CN108416677A - The method and device of Claims Resolution investigation - Google Patents
The method and device of Claims Resolution investigation Download PDFInfo
- Publication number
- CN108416677A CN108416677A CN201710147700.7A CN201710147700A CN108416677A CN 108416677 A CN108416677 A CN 108416677A CN 201710147700 A CN201710147700 A CN 201710147700A CN 108416677 A CN108416677 A CN 108416677A
- Authority
- CN
- China
- Prior art keywords
- resolution
- risk
- error
- test
- training
- 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
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)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of method and device of Claims Resolution investigation, the method for the Claims Resolution investigation includes：After receiving the Claims Resolution for carrying subscriber identity information request, client characteristics information corresponding with the subscriber identity information is extracted from scheduled database；The Claims Resolution risk Severity Analysis model obtained using advance training analyzes the client characteristics information, show that corresponding risk of fraud probability value is asked in the Claims Resolution with analysis；If the risk of fraud probability value is more than predetermined probabilities threshold value, the instruction investigated for the Claims Resolution request is sent to scheduled terminal.The present invention can reduce Claims Resolution investigation rate, accurately lift investigation to Claims Resolution request, improve Claims Resolution efficiency.
Description
Technical field
The present invention relates to banking and insurance business technical field more particularly to a kind of method and devices of Claims Resolution investigation.
Background technology
In banking and insurance business field, before Claims Resolution claimant settles a claim, Claims Review operating personnel is needed to ask Claims Resolution
It asks and is audited to confirm whether Claims Resolution meets the requirements.And as the insurance awareness of the public constantly enhances, the insurance of insurance company
Business also constantly increases so that Claims Review operating personnel needs to handle a large amount of settlement of insurance claim case.Existing settlement of insurance claim
The audit scheme of request is：By Claims Review operating personnel audit by single case, Claims Resolution fraud in order to prevent, when to some
When Claims Resolution request is uncertain, Claims Review operating personnel can lift investigation.This existing Claims Resolution request checking method is easy to lead
It causes Claims Resolution investigation rate high, compensates that the time is long, and the Claims Resolution investigation lifted is due to subjectivity, it can not be accurately to reason
It pays for request and lifts investigation.
Invention content
The purpose of the present invention is to provide a kind of method and devices of Claims Resolution investigation, it is intended to reduce Claims Resolution investigation rate, accurately
Ground lifts investigation to Claims Resolution request, improves Claims Resolution efficiency.
To achieve the above object, the present invention provides a kind of method of Claims Resolution investigation, and the method for the Claims Resolution investigation includes：
S1, receive carry subscriber identity information Claims Resolution request after, extracted from scheduled database with it is described
The corresponding client characteristics information of subscriber identity information；
S2, the Claims Resolution risk Severity Analysis model obtained using advance training analyze the client characteristics information, with
Analysis show that corresponding risk of fraud probability value is asked in the Claims Resolution；
S3 sends to scheduled terminal if the risk of fraud probability value is more than predetermined probabilities threshold value and is directed to the reason
Pay for the instruction that request is investigated.
Preferably, include after the step S2：
Claims Resolution request is used as if the risk of fraud probability value is less than or equal to predetermined probabilities threshold value and waits taking out by S4
Charlie pay for request preserve, and timing wait for selective examination Claims Resolution request in randomly select preset ratio wait for selective examination Claims Resolution request,
It is sent to scheduled terminal for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
Preferably, the Claims Resolution risk Severity Analysis model is Random Forest model, includes before the step S2：
S201, obtains the Claims Resolution message sample of preset quantity compensated extremely, and the Claims Resolution message sample includes Claims Resolution visitor
The subscriber identity information and benchmark risk of fraud probability value at family；
S202 extracts client characteristics information corresponding with the subscriber identity information, by each reason from scheduled database
It pays for message sample corresponding client characteristics information aggregate and is divided into the training set of preset first ratio and preset second ratio
Test set；
S203 trains scheduled Random Forest model using each client characteristics information in the training set, is trained
Random Forest model afterwards surveys the Random Forest model after training using each client characteristics information in the test set
Examination；
S204, if test passes through, training terminates, and is divided using the Random Forest model after training as the Claims Resolution risk
Model is analysed, if not passing through alternatively, testing, increases the quantity of the Claims Resolution message sample compensated extremely and re-starts instruction
Practice.
Preferably, each client characteristics information using in the test set carries out the Random Forest model after training
The step of test includes：
Each client characteristics information in the test set is analyzed based on the Random Forest model after training, with analysis
Obtain the corresponding risk of fraud probabilistic testing value of each client characteristics information；
Calculate the mistake of the risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value of each client characteristics information
Difference analyzes each error according to scheduled error analysis rule, tests whether to pass through with determination.
Preferably, the scheduled error analysis rule is：
If there is at least one error to be more than or equal to preset first error threshold, it is determined that test does not pass through, and,
If all errors are respectively less than preset first error threshold, it is determined that test passes through；Or
If all errors are respectively less than the first default error threshold, the average error of all errors is calculated, if the mistake
Poor average value is less than or equal to the second default error threshold, it is determined that and test passes through, and, if the average error is more than second
Default error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the error more than or equal to the second default error threshold is calculated
Quantity account for all errors quantity ratio, if the ratio is less than or equal to default ratio, it is determined that test passes through, and,
If the ratio is more than default ratio, it is determined that test does not pass through, wherein the second default error threshold is less than described first
Default error threshold.
