information corresponding with the target risk type before, comprising: reading
Target amount for which loss settled value in the record information；The mapping table of preset amount for which loss settled threshold value and risk classifications is called,
Obtain target amount for which loss settled threshold value corresponding with the target risk type；Judge whether the target amount for which loss settled value is greater than institute
State target amount for which loss settled threshold value；If so, executing the step: obtaining the corresponding target risk judgement of the target risk type
Target information corresponding with the target risk type in model and the record information.
In one embodiment, after above-mentioned processor executes described the step of determining the event for risk case, packet
It includes: the record information of the event being sent to the staff on list corresponding with the target risk type, with toilet
Staff is stated to handle the event.
In one embodiment, above-mentioned processor, which executes, described is sent to the record information of the event and the target
Before the step of staff on the corresponding list of risk classifications, comprising: encrypt the record information of the event.
Event in conclusion the application computer readable storage medium, the automatic record information for reading event, and according to
The model of user's selection calculates the corresponding data recorded in information, and user is objectively helped to calculate the risk of outgoing event
System is conducive to that user is helped quickly to determine whether the event is risk case, saves the time of user, improves risk case
Determination rate of accuracy.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
Any reference used in provided herein and embodiment to memory, storage, database or other media,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double speed are according to rate SDRAM (SSRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that the process, device, article or the method that include a series of elements not only include those elements, and
And further include other elements that are not explicitly listed, or further include for this process, device, article or method institute it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including being somebody's turn to do
There is also other identical elements in the process, device of element, article or method.
The foregoing is merely preferred embodiment of the present application, are not intended to limit the scope of the patents of the application, all utilizations
Equivalent structure or equivalent flow shift made by present specification and accompanying drawing content is applied directly or indirectly in other correlations
Technical field, similarly include in the scope of patent protection of the application.
Claims (10)
1. a kind of method for judging risk case based on neural network characterized by comprising
The record information of reading event；
Receive the target risk type that user terminal selects in multiple risk classifications；
Obtain in the corresponding target risk judgment models of the target risk type and the record information with the target wind
The corresponding target information of dangerous type；
The target information is input in the target risk judgment models；
The model coefficient of the target risk judgment models output is received, and risk case is calculated according to the model coefficient
Coefficient, the model coefficient refer to the data that target risk judgment models are obtained according to the associated information calculation of event；
If the risk case coefficient is more than preset the first risk threshold value corresponding with the target risk type, institute is determined
Stating event is risk case.
2. the method for risk case is judged based on neural network as described in claim 1, which is characterized in that described by the mesh
Before mark information input to the step in the target risk judgment models, comprising:
Acquire the record information of multiple events, the multiple event includes the risk cases of multiple same risk classifications and more
A non-risk case；
The record information input of the multiple event is trained to convolutional neural networks model, obtains the risk classifications
Risk judgment model.
3. the method for risk case is judged based on neural network as claimed in claim 2, which is characterized in that it is described will be described more
The step of record information input of a event is trained to convolutional neural networks model, comprising:
Read putting on record the date in the record information；
According to put on record date and the preset computation rule, calculate when managing the corresponding process of process everywhere in the event
Between the stage；
The record date for reading the record data in record information respectively, according to the record date of each record data and each process
The record data are added the label in corresponding flow time stage by time phase respectively；
According to the training type that user selects, the record number of the label in the trained type corresponding flow time stage will be had
It is trained according to convolutional neural networks model is input to.
4. the method for risk case is judged based on neural network as described in claim 1, which is characterized in that described in the reception
The model coefficient of target risk judgment models output, and the step of risk case coefficient is calculated according to the model coefficient,
Include:
Read party's information in the record information；
Count the object event quantity of the event in preset time period with party's information；
According to the object event quantity, target weight coefficient is determined in the corresponding relationship of event number and weight coefficient；
The model coefficient for receiving the target risk judgment models output, by the target weight coefficient multiplied by the model system
Number, obtains the risk case coefficient.
