CN105512938A - Online credit risk assessment method based on long-term using behavior of user - Google Patents
Online credit risk assessment method based on long-term using behavior of user Download PDFInfo
- Publication number
- CN105512938A CN105512938A CN201610076202.3A CN201610076202A CN105512938A CN 105512938 A CN105512938 A CN 105512938A CN 201610076202 A CN201610076202 A CN 201610076202A CN 105512938 A CN105512938 A CN 105512938A
- Authority
- CN
- China
- Prior art keywords
- user
- input
- behavior
- signal
- module
- 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
- G06Q40/03—Credit; Loans; Processing thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention provides an online credit risk assessment method based on the long-term using behavior of a user. The long-term using habit of the user on a mobile application APP is used as an anti-fraud analysis means, effective signals are extracted by acquiring a real-time geographic position, input time, an input method and input hint content which are input by the user when the user fills in a loan application form, and comprise an input method signal, a behavior signal and a phrase selecting signal, then the signals are converted into characteristics, and therefore a decision-making tree or other risk models are utilized to perform classification judgment. On the basis of the original application process, many characteristics can be provided for an anti-fraud algorithm without improving the application complexity degree; the natural using habit of the user is extracted and cannot be faked easily.
Description
Technical field
The present invention relates to a kind of credit estimation method of applied mathematical model, especially relate to a kind of online assessing credit risks method based on the behavior of user's Long-Time Service.
Background technology
Along with the violent fast development of mobile Internet, people more and more get used to utilizing mobile phone to carry out doing shopping, consume, amusement.Increasing lend-borrow action has also started to shift on line.P2P specially refers to that individual borrows or lends money with the small amount in a human world conclude the business, and the E-commerce Specialty network platform helps debtor and creditor establish debtor-creditor relationship and complete relationship trading formality.In order to ensure the interests of lender, platform side needs to carry out comprehensive assessment to the credit qualification of debtor, because domestic credit system development relatively lags behind, causes there is a large amount of swindle phenomenons in the middle of each P2P platform current.
Current identification swindle method has two defects:
1. identify the main personal information relying on user to provide of swindle, utilize these information extractions to go out and instead swindle signal accordingly.These methods depend on the every terms of information that user fills in.Want to reduce probability of cheating, will require that user fills in more user profile, and make application process more and more loaded down with trivial details.The experience of loaded down with trivial details application process meeting negative effect user apply for loan, can reduce the achievement of company.
2. what prior art mainly relied on is the unsolicited data of user, and these data fabrication costs are low, can forge easily.
When lending agency and loan user are interviewed, body language and application behavior are important anti-fraud information sources.In application process on line, the application behavior of user can be captured and digitizing, for credit approval decisionmaking in the software of user's use.
User has its input habit formed for a long time, and the change of custom is difficult, and habitual movement is simultaneously that user makes without thinking, and such signal has reacted the information of user self really.Ordinary people needs thinking, needs the correctness constantly verifying input content simultaneously.
Summary of the invention
The invention provides a kind of online assessing credit risks method based on the behavior of user's Long-Time Service, the method gathers the real behavior of user when user does not have perception, utilize these behavior construction features then to go to find potential fraudulent user by decision engine.The data that the present invention gathers are collections when user carries out loan application, do not invade the privacy information of user.Its technical scheme is as described below:
Based on an online assessing credit risks method for user's Long-Time Service behavior, comprise the following steps:
(1) user starts Mobile solution App at mobile platform, the SDK that Mobile solution App inserts is called when filling in list, described SDK gathers the input information of user when filling in loan application list, real-time geographical locations when comprising input residing for user, input time, input method, input prompt content, described SDK is provided with information acquisition module, for input information is sent to server end;
(2) server end carries out analyzing for the content that user inputs and utilizes characteristic extracting module from input information, to extract effective signal respectively, comprises geographic location signal, input method signal, behavior signal, phrase selection signal;
(3) effective signal is converted into feature by described characteristic extracting module, be sent to off-line model training module, on-line decision engine, on-line study engine respectively, feature after process is sent to the risk model in on-line decision engine by described off-line model training module and on-line study engine, and then utilizes risk model to carry out classification judgement.
In step (2), described behavior signal comprises the letter of typewriting, the key position using touch screen dummy keyboard, time, rhythm, act of revision; Described input method signal comprise whole tone, Two bors d's oeuveres, five, nine grids, stroke, hand-written method choice; Described phrase selects signal to comprise frequency of utilization and the phrase option priority of special phrase, thus scores respectively to sensitive word and keyword.