To achieve the above object, the present invention also provides a kind of device of Claims Resolution investigation, the device of the Claims Resolution investigation includes：
Extraction module, for after receiving the Claims Resolution for carrying subscriber identity information request, being carried from scheduled database
Take out client characteristics information corresponding with the subscriber identity information；
Analysis module, for using the obtained Claims Resolution risk Severity Analysis model of training in advance to the client characteristics information into
Row analysis show that corresponding risk of fraud probability value is asked in the Claims Resolution with analysis；
First inquiry module, if being more than predetermined probabilities threshold value for the risk of fraud probability value, to scheduled terminal
Send the instruction investigated for the Claims Resolution request.
Preferably, the device of the Claims Resolution investigation further includes：
Second inquiry module, if being less than or equal to predetermined probabilities threshold value for the risk of fraud probability value, by the reason
Pay for request as waits for selective examination Claims Resolution request preserved, and timing wait for selective examination Claims Resolution request in randomly select waiting for for preset ratio
Selective examination Claims Resolution request is sent to scheduled terminal for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
Preferably, the device of the Claims Resolution investigation further includes：
Acquisition module, the Claims Resolution message sample compensated extremely for obtaining preset quantity, the Claims Resolution message sample packet
Include the subscriber identity information and benchmark risk of fraud probability value of Claims Resolution client；
Processing module, for extracting client characteristics letter corresponding with the subscriber identity information from scheduled database
The corresponding client characteristics information aggregate of each Claims Resolution message sample is divided into the training set and preset the of preset first ratio by breath
The test set of two ratios；
Training module, for training scheduled Random Forest model using each client characteristics information in the training set,
Random Forest model after being trained, using each client characteristics information in the test set to the random forest mould after training
Type is tested；
Test module, if passing through for testing, training terminates, using the Random Forest model after training as the Claims Resolution
Risk Severity Analysis model increases the quantity and again of the Claims Resolution message sample compensated extremely if alternatively, test does not pass through
It is trained.
Preferably, the test module is specifically used for based on the Random Forest model after training to each in the test set
Client characteristics information is analyzed, and the corresponding risk of fraud probabilistic testing value of each client characteristics information is obtained with analysis；Meter
The error for calculating the risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value of each client characteristics information, according to predetermined
Error analysis rule each error is analyzed, with determination test whether to pass through.
Preferably, the scheduled error analysis rule is：
If there is at least one error to be more than or equal to preset first error threshold, it is determined that test does not pass through, and,
If all errors are respectively less than preset first error threshold, it is determined that test passes through；Or
If all errors are respectively less than the first default error threshold, the average error of all errors is calculated, if the mistake
Poor average value is less than or equal to the second default error threshold, it is determined that and test passes through, and, if the average error is more than second
Default error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the error more than or equal to the second default error threshold is calculated
Quantity account for all errors quantity ratio, if the ratio is less than or equal to default ratio, it is determined that test passes through, and,
If the ratio is more than default ratio, it is determined that test does not pass through, wherein the second default error threshold is less than described first
Default error threshold.
The beneficial effects of the invention are as follows：The present invention receive carry subscriber identity information Claims Resolution request after, extraction with
The corresponding client characteristics information of the subscriber identity information, the Claims Resolution risk Severity Analysis model analysis client obtained using advance training
Characteristic information, and the corresponding risk of fraud probability value of the client is obtained, it is preset if the risk of fraud probability value is more than
Probability threshold value, then it is assumed that the client has the risk of Claims Resolution fraud, and the instruction of investigation is sent to investigate the visitor to scheduled terminal
The Claims Resolution at family is asked, and the present invention does not audit Claims Resolution request by Claims Review operating personnel, but passes through risk Severity Analysis of settling a claim
The client characteristics information of model analysis client, to obtain the risk of fraud probability of the client, to according to the risk of fraud probability
Determine whether that the Claims Resolution for investigating the client is asked, Claims Resolution investigation rate can be reduced to a certain extent, and can be accurately to Claims Resolution
Request lifts investigation, improves Claims Resolution efficiency.
Description of the drawings
Fig. 1 is the flow diagram of the method first embodiment of present invention Claims Resolution investigation；
Fig. 2 is the flow diagram of the method second embodiment of present invention Claims Resolution investigation；
Fig. 3 is the flow diagram of the method 3rd embodiment of present invention Claims Resolution investigation；
Fig. 4 is the structural schematic diagram of the device first embodiment of present invention Claims Resolution investigation；
Fig. 5 is the structural schematic diagram of the device first embodiment of present invention Claims Resolution investigation.
Specific implementation mode
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the present invention.
The flow diagram for one embodiment of method of investigation of settling a claim as shown in FIG. 1, FIG. 1 is the present invention, Claims Resolution investigation
Method includes the following steps：
Step S1, receive carry subscriber identity information Claims Resolution request after, extracted from scheduled database with
The corresponding client characteristics information of the subscriber identity information；
In the present embodiment, client initiates Claims Resolution request to Claims Resolution server, and Claims Resolution server reception carries customer ID
The Claims Resolution of information is asked, which can be identification card number, passport No. or insurance odd numbers of client etc..Claims Resolution service
Device extracts client characteristics information corresponding with subscriber identity information from scheduled database, such as from insurance business data
Corresponding client characteristics information is extracted in library, banking business data or collage-credit data library etc..