5. the method for risk case is judged based on neural network as described in claim 1, which is characterized in that described in the acquisition
It is corresponding with the target risk type in the corresponding target risk judgment models of target risk type and the record information
Before the step of target information, comprising:
Read the target amount for which loss settled value in the record information；
The mapping table of preset amount for which loss settled threshold value and risk classifications is called, is obtained corresponding with the target risk type
Target amount for which loss settled threshold value；
Judge whether the target amount for which loss settled value is greater than the target amount for which loss settled threshold value；
If so, executing the step: obtaining the corresponding target risk judgment models of the target risk type and the note
Record target information corresponding with the target risk type in information.
6. a kind of device for judging risk case based on neural network characterized by comprising
First read module, for reading the record information of event；
Receiving module, the target risk type selected in multiple risk classifications for receiving user terminal；
First obtains module, for obtaining the corresponding target risk judgment models of the target risk type and the record
Target information corresponding with the target risk type in information；
Input module, for the target information to be input in the target risk judgment models；
Computing module, for receiving the model coefficient of the target risk judgment models output, and according to the model coefficient meter
Calculation obtains risk case coefficient, and the model coefficient refers to that target risk judgment models are obtained according to the associated information calculation of event
Data；
Determination module, if being more than preset the first risk corresponding with the target risk type for the risk case coefficient
Threshold value then determines the event for risk case.
7. the device of risk case is judged based on neural network as claimed in claim 6, which is characterized in that further include:
Acquisition module, for acquiring the record information of multiple events, the multiple event includes multiple same risk classifications
Risk case and multiple non-risk cases；
Training module is obtained for the record information input of the multiple event to be trained to convolutional neural networks model
The risk judgment model of the risk classifications.
8. the device of risk case is judged based on neural network as claimed in claim 7, which is characterized in that the training module
Include:
Date unit is read, for reading putting on record the date in the record information；
It is corresponding to calculate each process flow for date and the preset computation rule of putting on record according to for calculation stages unit
The flow time stage；
Adding unit, for reading the record date of the record data in record information respectively, according to the record of each record data
The record data are added the label in corresponding flow time stage by date and each flow time stage respectively；
Training unit, the training type for being selected according to user will have the trained type corresponding flow time stage
The record data of label be input to convolutional neural networks model and be trained.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the step of processor realizes any one of claims 1 to 5 the method when executing the computer program.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 5 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910314537.8A CN110136010A (en) | 2019-04-18 | 2019-04-18 | The method, apparatus and computer equipment of risk case are judged based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910314537.8A CN110136010A (en) | 2019-04-18 | 2019-04-18 | The method, apparatus and computer equipment of risk case are judged based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110136010A true CN110136010A (en) | 2019-08-16 |
Family
ID=67570312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910314537.8A Pending CN110136010A (en) | 2019-04-18 | 2019-04-18 | The method, apparatus and computer equipment of risk case are judged based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110136010A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826621A (en) * | 2019-11-01 | 2020-02-21 | 北京芯盾时代科技有限公司 | Risk event processing method and device |
| CN113515486A (en) * | 2020-04-10 | 2021-10-19 | 华晨宝马汽车有限公司 | Method, system and computer readable medium for event double check |
| CN116012169A (en) * | 2022-12-21 | 2023-04-25 | 南京睿聚科技发展有限公司 | Method and system for screening risk of insurance claim settlement based on position data |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105590261A (en) * | 2014-12-31 | 2016-05-18 | 中国银联股份有限公司 | Merchant risk estimation method and system |