Further, in step (1), described input method signal directly can obtain the input method type of mobile phone, or it is corresponding to carry out identification by the key position information of input, the key position information of described input comprises letter, the act of revision of the typewriting in behavior signal, the input method signal identified is given a mark, judge to use the crowd characteristic of different input method by statistical method, and this feature is added off-line model training module, on-line decision engine, on-line study engine carry out classification judgement.
User can record in the time of each form of input, information input time extracted can convert the tempo signal of behavior signal to, time when at every turn knocking keyboard goes on record, time is millisecond rank, and then form a time series, utilize this seasonal effect in time series feature, judge that sequence averaging time of user calculates similarity.
Obtain phrase according to input method suggestion content and select signal, described phrase selects signal to carry out sensitive word score for the input text of user, and a responsive dictionary is built on backstage, the input prompt that user is each, the position cumulative calculation comprising cue and this word hits the position of the word of responsive dictionary, obtains susceptibility score.
Obtain phrase according to input method suggestion content and select signal, described phrase selects signal to carry out keyword score for the input text of user, comprises singularity and the actual selection priority difference of inputted phrase, obtains the score of crucial degree.
In step (3), described off-line model training module is that financial institution utilizes the lending historical record of oneself accumulation to train model, described lending historical record comprises the refund behavior of user, the content of user's input and input behavior record, the algorithm that described off-line model training module can adopt comprises logistic regression, support vector machine, decision tree, using the algorithm selected as risk model, expression behaviour when utilizing user's apply for loan of collecting to fill in list matches the refund performance of this user after obtaining loan again, constructs training dataset.
In step (3), the risk model that on-line decision engine uses off-line model training module to build based on historical data, and run up to present sensitive word dictionary and the next score each user of line computation of keyword dictionary always.
Further, in step (3), the algorithm that described risk model adopts comprises logistic regression, support vector machine, decision tree.
Described information acquisition module comprises the input method module in mobile platform, the characteristic extracting module be connected with input method module, the information issuing module be connected with characteristic extracting module, and accept module in the information of server end, described information accepts module and is connected with characteristic extracting module.
The present invention directly gathers the behavior of user when carrying out loan application, has following advantage:
1, on original application process basis, do not improve application fussy degree, just can provide multinomial feature for anti-algorithm of swindling; 2, feature extraction is in the natural use habit of user, is not easy to forge.
Accompanying drawing explanation
Fig. 1 is that the data of the described online assessing credit risks method based on the behavior of user's Long-Time Service transmit schematic diagram.
Embodiment
The invention provides a kind of online assessing credit risks method based on the behavior of user's Long-Time Service, the real behavior of user can be gathered when user does not have perception, utilize these behavior construction features then to go to find potential fraudulent user by decision engine.The data that the present invention gathers are collections when user carries out loan application, do not invade the privacy information of user.
Real-time geographical locations when the present invention is by gathering the input of user when filling in loan application list residing for user, input time, input method, input prompt content, extract effective signal respectively, comprise geographic location signal, behavior signal, input method signal, phrase selection signal, and the identification of credit applications is one of application of these four layers of features.Practical application needs a whole set of data science analytical framework, as shown in Figure 1, comprise: the Mobile solution APP102 installed in mobile platform 101, described Mobile solution APP102 calls the input method module 103 be connected with phrase database 104, input method module passes the signal along to characteristic extracting module 112, again by features convey to information issuing module 105, further, beyond the clouds in server 201, the information receiving module 106 be connected with information issuing module 105, also comprise characteristic extracting module 107, off-line model training module 108, on-line decision engine 109, on-line decision model 110, on-line study engine 111, effective signal is converted into feature by described characteristic extracting module 107, be sent to off-line model training module 108 respectively, on-line decision engine 109, on-line study engine 111, feature after process is sent to the on-line decision model 110 of on-line decision engine 109 by described off-line model training module 108 and on-line study engine 111, and then utilize decision tree risk model to carry out classification judgement.
Described characteristic extracting module 112 for carrying out the extraction of feature in Mobile solution to signal, described characteristic extracting module 107 is for carrying out feature extraction at server end, and described characteristic extracting module 107 can be further processed characteristic extracting module 112.Further, described characteristic extracting module 107 and characteristic extracting module 112 also only can exist one, if only existing characteristics extraction module 112 time, then the feature of extraction that what information issuing module 105 was transmitted is exactly.
When concrete operation, loan application people refers in particular to the user using mobile phone A pp, and these users use certain P2PApp, financing App, bank house keeper App, the fiduciary loan of being mortgaged to information service platform or financial institution's application nothing as medium by mobile phone A pp.
The financial credit product that described loan application list refers to all needs user initiatively to fill in some essential informations, includes but not limited to the name of user, identification card number, telephone number, work unit, monthly salary, use of the loan, work address, home address, lineal relative's name, lineal relative's telephone number etc.
Call input method in mobile platform and fill in list, any user can call input method input when filling in list.
For information acquisition module, user can call the SDK (full name: SoftwareDevelopmentKit that App inserts when filling in list, corresponding Chinese is SDK (Software Development Kit)), information acquisition module is just built in this SDK, the usage behavior signal that this SDK produces when carrying out list and filling in mainly for user on various entering method keyboard, the position (position) that such as user knocks when inputting oneself name, time (time) and content (text), and the information that input produces.
A such as virtual user being " Zheng Xiujing ", she directly may input " zxj " these three letters when inputting oneself name, information after this method also can input user " zxj " three letters gathers out, generates prompting queue (hintqueue).
Further, the content (test) that described characteristic extracting module inputs for user carries out analyzing and extracting feature, such as user's input " zhengxiujing ", also likely directly input " zxj " these three letters, input the name spelling of oneself by user or the difference of initials behavior extracts correlated characteristic.
Input " zxj " these three letters, utilize spelling input method, the phrase prompting order obtained is " the lower left corner, oneself, be Myself, scholarship, camera, philosopher, Zheng Xiujing, in Xinjiang, in foot bath, Miss Zhao, Miss Zhang, center street, meet each other again, Miss Zhou, meet together again, look for Miss " the prompting queue hint_queue of 16 cues altogether, and the actual text of last phrase " Zheng Xiujing " (word_pos=7) and user's input matches, then keyword score is keyword_score=1-word_pos/size (hint_queue).
Prompting text when user is inputted, also sensitive word score can be carried out, backstage builds a responsive dictionary and [lottery ticket, competing coloured silk, game, gambling, drugs, Miss, to raise ... ], the input prompt hintqueue that user is each, comprise the position of cue and this word, the last cue in following table " is looked for Miss " and is belonged to sensitive word.
| Cue | Cue sorts |
| The lower left corner | 1 |
| Oneself | 2 |
| Be Myself | 3 |
| Scholarship | 4 |
| Camera | 5 |
| Philosopher | 6 |
| Zheng Xiujing | 7 |
| In Xinjiang | 8 |
| In foot bath | 9 |
| Miss Zhao | 10 |
| Miss Zhang | 11 |
| Center street | 12 |
| Meet each other again | 13 |
| Miss Zhou | 14 |
| Meet together again | 15 |
| Look for Miss | 16 |
Cumulative calculation hits position Sum (the 1/word_pos)=sensitive_score of the word of responsive dictionary, obtains susceptibility score.
Identification is carried out for key position (position) information by input corresponding, will corresponding input method type be identified, such as phonetic, five, these three kinds of input modes hand-written.Such as, input " zhengxiujing " or " udbftebjjjf ", the former is spelling input method, and the latter is five-stroke input method.The probability using different input method is judged by statistical method, and using this probable value as feature.Again according to probable value height, the input method type (input_type) identified is given a mark, such as phonetic=2, five=1, hand-written=0.
Rhythm (rhythm) signal can be converted to for the time (time) of extracting, for identification card number, assuming that I.D. be input as 990440199009093344, time when at every turn knocking keyboard has been recorded, this time be Millisecond other.Such as:
A time series can be formed like this:
T=[500,500,700,500,500,700,500,500,500,700,500,500,500,700,500,500,500], after obtaining such time series, utilize sequence t_avg averaging time of this time series and user to calculate similarity, utilize similarity cos (t, t_avg) that the Similarity value cos_val of user can be calculated.
The real-time geographic location signal (location) of user can be collected, such as capture the location=Hongkong of user, list is filled in middle meeting and is required that user fills in work address simultaneously, household resident address, work_address=Beijing, home_address=Beijing, can convert consistance signal to.If work_address unequal to is location, so work_address_match=0, if home_address=location, so home_address_match=1.
User can record in the time of each form of input, and the content of form comprises name, identification card number, telephone number, work unit, monthly salary, use of the loan, lineal relative's name, lineal relative's telephone number, as follows:
tinput_name,tinput_id,tinput_phone,tinput_company,tinput_salary,tinput_purpose,tinput_address,tinput_family,tinput_family_phone。
Off-line training module, it is the lending historical record that financial institution utilizes oneself accumulation, train model, described lending historical record comprises the refund behavior of user and the input content of user and input behavior, this case uses decision tree as risk model, expression behaviour when utilizing user's apply for loan of collecting to fill in list matches the refund performance of this user after obtaining loan again, builds training dataset.For the definition of the good user of classification samples and fraudulent user, this case uses user to obtain after loan the refund in N number of month future the term of execution, if user has at most only repaid the N/3 phase and has provided a loan and just think that they are fraudulent user, if user repays the loan, the longest number of days that once exceeds the time limit is less than N*6 days and has so just been defined as user, the user of other scope is thought of as grey user, it goes without doing training data, wherein N>1.Off-line training model adopts logistic regression to train and carries out tune ginseng to model.
The risk model that on-line decision engine mainly uses off-line training module to build based on the history user data of 6 months, and run up to present sensitive word dictionary and the next score each user of line computation of keyword dictionary always.And the user data of 6 months also can select the quantitative value in other months as required.
Signal, after changing feature into, will enter model and calculate.For virtual " Zheng Xiujing " user, the various features of this user generates a proper vector:
Vec=[input_type, cos_val, work_address_match, home_address_match, tinput_name, tinput_id, tinput_phone, tinput_company, tinput_salary, tinput_purpose, tinput_address, tinput_family, tinput_family_phone, sensitive_score]=[0, 0.3, 0, 0, 14.6, 18.2, 14.4, 83.4, 2.6, 317, 14.6, 14.9, 15.1, 0.32], call LogisticRegression (later referred to as LR, corresponding Chinese is logistic regression) model, user_fraud_score=LR (vec)=0.9, the value that LR calculates is interval [0, 1.0], historical data can calculate one and distinguish threshold value, this case is distinguish threshold value with threshold (threshold value)=0.85, because fraudulent user number is statistically a small amount of, threshold value usually can be higher, because the swindle mark of " Zheng Xiujing " user is 0.9>threshold, therefore refusal is carried out batch loan to this user by decision engine.
For LR model wherein at off-line model training module, a large amount of vec (proper vector) to be input to and to obtain in model, input packet is wherein containing vec and corresponding mark.Such as a complete training data should be [vec; label]; wherein label is the mark of expression behaviour after user provides a loan; user normally refunds label=0; user does not normally refund label=1; the training data utilizing multirow such builds the matrix M of the capable m row of n, and wherein line number is exactly training sample data, and columns is exactly mark corresponding to the characteristic sum of sample.After unknown subscriber's feature enters model LR, we just can obtain the prediction as the preceding paragraph.
Predicting the outcome is after judging through decision engine, and user can be swindled or the result of non-swindle, once user is judged as fraudulent user, the loan application of user will be rejected.
The present invention utilizes the Long-Time Service custom of user on mobile phone A pp to make the means of anti-swindle identification, by gathering the input position of user when filling in loan application list, input time, input method, input method suggestion content, extract effective signal, then these signals are converted into feature, and then utilize decision tree risk model to carry out classification judgement.
There is not other replacement scheme in this case, the risk model of this case can adopt other algorithm, such as logistic regression, support vector machine, neuroid, random forest and naive Bayesian etc. in data acquisition.
Different application crowds and application scenarios can select the support of different machines learning algorithm.At the product development initial stage, when Case comparison is rare, decision Tree algorithms can be selected.Swindle case have necessarily accumulate time, logistic regression algorithm can be selected.When Data Collection is more complete, algorithm of support vector machine and NB Algorithm can be selected.Imperfect at Data Collection, under having disappearance scene, can random forests algorithm be selected.When case is very sufficient, the available Integrated Algorithm incorporating neural network algorithm.
Claims (10)
1., based on an online assessing credit risks method for user's Long-Time Service behavior, comprise the following steps:
(1) user starts Mobile solution App at mobile platform, the SDK that Mobile solution App inserts is called when filling in list, described SDK gathers the input information of user when filling in loan application list, real-time geographical locations when comprising input residing for user, input time, input method, input prompt content, described SDK is provided with information acquisition module, for input information is sent to server end;
(2) server end carries out analyzing for the content that user inputs and utilizes characteristic extracting module from input information, to extract effective signal respectively, comprises geographic location signal, input method signal, behavior signal, phrase selection signal;
(3) effective signal is converted into feature by described characteristic extracting module, be sent to off-line model training module, on-line decision engine, on-line study engine respectively, feature after process is sent to the on-line decision model of on-line decision engine by described off-line model training module and on-line study engine, and then utilizes risk model to carry out classification judgement.
2. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, it is characterized in that: in step (2), described behavior signal comprises the letter of typewriting, the key position using touch screen dummy keyboard, time, rhythm, act of revision; Described input method signal comprise whole tone, Two bors d's oeuveres, five, nine grids, stroke, hand-written method choice; Described phrase selects signal to comprise frequency of utilization and the phrase option priority of special phrase, thus scores respectively to sensitive word and keyword.
3. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, it is characterized in that: in step (1), described input method signal directly can obtain the input method type of mobile phone, or it is corresponding to carry out identification by the key position information of input, the key position information of described input comprises the letter of the typewriting in behavior signal, act of revision, the input method signal identified is given a mark, the crowd characteristic using different input method is judged by statistical method, and this feature is added off-line model training module, on-line decision engine, on-line study engine carries out classification and judges.
4. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 2, it is characterized in that: user can record in the time of each form of input, information input time extracted can convert the tempo signal of behavior signal to, time when at every turn knocking keyboard goes on record, time is millisecond rank, and then formed a time series, utilize this seasonal effect in time series feature, judge user averaging time sequence thus calculating similarity.
5. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 2, it is characterized in that: obtain phrase according to input method suggestion content and select signal, described phrase selects signal to carry out sensitive word score for the input text of user, and a responsive dictionary is built on backstage, the input prompt that user is each, comprise the position of cue and this word, the position of the word of the responsive dictionary of accumulative hit, obtain susceptibility score.
6. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 2, it is characterized in that: obtain phrase according to input method suggestion content and select signal, described phrase selects signal to carry out keyword score for the input text of user, comprise singularity and the actual selection priority difference of inputted phrase, obtain the score of crucial degree.
7. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, it is characterized in that: in step (3), described off-line model training module is that financial institution utilizes the lending historical record of oneself accumulation to train model, described lending historical record comprises the refund behavior of user, the content of user's input and input behavior record, the algorithm that described off-line model training module can adopt comprises logistic regression, support vector machine, decision tree, using the algorithm selected as risk model, expression behaviour when utilizing user's apply for loan of collecting to fill in list matches the refund performance of this user after obtaining loan again, construct training dataset.
8. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, it is characterized in that: in step (3), the risk model that on-line decision engine uses off-line model training module to build based on historical data, and run up to present sensitive word dictionary and the next score each user of line computation of keyword dictionary always.
9. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, is characterized in that: in step (3), and the algorithm that described risk model adopts comprises logistic regression, support vector machine, decision tree.
10. the online assessing credit risks method based on the behavior of user's Long-Time Service according to claim 1, it is characterized in that: described information acquisition module comprises the input method module in mobile platform, the characteristic extracting module be connected with input method module, the information issuing module be connected with characteristic extracting module, and accepting module in the information of server end, described information accepts module and is connected with characteristic extracting module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610076202.3A CN105512938A (en) | 2016-02-03 | 2016-02-03 | Online credit risk assessment method based on long-term using behavior of user |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610076202.3A CN105512938A (en) | 2016-02-03 | 2016-02-03 | Online credit risk assessment method based on long-term using behavior of user |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105512938A true CN105512938A (en) | 2016-04-20 |
Family
ID=55720897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610076202.3A Pending CN105512938A (en) | 2016-02-03 | 2016-02-03 | Online credit risk assessment method based on long-term using behavior of user |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105512938A (en) |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106095915A (en) * | 2016-06-08 | 2016-11-09 | 百度在线网络技术（北京）有限公司 | The processing method and processing device of user identity |
| CN106296389A (en) * | 2016-07-28 | 2017-01-04 | 联动优势科技有限公司 | The appraisal procedure of a kind of user credit degree and device |
| CN106529773A (en) * | 2016-10-31 | 2017-03-22 | 宜人恒业科技发展（北京）有限公司 | Online credit and fraud risk evaluation method based on identifying code type question answering |
| CN107133867A (en) * | 2017-04-16 | 2017-09-05 | 信阳师范学院 | Credit method for anti-counterfeit based on SVMs |
| CN107330785A (en) * | 2017-07-10 | 2017-11-07 | 广州市触通软件科技股份有限公司 | A kind of petty load system and method based on the intelligent air control of big data |
| CN107346463A (en) * | 2016-05-04 | 2017-11-14 | 阿里巴巴集团控股有限公司 | Training, mode input data determination method and the device of risk control model |
| WO2018014789A1 (en) * | 2016-07-22 | 2018-01-25 | 阿里巴巴集团控股有限公司 | Method and device for controlling service operation risk |
| CN107741899A (en) * | 2017-10-16 | 2018-02-27 | 北京小米移动软件有限公司 | The method, apparatus and system of processing terminal data |
| CN108510280A (en) * | 2018-03-23 | 2018-09-07 | 上海氪信信息技术有限公司 | A kind of financial fraud behavior prediction method based on mobile device behavioral data |
| CN108564460A (en) * | 2018-01-12 | 2018-09-21 | 阳光财产保险股份有限公司 | Real-time fraud detection method under internet credit scene and device |
| CN108596434A (en) * | 2018-03-23 | 2018-09-28 | 卫盈联信息技术（深圳）有限公司 | Fraud detection and methods of risk assessment, system, equipment and storage medium |
| CN109426894A (en) * | 2017-08-22 | 2019-03-05 | 阿里巴巴集团控股有限公司 | User information is shared, price competing method, device, system and electronic equipment |
| CN109428900A (en) * | 2017-08-21 | 2019-03-05 | 阿里巴巴集团控股有限公司 | A kind of method and device of data processing |
| CN109636233A (en) * | 2018-12-20 | 2019-04-16 | 四川新网银行股份有限公司 | The method and system of comparison threshold value identification behavioural characteristic confidence level based on building |
| CN109670929A (en) * | 2018-09-13 | 2019-04-23 | 深圳壹账通智能科技有限公司 | Control method, device, equipment and the computer readable storage medium of loan early warning |
| CN110109740A (en) * | 2019-05-08 | 2019-08-09 | 宜人恒业科技发展（北京）有限公司 | A kind of execution method and device of Crontab order |
| CN110363649A (en) * | 2019-06-27 | 2019-10-22 | 上海淇馥信息技术有限公司 | A kind of method for prewarning risk based on user operation case, device, electronic equipment |
| CN110827033A (en) * | 2019-10-11 | 2020-02-21 | 支付宝(杭州)信息技术有限公司 | Information processing method and device and electronic equipment |
| WO2020042290A1 (en) * | 2018-08-28 | 2020-03-05 | 卫盈联信息技术（深圳）有限公司 | Risk management method, and apparatus and computer-readable storage medium |
| CN111046374A (en) * | 2019-11-26 | 2020-04-21 | 山东浪潮人工智能研究院有限公司 | Method and system for improving user password security based on machine learning |
| CN111080873A (en) * | 2019-12-31 | 2020-04-28 | 中国银行股份有限公司 | Key characteristic-based bank vault control method and device |
| CN111161042A (en) * | 2019-11-26 | 2020-05-15 | 深圳壹账通智能科技有限公司 | Personal risk assessment method, device, terminal and storage medium |
| CN111784348A (en) * | 2016-04-26 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Account risk identification method and device |
| CN112508696A (en) * | 2021-02-05 | 2021-03-16 | 北京淇瑀信息科技有限公司 | Channel user quality evaluation method and device and electronic equipment |
| CN113177844A (en) * | 2021-05-28 | 2021-07-27 | 成都新希望金融信息有限公司 | Credit risk prediction method, credit risk prediction device, credit risk prediction equipment and credit risk prediction storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040044615A1 (en) * | 2002-09-03 | 2004-03-04 | Xue Xun Sean | Multiple severity and urgency risk events credit scoring system |
| CN103413223A (en) * | 2013-07-24 | 2013-11-27 | 通联支付网络服务股份有限公司 | Personal-reputation and financing-credit evaluation system for non-face-to-face trading |
| CN104616198A (en) * | 2015-02-12 | 2015-05-13 | 哈尔滨工业大学 | P2P (peer-to-peer) network lending risk prediction system based on text analysis |
| CN105049421A (en) * | 2015-06-24 | 2015-11-11 | 百度在线网络技术（北京）有限公司 | Authentication method based on use behavior characteristic of user, server, terminal, and system |
-
2016
- 2016-02-03 CN CN201610076202.3A patent/CN105512938A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040044615A1 (en) * | 2002-09-03 | 2004-03-04 | Xue Xun Sean | Multiple severity and urgency risk events credit scoring system |
| CN103413223A (en) * | 2013-07-24 | 2013-11-27 | 通联支付网络服务股份有限公司 | Personal-reputation and financing-credit evaluation system for non-face-to-face trading |
| CN104616198A (en) * | 2015-02-12 | 2015-05-13 | 哈尔滨工业大学 | P2P (peer-to-peer) network lending risk prediction system based on text analysis |
| CN105049421A (en) * | 2015-06-24 | 2015-11-11 | 百度在线网络技术（北京）有限公司 | Authentication method based on use behavior characteristic of user, server, terminal, and system |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111784348A (en) * | 2016-04-26 | 2020-10-16 | 阿里巴巴集团控股有限公司 | Account risk identification method and device |
| CN107346463A (en) * | 2016-05-04 | 2017-11-14 | 阿里巴巴集团控股有限公司 | Training, mode input data determination method and the device of risk control model |
| CN107346463B (en) * | 2016-05-04 | 2020-08-28 | 阿里巴巴集团控股有限公司 | Method and device for training risk control model and determining model input data |
| CN106095915A (en) * | 2016-06-08 | 2016-11-09 | 百度在线网络技术（北京）有限公司 | The processing method and processing device of user identity |
| WO2018014789A1 (en) * | 2016-07-22 | 2018-01-25 | 阿里巴巴集团控股有限公司 | Method and device for controlling service operation risk |
| TWI699720B (en) * | 2016-07-22 | 2020-07-21 | 香港商阿里巴巴集團服務有限公司 | Risk control method and device for business operation |
| CN106296389A (en) * | 2016-07-28 | 2017-01-04 | 联动优势科技有限公司 | The appraisal procedure of a kind of user credit degree and device |
| CN106529773A (en) * | 2016-10-31 | 2017-03-22 | 宜人恒业科技发展（北京）有限公司 | Online credit and fraud risk evaluation method based on identifying code type question answering |
| CN107133867A (en) * | 2017-04-16 | 2017-09-05 | 信阳师范学院 | Credit method for anti-counterfeit based on SVMs |
| CN107330785A (en) * | 2017-07-10 | 2017-11-07 | 广州市触通软件科技股份有限公司 | A kind of petty load system and method based on the intelligent air control of big data |
| CN109428900B (en) * | 2017-08-21 | 2022-05-13 | 创新先进技术有限公司 | Data processing method and device |
| CN109428900A (en) * | 2017-08-21 | 2019-03-05 | 阿里巴巴集团控股有限公司 | A kind of method and device of data processing |
| CN109426894A (en) * | 2017-08-22 | 2019-03-05 | 阿里巴巴集团控股有限公司 | User information is shared, price competing method, device, system and electronic equipment |
| CN107741899B (en) * | 2017-10-16 | 2021-07-27 | 北京小米移动软件有限公司 | Method, device and system for processing terminal data |
| CN107741899A (en) * | 2017-10-16 | 2018-02-27 | 北京小米移动软件有限公司 | The method, apparatus and system of processing terminal data |
| CN108564460B (en) * | 2018-01-12 | 2020-10-30 | 阳光财产保险股份有限公司 | Real-time fraud detection method and device in internet credit scene |
| WO2019137050A1 (en) * | 2018-01-12 | 2019-07-18 | 阳光财产保险股份有限公司 | Real-time fraud detection method and device under internet credit scene, and server |
| CN108564460A (en) * | 2018-01-12 | 2018-09-21 | 阳光财产保险股份有限公司 | Real-time fraud detection method under internet credit scene and device |
| CN108510280B (en) * | 2018-03-23 | 2020-07-31 | 上海氪信信息技术有限公司 | Financial fraud behavior prediction method based on mobile equipment behavior data |
| CN108510280A (en) * | 2018-03-23 | 2018-09-07 | 上海氪信信息技术有限公司 | A kind of financial fraud behavior prediction method based on mobile device behavioral data |
| CN108596434A (en) * | 2018-03-23 | 2018-09-28 | 卫盈联信息技术（深圳）有限公司 | Fraud detection and methods of risk assessment, system, equipment and storage medium |
| WO2020042290A1 (en) * | 2018-08-28 | 2020-03-05 | 卫盈联信息技术（深圳）有限公司 | Risk management method, and apparatus and computer-readable storage medium |
| CN109670929A (en) * | 2018-09-13 | 2019-04-23 | 深圳壹账通智能科技有限公司 | Control method, device, equipment and the computer readable storage medium of loan early warning |
| CN109636233A (en) * | 2018-12-20 | 2019-04-16 | 四川新网银行股份有限公司 | The method and system of comparison threshold value identification behavioural characteristic confidence level based on building |
| CN110109740A (en) * | 2019-05-08 | 2019-08-09 | 宜人恒业科技发展（北京）有限公司 | A kind of execution method and device of Crontab order |
| CN110363649A (en) * | 2019-06-27 | 2019-10-22 | 上海淇馥信息技术有限公司 | A kind of method for prewarning risk based on user operation case, device, electronic equipment |
| CN110827033A (en) * | 2019-10-11 | 2020-02-21 | 支付宝(杭州)信息技术有限公司 | Information processing method and device and electronic equipment |
| CN111161042A (en) * | 2019-11-26 | 2020-05-15 | 深圳壹账通智能科技有限公司 | Personal risk assessment method, device, terminal and storage medium |
| CN111046374A (en) * | 2019-11-26 | 2020-04-21 | 山东浪潮人工智能研究院有限公司 | Method and system for improving user password security based on machine learning |
| CN111080873A (en) * | 2019-12-31 | 2020-04-28 | 中国银行股份有限公司 | Key characteristic-based bank vault control method and device |
| CN112508696A (en) * | 2021-02-05 | 2021-03-16 | 北京淇瑀信息科技有限公司 | Channel user quality evaluation method and device and electronic equipment |
| CN113177844A (en) * | 2021-05-28 | 2021-07-27 | 成都新希望金融信息有限公司 | Credit risk prediction method, credit risk prediction device, credit risk prediction equipment and credit risk prediction storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105512938A (en) | Online credit risk assessment method based on long-term using behavior of user | |
| TWI712981B (en) | Risk identification model training method, device and server | |
| Bansal et al. | On predicting elections with hybrid topic based sentiment analysis of tweets | |
| CN110956547B (en) | Method and system for identifying fraudulent party in real time based on search engine | |
| CN108475393A (en) | The system and method that decision tree is predicted are promoted by composite character and gradient | |
| CN105787025B (en) | Network platform public account classification method and device | |
| CN108921686A (en) | A kind of credit-graded approach and device of personal user | |
| CN106529773A (en) | Online credit and fraud risk evaluation method based on identifying code type question answering | |
| CN105095187A (en) | Search intention identification method and device | |
| CN108961032A (en) | Borrow or lend money processing method, device and server | |
| CN102567534B (en) | Interactive product user generated content intercepting system and intercepting method for the same | |
| CN110659961A (en) | Method and device for identifying off-line commercial tenant | |
| CN107135314A (en) | Harass detection method, system, mobile terminal and the server of short message | |
| CN113011884B (en) | Account feature extraction method, device, equipment and readable storage medium | |
| CN110399479A (en) | Search for data processing method, device, electronic equipment and computer-readable medium | |
| CN112163944A (en) | Loan qualification scoring method and device for customer, computer equipment and storage medium | |
| CN115577172A (en) | Article recommendation method, device, equipment and medium | |
| CN116821759A (en) | Identification prediction method and device for category labels, processor and electronic equipment | |
| CN110389963A (en) | The recognition methods of channel effect, device, equipment and storage medium based on big data | |
| CN115687754A (en) | Active network information mining method based on intelligent conversation | |
| CN110570301B (en) | Risk identification method, device, equipment and medium | |
| CN112115258B (en) | Credit evaluation method and device for user, server and storage medium | |
| CN113420789A (en) | Method, device, storage medium and computer equipment for predicting risk account | |
| CN113327161A (en) | Intelligent wind control decision system for credit business | |
| CN113254775A (en) | Credit card product recommendation method based on client browsing behavior sequence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20160420 |