Wherein, client characteristics information includes but is not limited to client properties information, client's financial situation information, settlement of insurance claim
Enliven information, values information and reference information.Client properties information includes age, gender, occupation, annual income and native place etc.,
Client's financial situation information includes do not repay information, credit card debt information etc., and it includes nearest one that settlement of insurance claim, which enlivens information,
It is proposed number, Claims Resolution amount etc. of Claims Resolution year, values information includes to (values information such as the pessimistic or optimistic degree of society
Can be by being obtained in client's transacting business by allowing client to fill in evaluation questionnaire analysis), reference information includes that whether there is
(reference information can be from predetermined collage-credit data library " for example, the personal collage-credit data of People's Bank of China for situation of breaking one's promise etc.
Library " obtains).
Step S2, the Claims Resolution risk Severity Analysis model obtained using advance training divide the client characteristics information
Analysis show that corresponding risk of fraud probability value is asked in the Claims Resolution with analysis；
In the present embodiment, training in advance obtains a Claims Resolution risk Severity Analysis model, the Claims Resolution risk which obtains point
The parameter of analysis model can determine.The Claims Resolution risk Severity Analysis model can analyze client characteristics information, and according to point
Analysis result provides corresponding risk of fraud probability value, it is preferable that the Claims Resolution risk Severity Analysis model is Random Forest model, certainly
Can also other models analyze to obtain corresponding risk of fraud probability value.
It wherein, can be by client characteristics information input to the model, by more in model by taking Random Forest model as an example
Decision tree is trained the client characteristics information of extraction, is trained from the root node of each decision tree up in a certain section
Point reaches end condition, and every decision tree exports predicted value after training, and the predicted value exported by every decision tree can be counted
Calculation obtains risk of fraud probability value.
Step S3 sends to scheduled terminal if the risk of fraud probability value is more than predetermined probabilities threshold value and is directed to institute
State the instruction that Claims Resolution request is investigated.
In the present embodiment, corresponding risk of fraud probability value is obtained in Claims Resolution risk Severity Analysis model analysis, if this is taken advantage of
It cheats threat probability values and is more than predetermined probabilities threshold value, be greater than 0.6, then it is assumed that the corresponding client of the client characteristics information has reason
The risk of fraud is paid for, investigation instruction can be sent from Claims Resolution server to predetermined terminal, with immediately to the reason of the client
Request is paid for be investigated.
Compared with prior art, the present embodiment is extracted and is somebody's turn to do after receiving the Claims Resolution for carrying subscriber identity information request
The corresponding client characteristics information of subscriber identity information, the Claims Resolution risk Severity Analysis model analysis client obtained using advance training are special
Reference ceases, and obtains the corresponding risk of fraud probability value of the client, is preset generally if the risk of fraud probability value is more than
Rate threshold value, then it is assumed that the client has the risk of Claims Resolution fraud, and the instruction of investigation is sent to investigate the client to scheduled terminal
Claims Resolution request, the present embodiment not by Claims Review operating personnel audit Claims Resolution request, but pass through settle a claim risk Severity Analysis
The client characteristics information of model analysis client, to obtain the risk of fraud probability of the client, to according to the risk of fraud probability
Determine whether that the Claims Resolution for investigating the client is asked, Claims Resolution investigation rate can be reduced to a certain extent, and can be accurately to Claims Resolution
Request lifts investigation, improves Claims Resolution efficiency.
In a preferred embodiment, as shown in Fig. 2, on the basis of the embodiment of above-mentioned Fig. 1, above-mentioned steps S2 it
After include：
Claims Resolution request is used as if the risk of fraud probability value is less than or equal to predetermined probabilities threshold value and waits taking out by S4
Charlie pay for request preserve, and timing wait for selective examination Claims Resolution request in randomly select preset ratio wait for selective examination Claims Resolution request,
It is sent to scheduled terminal for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
In the present embodiment, if risk of fraud probability value is less than or equal to predetermined probabilities threshold value (being, for example, less than 0.6), then it is assumed that
The client there is no the risk of Claims Resolution fraud, Claims Resolution request need not be investigated immediately (but the flow of Claims Resolution can be started,
To enable a customer to be compensated as early as possible), can be by it to wait for that selective examination Claims Resolution request preserves, then timing is waiting for selective examination Claims Resolution
That preset ratio (such as 10%) is randomly selected in request waits for selective examination Claims Resolution request, and investigation instruction is sent to scheduled terminal, with
Selective examination Claims Resolution request, which is investigated, to be waited for these extracted.
The Claims Resolution request that the present embodiment is less than or equal to risk of fraud probability predetermined probabilities threshold value is done timing selective examination and is adjusted
Investigate and prosecute reason so that Claims Resolution investigation is more comprehensive.
In a preferred embodiment, as shown in figure 3, on the basis of the embodiment of above-mentioned Fig. 1, the step S2 it
Before include：
S201 obtains the Claims Resolution message sample of preset quantity compensated extremely；
S202 extracts client characteristics information corresponding with the subscriber identity information, by each reason from scheduled database
It pays for message sample corresponding client characteristics information aggregate and is divided into the training set of preset first ratio and preset second ratio
Test set；
S203 trains scheduled Random Forest model using each client characteristics information in the training set, is trained
Random Forest model afterwards surveys the Random Forest model after training using each client characteristics information in the test set
Examination；
S204, if test passes through, training terminates, and is divided using the Random Forest model after training as the Claims Resolution risk
Model is analysed, if not passing through alternatively, testing, increases the quantity of the Claims Resolution message sample compensated extremely and re-starts instruction
Practice.
In the present embodiment, the Claims Resolution message sample of preset quantity (for example, 100) compensated extremely is obtained, it is abnormal to compensate
It generally refers to refuse to pay and improper pay.It is each Claims Resolution message sample in further include settle a claim client subscriber identity information
And benchmark risk of fraud probability value, the benchmark risk of fraud probability value can be that Claims Review operating personnel manages according to customer historical
Case is paid for analyze to obtain.Client characteristics letter corresponding with the subscriber identity information is extracted from predetermined database
Breath, the client characteristics information include but is not limited to that client properties information, client's financial situation information, settlement of insurance claim are actively believed
Breath, values information and/or reference information.Then, the corresponding client characteristics information of each Claims Resolution message sample is divided into default
The training set of the first ratio (for example, 70%), preset second ratio (for example, 30%) test set, using in training set
Each client characteristics information train advance Random Forest model to utilize test set to obtain trained Random Forest model
In each client characteristics information trained Random Forest model is tested, if test pass through, training terminate, should
Random Forest model after test can be applied to practical operation, if test does not pass through, increase the Claims Resolution information compensated extremely
The quantity of sample, and operations of the above-mentioned steps S201 to step S204 is re-executed, until test passes through.
The present embodiment by compensate extremely Claims Resolution message sample carry out client characteristics information extraction, to training set into
Row training and test set is carried out the operation such as to test, to training obtain accurately analyzing the risk of fraud probability of client with
Machine forest model.
Preferably, utilize each client characteristics information in the test set to random gloomy after training in above-mentioned steps S203
The step of woods model is tested include：Each client characteristics in the test set are believed based on the Random Forest model after training
Breath is analyzed, and the corresponding risk of fraud probabilistic testing value of each client characteristics information is obtained with analysis；It is special to calculate each client
The error of the risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value of reference breath, according to scheduled error analysis
Rule analyzes each error, tests whether to pass through with determination.
In the present embodiment, by the Random Forest model after the client characteristics information input to training in test set, pass through
Random Forest model after training analyzes the client characteristics information in test set to obtain risk of fraud probabilistic testing value, so
The error of risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value is calculated afterwards, such as calculates the formula of error
For：Error=(risk of fraud probabilistic testing value-benchmark risk of fraud probability value)/benchmark risk of fraud probability value × 100%, with
It tests whether to pass through according to the error analysis.
Preferably, above-mentioned scheduled error analysis rule is：If there is at least one error to be more than or equal to preset the
One error threshold (for example, 0.03), it is determined that test does not pass through, and, if all errors are respectively less than preset first error threshold
Value, it is determined that test passes through；Or
If all errors are respectively less than the first default error threshold, the average error of all errors is calculated, if the mistake
Poor average value is less than or equal to the second default error threshold (0.02), it is determined that and test passes through, and, if the average error is big
In the second default error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the error more than or equal to the second default error threshold is calculated
Quantity account for all errors quantity ratio, if the ratio is less than or equal to default ratio (such as 0.5), it is determined that test is logical
It crosses, and, if the ratio is more than default ratio, it is determined that test does not pass through, wherein the second above-mentioned default error threshold is small
In the first default error threshold.
As shown in figure 4, Fig. 4 is the structural schematic diagram of one embodiment of device of present invention Claims Resolution investigation, Claims Resolution investigation
Device includes：
Extraction module 101, after being asked in the Claims Resolution for receiving carrying subscriber identity information, from scheduled database
Extract client characteristics information corresponding with the subscriber identity information；
In the present embodiment, client initiates Claims Resolution request to Claims Resolution server, and Claims Resolution server reception carries customer ID
The Claims Resolution of information is asked, which can be identification card number, passport No. or insurance odd numbers of client etc..Claims Resolution service
Device extracts client characteristics information corresponding with subscriber identity information from scheduled database, such as from insurance business data
Corresponding client characteristics information is extracted in library, banking business data or collage-credit data library etc..
Wherein, client characteristics information includes but is not limited to client properties information, client's financial situation information, settlement of insurance claim
Enliven information, values information and reference information.Client properties information includes age, gender, occupation, annual income and native place etc.,
Client's financial situation information includes do not repay information, credit card debt information etc., and it includes nearest one that settlement of insurance claim, which enlivens information,
It is proposed number, Claims Resolution amount etc. of Claims Resolution year, values information includes to (values information such as the pessimistic or optimistic degree of society
Can be by being obtained in client's transacting business by allowing client to fill in evaluation questionnaire analysis), reference information includes that whether there is
(reference information can be from predetermined collage-credit data library " for example, the personal collage-credit data of People's Bank of China for situation of breaking one's promise etc.
Library " obtains).
Analysis module 102, for being believed the client characteristics using the Claims Resolution risk Severity Analysis model that training obtains in advance
Breath is analyzed, and show that corresponding risk of fraud probability value is asked in the Claims Resolution with analysis；
In the present embodiment, training in advance obtains a Claims Resolution risk Severity Analysis model, the Claims Resolution risk which obtains point
The parameter of analysis model can determine.The Claims Resolution risk Severity Analysis model can analyze client characteristics information, and according to point
Analysis result provides corresponding risk of fraud probability value, it is preferable that the Claims Resolution risk Severity Analysis model is Random Forest model, certainly
Can also other models analyze to obtain corresponding risk of fraud probability value.
It wherein, can be by client characteristics information input to the model, by more in model by taking Random Forest model as an example
Decision tree is trained the client characteristics information of extraction, is trained from the root node of each decision tree up in a certain section
Point reaches end condition, and every decision tree exports predicted value after training, and the predicted value exported by every decision tree can be counted
Calculation obtains risk of fraud probability value.
First inquiry module 103, if being more than predetermined probabilities threshold value for the risk of fraud probability value, to scheduled end
End sends the instruction investigated for the Claims Resolution request.
In the present embodiment, corresponding risk of fraud probability value is obtained in Claims Resolution risk Severity Analysis model analysis, if this is taken advantage of
It cheats threat probability values and is more than predetermined probabilities threshold value, be greater than 0.6, then it is assumed that the corresponding client of the client characteristics information has reason
The risk of fraud is paid for, investigation instruction can be sent from Claims Resolution server to predetermined terminal, with immediately to the reason of the client
Request is paid for be investigated.
In a preferred embodiment, as shown in figure 5, on the basis of the embodiment of above-mentioned Fig. 4, above-mentioned Claims Resolution investigation
Device further includes：
Second inquiry module 104 will be described if being less than or equal to predetermined probabilities threshold value for the risk of fraud probability value
Claims Resolution request as waits for selective examination Claims Resolution request preserved, and timing wait for selective examination Claims Resolution request in randomly select preset ratio
It waits for selective examination Claims Resolution request, is sent to scheduled terminal for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
In the present embodiment, if risk of fraud probability value is less than or equal to predetermined probabilities threshold value (being, for example, less than 0.6), then it is assumed that
The client there is no the risk of Claims Resolution fraud, Claims Resolution request need not be investigated immediately (but the flow of Claims Resolution can be started,
To enable a customer to be compensated as early as possible), can be by it to wait for that selective examination Claims Resolution request preserves, then timing is waiting for selective examination Claims Resolution
That preset ratio (such as 10%) is randomly selected in request waits for selective examination Claims Resolution request, and investigation instruction is sent to scheduled terminal, with
Selective examination Claims Resolution request, which is investigated, to be waited for these extracted.
The Claims Resolution request that the present embodiment is less than or equal to risk of fraud probability predetermined probabilities threshold value is done timing selective examination and is adjusted
Investigate and prosecute reason so that Claims Resolution investigation is more comprehensive.
In a preferred embodiment, on the basis of the embodiment of above-mentioned Fig. 4, the device of the Claims Resolution investigation also wraps
It includes：
Acquisition module, the Claims Resolution message sample compensated extremely for obtaining preset quantity, the Claims Resolution message sample packet
Include the subscriber identity information and benchmark risk of fraud probability value of Claims Resolution client；
Processing module, for extracting client characteristics letter corresponding with the subscriber identity information from scheduled database
The corresponding client characteristics information aggregate of each Claims Resolution message sample is divided into the training set and preset the of preset first ratio by breath
The test set of two ratios；
Training module, for training scheduled Random Forest model using each client characteristics information in the training set,
Random Forest model after being trained, using each client characteristics information in the test set to the random forest mould after training
Type is tested；
Test module, if passing through for testing, training terminates, using the Random Forest model after training as the Claims Resolution
Risk Severity Analysis model increases the quantity and again of the Claims Resolution message sample compensated extremely if alternatively, test does not pass through
It is trained.
In the present embodiment, the Claims Resolution message sample of preset quantity (for example, 100) compensated extremely is obtained, it is abnormal to compensate
It generally refers to refuse to pay and improper pay.It is each Claims Resolution message sample in further include settle a claim client subscriber identity information
And benchmark risk of fraud probability value, the benchmark risk of fraud probability value can be that Claims Review operating personnel manages according to customer historical
Case is paid for analyze to obtain.Client characteristics letter corresponding with the subscriber identity information is extracted from predetermined database
Breath, the client characteristics information include but is not limited to that client properties information, client's financial situation information, settlement of insurance claim are actively believed
Breath, values information and/or reference information.Then, the corresponding client characteristics information of each Claims Resolution message sample is divided into default
The training set of the first ratio (for example, 70%), preset second ratio (for example, 30%) test set, using in training set
Each client characteristics information train advance Random Forest model to utilize test set to obtain trained Random Forest model
In each client characteristics information trained Random Forest model is tested, if test pass through, training terminate, should
Random Forest model after test can be applied to practical operation, if test does not pass through, increase the Claims Resolution information compensated extremely
The quantity of sample, and aforesaid operations are re-executed, until test passes through.
The present embodiment by compensate extremely Claims Resolution message sample carry out client characteristics information extraction, to training set into
Row training and test set is carried out the operation such as to test, to training obtain accurately analyzing the risk of fraud probability of client with
Machine forest model.
Preferably, above-mentioned test module is specifically used for based on the Random Forest model after training to each in the test set
Client characteristics information is analyzed, and the corresponding risk of fraud probabilistic testing value of each client characteristics information is obtained with analysis；Meter
The error for calculating the risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value of each client characteristics information, according to predetermined
Error analysis rule each error is analyzed, with determination test whether to pass through.
In the present embodiment, by the Random Forest model after the client characteristics information input to training in test set, pass through
Random Forest model after training analyzes the client characteristics information in test set to obtain risk of fraud probabilistic testing value, so
The error of risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value is calculated afterwards, such as calculates the formula of error
For：Error=(risk of fraud probabilistic testing value-benchmark risk of fraud probability value)/benchmark risk of fraud probability value × 100%, with
It tests whether to pass through according to the error analysis.
Preferably, the scheduled error analysis rule is：If there is at least one error to be more than or equal to preset the
One error threshold, it is determined that test does not pass through, and, if all errors are respectively less than preset first error threshold, it is determined that survey
It pinged；Or
If all errors are respectively less than the first default error threshold, the average error of all errors is calculated, if the mistake
Poor average value is less than or equal to the second default error threshold, it is determined that and test passes through, and, if the average error is more than second
Default error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the error more than or equal to the second default error threshold is calculated
Quantity account for all errors quantity ratio, if the ratio is less than or equal to default ratio, it is determined that test passes through, and,
If the ratio is more than default ratio, it is determined that test does not pass through, wherein the second default error threshold is less than described first
Default error threshold.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (10)
1. it is a kind of Claims Resolution investigation method, which is characterized in that it is described Claims Resolution investigation method include：
S1 is extracted and the client after receiving the Claims Resolution for carrying subscriber identity information request from scheduled database
Client characteristics information corresponding to identification information；
S2, the Claims Resolution risk Severity Analysis model obtained using advance training analyzes the client characteristics information, with analysis
Show that corresponding risk of fraud probability value is asked in the Claims Resolution；
S3 sends to scheduled terminal and is asked for the Claims Resolution if the risk of fraud probability value is more than predetermined probabilities threshold value
Ask the instruction investigated.
2. the method for Claims Resolution investigation according to claim 1, which is characterized in that include after the step S2：
Claims Resolution request is used as if the risk of fraud probability value is less than or equal to predetermined probabilities threshold value and waits for selective examination reason by S4
Request is paid for be preserved, and timing randomly select preset ratio in waiting for selective examination Claims Resolution request wait for selective examination Claims Resolution request, to pre-
Fixed terminal is sent for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
3. the method for Claims Resolution investigation according to claim 1 or 2, which is characterized in that the Claims Resolution risk Severity Analysis model
For Random Forest model, the step S2 includes before：
S201, obtains the Claims Resolution message sample of preset quantity compensated extremely, and the Claims Resolution message sample includes Claims Resolution client
Subscriber identity information and benchmark risk of fraud probability value；
S202 extracts client characteristics information corresponding with the subscriber identity information from scheduled database, and each Claims Resolution is believed
The corresponding client characteristics information aggregate of breath sample is divided into the training set of preset first ratio and the test of preset second ratio
Collection；
S203 trains scheduled Random Forest model, after being trained using each client characteristics information in the training set
Random Forest model tests the Random Forest model after training using each client characteristics information in the test set；
S204, if test passes through, training terminates, using the Random Forest model after training as the Claims Resolution risk Severity Analysis mould
Type increases the quantity of the Claims Resolution message sample compensated extremely and is trained again if not passing through alternatively, testing.
4. the method for Claims Resolution investigation according to claim 3, which is characterized in that each visitor using in the test set
The step of family characteristic information tests the Random Forest model after training include：
Each client characteristics information in the test set is analyzed based on the Random Forest model after training, is obtained with analysis
The corresponding risk of fraud probabilistic testing value of each client characteristics information；
Calculate the error of the risk of fraud probabilistic testing value and corresponding benchmark risk of fraud probability value of each client characteristics information, root
Each error is analyzed according to scheduled error analysis rule, tests whether to pass through with determination.
5. the method for Claims Resolution investigation according to claim 4, which is characterized in that the scheduled error analysis rule is：
If there is at least one error to be more than or equal to preset first error threshold, it is determined that test does not pass through, and, if institute
There is error to be respectively less than preset first error threshold, it is determined that test passes through；If all errors are respectively less than the first default mistake
Poor threshold value then calculates the average error of all errors, if the average error is less than or equal to the second default error threshold,
Determine that test passes through, and, if the average error is more than the second default error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the number of the error more than or equal to the second default error threshold is calculated
Amount accounts for the ratio of the quantity of all errors, if the ratio is less than or equal to default ratio, it is determined that and test passes through, and, if institute
It states ratio and is more than default ratio, it is determined that test does not pass through, wherein it is default that the second default error threshold is less than described first
Error threshold.
6. it is a kind of Claims Resolution investigation device, which is characterized in that it is described Claims Resolution investigation device include：
Extraction module, for after receiving the Claims Resolution for carrying subscriber identity information request, being extracted from scheduled database
Client characteristics information corresponding with the subscriber identity information；
Analysis module, the Claims Resolution risk Severity Analysis model for being obtained using advance training divide the client characteristics information
Analysis show that corresponding risk of fraud probability value is asked in the Claims Resolution with analysis；
First inquiry module is sent if being more than predetermined probabilities threshold value for the risk of fraud probability value to scheduled terminal
The instruction investigated for the Claims Resolution request.
7. it is according to claim 6 Claims Resolution investigation device, which is characterized in that it is described Claims Resolution investigation device further include：
Second inquiry module asks the Claims Resolution if being less than or equal to predetermined probabilities threshold value for the risk of fraud probability value
It asks as waiting for that selective examination Claims Resolution request is preserved, and timing randomly selects preset ratio in waiting for selective examination Claims Resolution request and waits spot-check
Claims Resolution request is sent to scheduled terminal for the instruction for respectively waiting for selective examination Claims Resolution request and being investigated extracted.
8. the device of the Claims Resolution investigation described according to claim 6 or 7, which is characterized in that the device of the Claims Resolution investigation also wraps
It includes：
Acquisition module, the Claims Resolution message sample compensated extremely for obtaining preset quantity, the Claims Resolution message sample include reason
Pay for the subscriber identity information and benchmark risk of fraud probability value of client；
Processing module will for extracting client characteristics information corresponding with the subscriber identity information from scheduled database
Each corresponding client characteristics information aggregate of message sample of settling a claim is divided into the training set of preset first ratio and preset second ratio
The test set of example；
Training module is obtained for training scheduled Random Forest model using each client characteristics information in the training set
Random Forest model after training, using each client characteristics information in the test set to the Random Forest model after training into
Row test；
Test module, if passing through for testing, training terminates, using the Random Forest model after training as the Claims Resolution risk
Analysis model is spent, if not passing through alternatively, testing, increases the quantity of the Claims Resolution message sample compensated extremely and re-starts
Training.
9. the device of Claims Resolution investigation according to claim 8, which is characterized in that the test module is specifically used for based on instruction
Random Forest model after white silk analyzes each client characteristics information in the test set, and each client is obtained with analysis
The corresponding risk of fraud probabilistic testing value of characteristic information；Calculate the risk of fraud probabilistic testing value of each client characteristics information with it is corresponding
Benchmark risk of fraud probability value error, each error is analyzed according to scheduled error analysis rule, with determine test
Whether pass through.
10. the device of Claims Resolution investigation according to claim 9, which is characterized in that the scheduled error analysis rule is：
If there is at least one error to be more than or equal to preset first error threshold, it is determined that test does not pass through, and, if institute
There is error to be respectively less than preset first error threshold, it is determined that test passes through；Or
If all errors are respectively less than the first default error threshold, the average error of all errors is calculated, if the error is flat
Mean value is less than or equal to the second default error threshold, it is determined that and test passes through, and, it is preset if the average error is more than second
Error threshold, it is determined that test does not pass through；Or
If all errors are respectively less than the first default error threshold, the number of the error more than or equal to the second default error threshold is calculated
Amount accounts for the ratio of the quantity of all errors, if the ratio is less than or equal to default ratio, it is determined that and test passes through, and, if institute
It states ratio and is more than default ratio, it is determined that test does not pass through, wherein it is default that the second default error threshold is less than described first
Error threshold.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710147700.7A CN108416677A (en) | 2017-03-13 | 2017-03-13 | The method and device of Claims Resolution investigation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710147700.7A CN108416677A (en) | 2017-03-13 | 2017-03-13 | The method and device of Claims Resolution investigation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108416677A true CN108416677A (en) | 2018-08-17 |
Family
ID=63125321
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710147700.7A Pending CN108416677A (en) | 2017-03-13 | 2017-03-13 | The method and device of Claims Resolution investigation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108416677A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360109A (en) * | 2018-09-26 | 2019-02-19 | 平安医疗健康管理股份有限公司 | Intelligent Claims Review householder method and system based on big data analysis and processing |
| CN109377361A (en) * | 2018-09-18 | 2019-02-22 | 中国平安财产保险股份有限公司 | The building method and device in the settlement of insurance claim supervision library based on big data analysis |
| CN109509106A (en) * | 2018-10-30 | 2019-03-22 | 平安科技（深圳）有限公司 | Flat type determines method and Related product |
| CN109859060A (en) * | 2019-01-31 | 2019-06-07 | 泰康保险集团股份有限公司 | Risk determines method, apparatus, medium and electronic equipment |
| CN109872071A (en) * | 2019-02-22 | 2019-06-11 | 德联易控科技(北京)有限公司 | Auditing method, device, equipment and the storage medium of vehicle insurance Claims Resolution case |
| CN110399818A (en) * | 2019-07-15 | 2019-11-01 | 联动优势科技有限公司 | A kind of method and apparatus of risk profile |
| CN110827157A (en) * | 2019-11-05 | 2020-02-21 | 泰康保险集团股份有限公司 | Data processing method and device, storage medium and electronic equipment |
| CN111524017A (en) * | 2019-02-01 | 2020-08-11 | 天津幸福生命科技有限公司 | Claim data auditing method, device and medium |
| CN111861767A (en) * | 2020-07-29 | 2020-10-30 | 贵州力创科技发展有限公司 | System and method for monitoring vehicle insurance fraud behaviors |
| CN111861765A (en) * | 2020-07-29 | 2020-10-30 | 贵州力创科技发展有限公司 | Intelligent anti-fraud method for vehicle insurance claim settlement |
| CN112116356A (en) * | 2020-09-28 | 2020-12-22 | 中国银行股份有限公司 | Asset characteristic information processing method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101925919A (en) * | 2007-11-28 | 2010-12-22 | 安信龙股份公司 | Automated claims processing system |
| CN102013084A (en) * | 2010-12-14 | 2011-04-13 | 江苏大学 | System and method for detecting fraudulent transactions in medical insurance outpatient services |
| US20140164025A1 (en) * | 2001-08-15 | 2014-06-12 | Meridian Enterprises Corporation | Insurance claim payment card system |
| CN106295151A (en) * | 2016-08-03 | 2017-01-04 | 合肥奇也信息科技有限公司 | A kind of medical care insurance is settled a claim system online |
-
2017
- 2017-03-13 CN CN201710147700.7A patent/CN108416677A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140164025A1 (en) * | 2001-08-15 | 2014-06-12 | Meridian Enterprises Corporation | Insurance claim payment card system |
| CN101925919A (en) * | 2007-11-28 | 2010-12-22 | 安信龙股份公司 | Automated claims processing system |
| CN102013084A (en) * | 2010-12-14 | 2011-04-13 | 江苏大学 | System and method for detecting fraudulent transactions in medical insurance outpatient services |
| CN106295151A (en) * | 2016-08-03 | 2017-01-04 | 合肥奇也信息科技有限公司 | A kind of medical care insurance is settled a claim system online |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109377361A (en) * | 2018-09-18 | 2019-02-22 | 中国平安财产保险股份有限公司 | The building method and device in the settlement of insurance claim supervision library based on big data analysis |
| CN109360109A (en) * | 2018-09-26 | 2019-02-19 | 平安医疗健康管理股份有限公司 | Intelligent Claims Review householder method and system based on big data analysis and processing |
| CN109509106A (en) * | 2018-10-30 | 2019-03-22 | 平安科技（深圳）有限公司 | Flat type determines method and Related product |
| CN109509106B (en) * | 2018-10-30 | 2024-07-19 | 平安科技（深圳）有限公司 | Unit type determining method and related product |
| CN109859060B (en) * | 2019-01-31 | 2021-09-10 | 泰康保险集团股份有限公司 | Risk determination method, risk determination device, risk determination medium and electronic equipment |
| CN109859060A (en) * | 2019-01-31 | 2019-06-07 | 泰康保险集团股份有限公司 | Risk determines method, apparatus, medium and electronic equipment |
| CN111524017A (en) * | 2019-02-01 | 2020-08-11 | 天津幸福生命科技有限公司 | Claim data auditing method, device and medium |
| CN111524017B (en) * | 2019-02-01 | 2023-09-22 | 北京懿医云科技有限公司 | Method, device and medium for auditing of claim data |
| CN109872071A (en) * | 2019-02-22 | 2019-06-11 | 德联易控科技(北京)有限公司 | Auditing method, device, equipment and the storage medium of vehicle insurance Claims Resolution case |
| CN110399818A (en) * | 2019-07-15 | 2019-11-01 | 联动优势科技有限公司 | A kind of method and apparatus of risk profile |
| CN110827157B (en) * | 2019-11-05 | 2023-04-07 | 泰康保险集团股份有限公司 | Data processing method and device, storage medium and electronic equipment |
| CN110827157A (en) * | 2019-11-05 | 2020-02-21 | 泰康保险集团股份有限公司 | Data processing method and device, storage medium and electronic equipment |
| CN111861765A (en) * | 2020-07-29 | 2020-10-30 | 贵州力创科技发展有限公司 | Intelligent anti-fraud method for vehicle insurance claim settlement |
| CN111861767A (en) * | 2020-07-29 | 2020-10-30 | 贵州力创科技发展有限公司 | System and method for monitoring vehicle insurance fraud behaviors |
| CN111861765B (en) * | 2020-07-29 | 2024-07-12 | 贵州力创科技发展有限公司 | Intelligent anti-fraud method for vehicle insurance claim settlement |
| CN112116356A (en) * | 2020-09-28 | 2020-12-22 | 中国银行股份有限公司 | Asset characteristic information processing method and device |
| CN112116356B (en) * | 2020-09-28 | 2023-11-21 | 中国银行股份有限公司 | Asset characteristic information processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416677A (en) | The method and device of Claims Resolution investigation | |
| CN109829776B (en) | Merchant risk assessment method, device, computer equipment and storage medium | |
| US8417600B2 (en) | Systems and methods for graduated suspicious activity detection | |
| CN107240024A (en) | The anti-fraud recognition methods of settlement of insurance claim and device | |
| CN109509093B (en) | Transaction security control method and system based on main body portrait | |
| CN106780012A (en) | A kind of internet credit methods and system | |
| US20100005013A1 (en) | Methods and systems for detecting fraudulent transactions in a customer-not-present environment | |
| CN101236638A (en) | Web based bank card risk monitoring method and system | |
| WO2010037030A1 (en) | Evaluating loan access using online business transaction data | |
| US20130018789A1 (en) | Systems and methods for estimating the risk that a real-time promissory payment will default | |
| CA2927849A1 (en) | Intelligent collections models | |
| CN110335144B (en) | Personal electronic bank account security detection method and device | |
| CN111598682A (en) | Credit risk assessment method and system for enterprise | |
| CN108876105A (en) | A kind of transaction risk control method and device | |
| CN109472491A (en) | A kind of inspection detection rating business credit software systems | |
| CN113034000A (en) | Wind control processing method and device, computing equipment and storage medium | |
| CN112330328A (en) | Credit card fraud detection method based on feature extraction | |
| CN110728570B (en) | Anti-fraud fund analysis method | |
| CN110991650A (en) | Method and device for training card maintenance identification model and identifying card maintenance behavior | |
| CN113065945A (en) | Method and system for classifying repayment willingness of customer for collection, verification and sale | |
| CN114998014A (en) | Information processing and analyzing method based on block chain | |
| CN113888321A (en) | Default probability analysis method based on third-party data source | |
| CN112823502B (en) | Real-time feedback service for resource access rule configuration | |
| CN107871275B (en) | Business approval method and device | |
| CN117952619B (en) | Risk behavior analysis method, system and computer readable medium based on digital RMB wallet account correlation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180817 |