| CN107240024A (en) * | 2017-05-22 | 2017-10-10 | 中国平安人寿保险股份有限公司 | The anti-fraud recognition methods of settlement of insurance claim and device |
| CN108256720A (en) * | 2017-11-07 | 2018-07-06 | 中国平安财产保险股份有限公司 | A kind of settlement of insurance claim methods of risk assessment and terminal device |
| CN109410036A (en) * | 2018-10-09 | 2019-03-01 | 北京芯盾时代科技有限公司 | A kind of fraud detection model training method and device and fraud detection method and device |
| CN109492095A (en) * | 2018-10-16 | 2019-03-19 | 平安健康保险股份有限公司 | Claims Resolution data processing method, device, computer equipment and storage medium |
-
2019
- 2019-04-18 CN CN201910314537.8A patent/CN110136010A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105590261A (en) * | 2014-12-31 | 2016-05-18 | 中国银联股份有限公司 | Merchant risk estimation method and system |
| CN107240024A (en) * | 2017-05-22 | 2017-10-10 | 中国平安人寿保险股份有限公司 | The anti-fraud recognition methods of settlement of insurance claim and device |
| CN108256720A (en) * | 2017-11-07 | 2018-07-06 | 中国平安财产保险股份有限公司 | A kind of settlement of insurance claim methods of risk assessment and terminal device |
| CN109410036A (en) * | 2018-10-09 | 2019-03-01 | 北京芯盾时代科技有限公司 | A kind of fraud detection model training method and device and fraud detection method and device |
| CN109492095A (en) * | 2018-10-16 | 2019-03-19 | 平安健康保险股份有限公司 | Claims Resolution data processing method, device, computer equipment and storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110826621A (en) * | 2019-11-01 | 2020-02-21 | 北京芯盾时代科技有限公司 | Risk event processing method and device |
| CN113515486A (en) * | 2020-04-10 | 2021-10-19 | 华晨宝马汽车有限公司 | Method, system and computer readable medium for event double check |
| CN113515486B (en) * | 2020-04-10 | 2024-03-08 | 华晨宝马汽车有限公司 | Method, system and computer readable medium for event duplication |
| CN116012169A (en) * | 2022-12-21 | 2023-04-25 | 南京睿聚科技发展有限公司 | Method and system for screening risk of insurance claim settlement based on position data |
| CN116012169B (en) * | 2022-12-21 | 2024-03-22 | 南京睿聚科技发展有限公司 | Method and system for screening risk of insurance claim settlement based on position data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11238176B1 (en) | System and methods for privacy management | |
| US20200004968A1 (en) | Data processing systems for data transfer risk identification and related methods | |
| US7818581B2 (en) | Data management system | |
| CN110136010A (en) | The method, apparatus and computer equipment of risk case are judged based on neural network | |
| CN108876133A (en) | Risk assessment processing method, device, server and medium based on business information | |
| CN107871083A (en) | Desensitize regular collocation method, application server and computer-readable recording medium | |
| US7693767B2 (en) | Method for generating predictive models for a business problem via supervised learning | |
| US20120036078A1 (en) | System and Method for Coordinating the Collection, Analysis and Storage of Payroll Information Provided to Government Agencies by Government Contractors | |
| CN113765881A (en) | Method and device for detecting abnormal network security behavior, electronic equipment and storage medium | |
| CN109801151B (en) | Financial falsification risk monitoring method, device, computer equipment and storage medium | |
| WO2004079539A2 (en) | System and method for generating and using a pooled knowledge base | |
| US20210351927A1 (en) | System, method and computer program product for mitigating customer onboarding risk | |
| WO2019210575A1 (en) | Real estate information enquiry method and apparatus, and computer device and storage medium | |
| CN109902747B (en) | Identity recognition method, device, equipment and computer readable storage medium | |
| CN110728301A (en) | Credit scoring method, device, terminal and storage medium for individual user | |
| CN109684863B (en) | Data leakage prevention method, device, equipment and storage medium | |
| WO2023006974A1 (en) | Optical fraud detector for automated detection of fraud in digital imaginary-based automobile claims, automated damage recognition, and method thereof | |
| CN114186275A (en) | Privacy protection method and device, computer equipment and storage medium | |
| Scala et al. | Evaluating mail‐based security for electoral processes using attack trees | |
| CN112036497A (en) | Method and device for training fraud case recognition model and computer equipment | |
| US20210224588A1 (en) | Recruitment process graph based unsupervised anomaly detection | |
| CN114338407B (en) | Operation and maintenance management method for enterprise information security | |
| CN111241139B (en) | Data statistical method, device, computer equipment and storage medium | |
| GB2544453A (en) | System for secure analysis of datasets | |
| CN113850456A (en) | Risk control management system based on alliance block chain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |