CN103793484B - The fraud identifying system based on machine learning in classification information website - Google Patents
The fraud identifying system based on machine learning in classification information website Download PDFInfo
- Publication number
- CN103793484B CN103793484B CN201410022138.1A CN201410022138A CN103793484B CN 103793484 B CN103793484 B CN 103793484B CN 201410022138 A CN201410022138 A CN 201410022138A CN 103793484 B CN103793484 B CN 103793484B
- Authority
- CN
- China
- Prior art keywords
- data
- user
- user behavior
- behavior
- model
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/958—Organisation or management of web site content, e.g. publishing, maintaining pages or automatic linking
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides in a kind of website for classification information based on machine learning fraud identifying system method, methods described comprises the steps：A) existing user behavior data sample drawn data are based on, for generation model first；B) for the multiple user behavior features of training data selective extraction of different service types；C) based on the user behavior feature that is extracted, vectorization is carried out to the sample training data；D) forecast model is produced using the sample training data of vectorization；E) data on line are detected based on classification and clustering rule using produced model；F) the detected abnormal user data for obtaining are processed.The behavior of user can be identified from multiple dimensions using the present invention, efficiently reduce the false amount of Transaction Information.Even if also, also can be good at being identified user behavior inferior in the case of containing noise data in training data.
Description
Technical field
The present invention relates to Internet technology, the fraud based on machine learning in particularly a kind of classification information website
Identifying system.
Background technology
Classification information net is a kind of website class of every aspect information for being related to daily life that the Internet newly rises
Type.Inside these websites user can obtain freely, easily information distribution service, including second-hand article trading, used car
Dealing, housing, house pet, recruitment, part-time, job hunting, make friend activity, life service information etc..Classification information is wide also known as classification
Accuse, the daily advertisement that is seen on TV, newpapers and periodicals of people, often no matter beholder is willing to be unwilling that it can all impose on sight
The person of seeing, this series advertisements are passive advertisement；And people actively go to inquire about the information of the aspect such as recruit, rent a house, travelling, these are believed
Breath, it is called active advertisement.In today that information-intensive society progressively develops, passive advertisement increasingly causes the dislike of people, and leads
Dynamic advertisement is but subject to the extensive favor of people.Almost each local evening paper, daily paper, life ＆ amusement report can't do without classification information
Figure, and do to obtain better newspaper, the length of classification information is often bigger.Thus classification information net is just generated.
In the user of classification information website orientation information, a part of poor quality user often occurs, they are issuing falseness
The mode such as information fraudulent user is obtaining interests.Therefore, classification information website can arrange some process rules to tinpot information
With filter logic etc..
The means of existing deceptive information identification are mainly based upon the recognition method of rule, some artificial interventions additional,
For example counted with issuing within a period of time by counting an ip, in information content whether containing illegal word, the commodity that issues
Or whether unreasonable etc. the rule of price range of service is come whether judge a user be the user inferior that issues deceptive information,
So as to take the processing means such as deletion information, warning, logging off users.However, common process rule and filter logic are generally all
It is to carry out the identification of behavior inferior using single dimension, therefore user inferior can sound out the critical of rule by every means
Point is so as to process and the filter logic to tinpot information around open system.
In addition, reaching the standard grade with various rules, spendable rule can be fewer and feweri, because rule is all can be obvious
Feature.The identification of rule can only be made a distinction using linear classifying face in existing method, so as to cause the bad of majority
Matter information is all without being recognized by the system and process.
Accordingly, it would be desirable to the fraud identifying system based on machine learning in a kind of classification information website, comes multiple
Dimension is identified to the behavior of user, so as to efficiently reduce the false amount of Transaction Information, improves the verity of Transaction Information.
Content of the invention
It is an object of the invention to provide the fraud identifying system based on machine learning in a kind of classification information website.
According to an aspect of the invention, there is provided the fraud row in a kind of website for classification information based on machine learning
For the method for identifying system, methods described comprises the steps：A) based on existing user behavior data sample drawn data, use
In generation model first；B) for the multiple user behavior features of training data selective extraction of different service types；C) based on institute
The sample training data are carried out vectorization by the user behavior feature of extraction；D) produced using the sample training data of vectorization
Raw forecast model；E) data on line are detected based on classification and clustering rule using produced model；F) to being detected
The abnormal user data for obtaining are processed.
Preferably, the sample data in step a includes positive sample data and negative sample data, corresponds respectively to high-quality
The user of behavior and the user of behavior inferior.
Preferably, in step b, user behavior feature includes user behavior data and the use for same cookie
The statistical magnitude of each dimension in family.
Preferably, selected to difference by way of calculating comentropy and model intersection data verification in step b
The user characteristicses that type of service is extracted.
Preferably, used in step d, the grader of probabilistic type carries out decision-making.
Preferably, the Probability Point of the abnormal probability for representing user behavior data is calculated using model in step e.
Preferably, the method for calculating the Probability Point is that multiple models enter to many stack features of user behavior data respectively
Row detection, and draw a point of Probability Point respectively, sum of products conversion operation is carried out to each point of Probability Point then, user behavior is drawn
The Probability Point of data.
Preferably, include setting a probability based on user's anomaly detection method of classifying ruless in step e
Line is used for judging whether user behavior data is bad data.
Preferably, included based on user's anomaly detection method of clustering rule in step e as follows：E1) to general
Rate point carries out clustering phenomenon monitoring；E2) Probability Point cluster is detected to a number of user behavior, to judge cluster extremely
Whether the user behavior of equal probabilities point is user behavior inferior；E3) according to testing result, abnormal user behavior discrimination model pair
The Probability Point of such user behavior is updated；E4 the new bad data for passing through detection discovery is added sample as training data)
This storehouse；E5) new training data training pattern is utilized.
Preferably, for the inaccurate user behavior data of Probability Point carries out off-line analysiss in step e5, find new
User behavior feature and select suitable feature.
Using the fraud identifying system based on machine learning in a kind of classification information website of the present invention, Neng Goucong
Multiple dimensions are identified to the behavior of user, so as to efficiently reduce the false amount of Transaction Information, improve the true of Transaction Information
Reality.Even if also, also can be good at carrying out user behavior inferior in the case of containing noise data in training data
Identification.
Description of the drawings
With reference to the accompanying drawing that encloses, the present invention more purpose, function and advantages are by by the as follows of embodiment of the present invention
Description is illustrated, wherein：
Fig. 1 diagrammatically illustrates the fraud identifying system based on machine learning in the classification information website of the present invention
Method flow diagram.
Specific embodiment
By reference to one exemplary embodiment, the purpose of the present invention and function and the side for realizing these purposes and function
Method will be illustrated.However, the present invention is not limited to one exemplary embodiment disclosed below；Can by multi-form come
Which is realized.The essence of description is only to aid in the detail of the various equivalent modifications Integrated Understanding present invention.
Hereinafter, embodiments of the invention will be described with reference to the drawings.In the accompanying drawings, identical reference represents identical
Or similar part, or same or like step.
The fraud information recognition methodss of the present invention have used the data produced based on user behavior, can immediately to user
The information data of issue is identified.The Model Identification of the machine learning that the present invention is adopted, can be in multiple dimensions to user's
Behavior is identified so that the user of the particularly tinpot information that releases news is difficult to know that what the dimension of identification is, so as to nothing
Method is evaded by getting around rule.The present invention can be predicted to data in the environment of a small amount of sample and high noisy,
And accuracy rate is high.Modeling is collected in the various actions to user, abnormal user is identified so as to reach.
Fig. 1 diagrammatically illustrates the fraud identifying system based on machine learning in the classification information website of the present invention
Method flow diagram.As shown in Figure 1：
Step 110, based on existing user behavior data sample drawn data, for generation model first.The extraction
Sample data can be extracted from the existing user behavior data storehouse that had audited, be mainly used in for user dividing into high-quality use
Family and user inferior, correspond respectively to positive sample data and negative sample data, and wherein, positive sample is through auditing the high-quality row for passing through
For the user behavior data of user, and negative sample is the behavioral data of the behavior user inferior through examination ＆ verification identification, for example offence
Some more serious rules（Wash sale information is for example issued）User behavior data.Sample number in existing examination ＆ verification storehouse
According to the data base set up for being classified to user behavior by some conventional user behavior recognition methods.Methods described is for example：
Whether whether contain in the text message that detection user issues in illegal word, the pictorial information that detection user issues containing illegal interior
Hold etc..
During the upper once model iteration after generation model, the method according to the invention can directly using institute
The positive and negative samples storehouse of extraction, and original examination ＆ verification storehouse information need not be reused.
Step 120, for the multiple user behavior features of training data selective extraction of different service types.Imitated by experiment
Really, which user behavior feature used in different business line judged.
The characteristic behavior of user is generally very more, in view of the requirement of computational accuracy and computational efficiency simultaneously, according to the present invention
User behavior feature generalizations be the feature for having discrimination to good and bad user, be therefore not required for a lot of features, model is got over
Little better, it is therefore an objective to data can be detected using multiple models in last identification.
User behavior is characterized in that the behavior characteristicss found by some off-line analysiss, typically will not be with other rule weights on line
Multiple, it can be understood as the feature that background audit personnel can not find, e.g. for the user behavior data of same cookie, bag
Include：Trans-city post number, using mobile phone number, time interval and click behavior, user's registration to the time interval that posts
Point, also have the statistical magnitude of the login behavior of user, the data of dimension such as user browsing behavior and each dimension of user such as：ip
Count within trans-city N days, cookie is counted etc..
Preferably for the training data of different types of service, system extracts different features.Such as used car business
Line using feature only can include：The time of posting is registered to, the user that posts fills in data time, the mouse track of user, 30
Its ip correlated count etc..Second-hand service line can only include following feature：User's registration login time, N days before the corresponding ip of user
Browsing pages number of times.
It is highly preferred that being selected to different business line drawing by way of calculating comentropy and model intersection data verification
Feature.
The model set up by the method according to the invention only have selected each that the feature of less dimension, i.e. model are used
Dimension be all by data analysiss produce to the good dimension of classifying quality, the data that therefore each dimension is produced are not
Sparse.The method of the present invention overcomes dimension in prior art excessively to be caused to calculate excessively complicated shortcoming, and prior art is led to
Be often by text is carried out participle distinguish user so as to produce a large amount of dimensions, each word as a dimension, so as to cause to instruct
Practice sample excessive, calculate complexity, also can be too high to the accuracy requirement of sample data.
Sample training data, according to the user behavior feature that selects in step 120, are carried out vectorization by step 130.Training
The result of the vectorization of data can be saved in file.The component of each dimension of training data is the feature of a selection.
Below by taking the vectorization procedure of this dimension of time that user fills in model as an example, the vectorization procedure of training data is described：
1. the data of each model fill in the time are obtained in sample data.
2. pair these data carry out data cleansing, its objective is to playback outlier.
3. the attribute of pair successive value carries out sliding-model control, using K mean cluster 100 times, is the cluster for having error minimum
Central point is used as discretization cut section.
4. and then last correction is carried out to data.
5. vectorization is completed.
Step 140, produces forecast model using the sample training data of vectorization.During training pattern, probability is preferably used
The grader of type carries out decision-making.Probabilistic type grader is used for the Probability Point for calculating user behavior data.
The reason for decision-making is carried out using the grader of probabilistic type is, as the purpose that last model is produced is to recognize and delete
Some Deviant Behavior information, so this model needs that there is very high accuracy rate, due to graders such as neutral net or decision trees
Situation about manslaughtering is had, so the grader using probabilistic type carries out decision-making.
The model for using is preferably and adopts Bayesian network model.Bayesian network (Bayesian network) is to be based on
The mathematical model of probability inference, which has a stronger generalization ability, and is capable of the layering and the output of probability of firm logic, institute
To be well suited for the scene of Activity recognition.
Preferably, the training of model, WEKA are carried out using the program WEKA of increasing income（Waikato Environment for
Knowledge Analysis）As disclosed data mining work platformses, gathered a large amount of can undertake data mining appoint
The machine learning algorithm of business, including data are carried out with pretreatment, classification is returned, cluster, correlation rule and in new interactive mode
Visualization on interface.
Data on line are detected based on classification and clustering rule by step 150 using produced multiple models, and right
User inferior is processed.When being detected to the user behavior data on line using abnormal user Activity recognition model, can be right
Every user behavior data is calculated, and generates a Probability Point, and Probability Point represents the probit that the data are bad data, this
Probability Point is more high, and the data are more likely to bad data.Wherein, the Probability Point of user behavior data is generated in the following manner：Each
Model many stack features respectively to user behavior data（The feature of multigroup different dimensions）Detected, each model is drawn respectively
One probit for representing that data are the probability that quality is inclined to（Hereinafter referred to as divide Probability Point）, finally each point of Probability Point is carried out
Sum of products conversion operation, draws the Probability Point of user behavior data.
The method of detection user's Deviant Behavior includes following two：
User's anomaly detection method based on classifying ruless.Set a probability line（That is probability threshold value）, for sentencing
Whether disconnected user behavior data is bad data.If the Probability Point of certain user behavior data exceedes probability threshold value, by user's row
For data judging be bad data, will the user be judged to user inferior.Otherwise then the user behavior data is judged to normally
The user is judged to high-quality user by data.Wherein, probability line is obtained by way of artificially verifying.
Data on line are detected based on clustering rule.Comprise the following steps that：
Step a, carries out clustering phenomenon monitoring to Probability Point.
Step b, Probability Point cluster is detected to a number of user behavior data turning-over operation personnel, cluster is judged extremely
Whether the user behavior of equal probabilities point is other user behaviors with the Probability Point in user behavior inferior, i.e. detection line
Whether data are all bad data.Cluster to the same class behavior of user behavior data or user's row of similar behavior of equal probabilities point
For data, they may only have the feature of less dimension distinct.Detection method is preferably, and detection has the use of the Probability Point
Whether family behavioral data is all fallen by other rule treatments.Wherein, other rules are the modes of the identification bad data outside invention.Example
Such as：The pictorial information that issues containing illegal word, user in the text message that user issues contains illegal contents etc..
Step c, according to operator's testing result, Probability Point of the abnormal user behavior discrimination model to such user behavior
It is updated.That is, if finding, this behavior is the behavior of user inferior, and the user behavior data is judged to bad data, improves
The Probability Point of the user behavior is for example improved Probability Point to 0.999 to certain higher probit.
The new bad data for passing through detection discovery is added Sample Storehouse as training data by step d.That is, will be judged as
When bad data user behavior data add Sample Storehouse in, as the training data of model next time, so as to the renewal for model
New training data is provided.
Step e, using new training data training pattern.
Preferably, in step e, for the inaccurate user behavior data of Probability Point carries out off-line analysiss, find new
User behavior feature simultaneously selects suitable feature.And do to the model for newly producing whether cross validation judgment models have preferably
Performance.
Using the above-mentioned detection method based on clustering rule to data on line, can realize not accurate enough in sample data
In the case of detection to user behavior data.Also, can be realized with semi-supervised engineering to step e using above-mentioned steps a
Habit mode carries out the renewal to model.Also, by present mechanism, can avoid due to containing in the sample data as training data
The inaccurate problem for having the model caused by the reasons such as noise data to calculate the Probability Point of data, even if so in sample number
According to also can be good at being identified user behavior inferior in the case of inaccurate.
Abnormal user data are processed by step 160.After determining that certain user behavior is Deviant Behavior, system meeting
User inferior is processed, for example, user is deleted etc. in the information of Web realease.
Using the fraud identifying system based on machine learning in a kind of classification information website of the present invention, Neng Goucong
Multiple dimensions are identified to the behavior of user, so as to efficiently reduce the false amount of Transaction Information, improve the true of Transaction Information
Reality.Even if also, also can be good at carrying out user behavior inferior in the case of containing noise data in training data
Identification.
In conjunction with the explanation and practice of the present invention for disclosing here, the other embodiment of the present invention is for those skilled in the art
All will be readily apparent and understand.Illustrate and embodiment be to be considered only as exemplary, the present invention true scope and purport equal
It is defined in the claims.
Claims (9)
1. in a kind of website for classification information based on machine learning fraud identifying system method, methods described includes
Following steps：
A) existing user behavior data sample drawn data are based on, for generation model first；
B) for the multiple user behavior features of training data selective extraction of different service types；
C) based on the user behavior feature that is extracted, vectorization is carried out to the sample training data；
D) forecast model is produced using the sample training data of vectorization；
E) data on line are detected based on classification and clustering rule using produced model, wherein,
Included based on user's anomaly detection method of clustering rule as follows：
E1) Probability Point is carried out clustering phenomenon monitoring；
E2) Probability Point cluster is detected to a number of user behavior, to judge the user clustered to equal probabilities point
Whether behavior is user behavior inferior；
E3) according to testing result, abnormal user behavior discrimination model is updated to the Probability Point of such user behavior；
E4 the new bad data for passing through detection discovery is added Sample Storehouse as training data)；E5) instructed using new training data
Practice model；
F) the detected abnormal user data for obtaining are processed.
2. the method for claim 1, the sample data in wherein described step a include positive sample data and negative sample number
According to corresponding respectively to the user of high-quality behavior and the user of behavior inferior.
3. the method for claim 1, in wherein described step b, user behavior feature is included for the use of same cookie
Family behavioral data and the statistical magnitude of each dimension of user.
4. the method for claim 1, intersects data verification by calculating comentropy and model in wherein described step b
Mode come select to different service types extract user characteristicses.
5. the method for claim 1, used in wherein described step d, the grader of probabilistic type carries out decision-making.
6. the method for claim 1, calculates the exception for representing user behavior data using model in wherein described step e
The Probability Point of probability.
7. method as claimed in claim 6, the method for wherein calculating the Probability Point is that multiple models are respectively to user behavior
Many stack features of data are detected, and draw a point of Probability Point respectively, then carry out sum of products conversion to each point of Probability Point
Operation, draws the Probability Point of user behavior data.
8. the method for claim 1, the user's anomaly detection method in wherein described step e based on classifying ruless
It is used for judging whether user behavior data is bad data including setting a probability line.
9. the method for claim 1, for the inaccurate user behavior data of Probability Point enters in wherein described step e5
Row off-line analysiss, find new user behavior feature and select suitable feature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410022138.1A CN103793484B (en) | 2014-01-17 | 2014-01-17 | The fraud identifying system based on machine learning in classification information website |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410022138.1A CN103793484B (en) | 2014-01-17 | 2014-01-17 | The fraud identifying system based on machine learning in classification information website |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103793484A CN103793484A (en) | 2014-05-14 |
| CN103793484B true CN103793484B (en) | 2017-03-15 |
Family
ID=50669150
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410022138.1A Active CN103793484B (en) | 2014-01-17 | 2014-01-17 | The fraud identifying system based on machine learning in classification information website |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103793484B (en) |
Families Citing this family (47)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447755B (en) * | 2014-09-01 | 2021-12-17 | 创新先进技术有限公司 | Transaction control method and device |
| CN104361010A (en) * | 2014-10-11 | 2015-02-18 | 北京中搜网络技术股份有限公司 | Automatic classification method for correcting news classification |
| CN104463668A (en) * | 2014-10-24 | 2015-03-25 | 南京邦科威信息科技有限公司 | Online credit checking method and device |
| CN105703966A (en) * | 2014-11-27 | 2016-06-22 | 阿里巴巴集团控股有限公司 | Internet behavior risk identification method and apparatus |
| CN104636912A (en) * | 2015-02-13 | 2015-05-20 | 银联智惠信息服务（上海）有限公司 | Identification method and device for withdrawal of credit cards |
| CN106296406A (en) * | 2015-05-13 | 2017-01-04 | 阿里巴巴集团控股有限公司 | The processing method and processing device of interaction data |
| CN106294508B (en) * | 2015-06-10 | 2020-02-11 | 深圳市腾讯计算机系统有限公司 | Brushing amount tool detection method and device |
| CN105306495B (en) * | 2015-11-30 | 2018-06-19 | 百度在线网络技术（北京）有限公司 | user identification method and device |
| CN109146151A (en) * | 2016-02-05 | 2019-01-04 | 第四范式（北京）技术有限公司 | There is provided or obtain the method, apparatus and forecasting system of prediction result |
| RU2626337C1 (en) * | 2016-02-18 | 2017-07-26 | Акционерное общество "Лаборатория Касперского" | Method of detecting fraudulent activity on user device |
| CN105808639B (en) * | 2016-02-24 | 2021-02-09 | 平安科技（深圳）有限公司 | Network access behavior identification method and device |
| CN107169768B (en) * | 2016-03-07 | 2021-07-27 | 阿里巴巴集团控股有限公司 | Method and device for acquiring abnormal transaction data |
| CN105844501A (en) * | 2016-05-18 | 2016-08-10 | 上海亿保健康管理有限公司 | Consumption behavior risk control system and method |
| CN107465648B (en) * | 2016-06-06 | 2020-09-04 | 腾讯科技（深圳）有限公司 | Abnormal equipment identification method and device |
| CN106096657B (en) * | 2016-06-13 | 2019-04-30 | 第四范式（北京）技术有限公司 | Based on machine learning come the method and system of prediction data audit target |
| CN106384282A (en) * | 2016-06-14 | 2017-02-08 | 平安科技（深圳）有限公司 | Method and device for building decision-making model |
| CN106228410A (en) * | 2016-07-29 | 2016-12-14 | 武汉斗鱼网络科技有限公司 | Virtual present task anti-brush system and method in a kind of live platform |
| CN106230849B (en) * | 2016-08-22 | 2019-04-19 | 中国科学院信息工程研究所 | A kind of smart machine machine learning safety monitoring system based on user behavior |
| CN106408411A (en) * | 2016-08-31 | 2017-02-15 | 北京城市网邻信息技术有限公司 | Credit assessment method and device |
| CN106548343B (en) * | 2016-10-21 | 2020-11-10 | 中国银联股份有限公司 | Illegal transaction detection method and device |
| CN106682067B (en) * | 2016-11-08 | 2018-05-01 | 浙江邦盛科技有限公司 | A kind of anti-fake monitoring system of machine learning based on transaction data |
| CN108108743B (en) * | 2016-11-24 | 2022-06-24 | 百度在线网络技术（北京）有限公司 | Abnormal user identification method and device for identifying abnormal user |
| CN106682985B (en) * | 2016-12-26 | 2020-03-27 | 深圳先进技术研究院 | Financial fraud identification method and system |
| CN108512682B (en) * | 2017-02-28 | 2021-02-26 | 腾讯科技（深圳）有限公司 | Method and device for determining false terminal identification |
| CN108875761B (en) * | 2017-05-11 | 2022-06-28 | 华为技术有限公司 | Method and device for expanding potential users |
| CN107335220B (en) * | 2017-06-06 | 2021-01-26 | 广州华多网络科技有限公司 | Negative user identification method and device and server |
| WO2019061376A1 (en) * | 2017-09-30 | 2019-04-04 | 深圳市得道健康管理有限公司 | Method for evaluating internet behavior and network terminal |
| CN109416700A (en) * | 2017-09-30 | 2019-03-01 | 深圳市得道健康管理有限公司 | A kind of the classification based training method and the network terminal of internet behavior |
| CN107730717B (en) * | 2017-10-31 | 2019-08-30 | 华中科技大学 | A kind of suspicious card identification method of public transport based on feature extraction |
| CN110097066B (en) * | 2018-01-31 | 2024-01-05 | 阿里巴巴集团控股有限公司 | User classification method and device and electronic equipment |
| CN108550052A (en) * | 2018-04-03 | 2018-09-18 | 杭州呯嘭智能技术有限公司 | Brush list detection method and system based on user behavior data feature |
| JP7091174B2 (en) * | 2018-07-09 | 2022-06-27 | キヤノン株式会社 | System, system control method and program |
| CN110796153B (en) * | 2018-08-01 | 2023-06-20 | 阿里巴巴集团控股有限公司 | Training sample processing method and device |
| CN109325779A (en) * | 2018-08-20 | 2019-02-12 | 北京数美时代科技有限公司 | A kind of read-write portrait method, system and portrait processing system cheated for counter |
| CN109389494B (en) * | 2018-10-25 | 2021-11-05 | 北京芯盾时代科技有限公司 | Loan fraud detection model training method, loan fraud detection method and device |
| CN109886284B (en) * | 2018-12-12 | 2021-02-12 | 同济大学 | Fraud detection method and system based on hierarchical clustering |
| CN109714636B (en) * | 2018-12-21 | 2021-04-23 | 武汉瓯越网视有限公司 | User identification method, device, equipment and medium |
| CN111476258B (en) * | 2019-01-24 | 2024-01-05 | 杭州海康威视数字技术股份有限公司 | Feature extraction method and device based on attention mechanism and electronic equipment |
| CN109829713B (en) * | 2019-01-28 | 2020-09-15 | 重庆邮电大学 | Mobile payment mode identification method based on common drive of knowledge and data |
| CN110557447B (en) * | 2019-08-26 | 2022-06-10 | 腾讯科技（武汉）有限公司 | User behavior identification method and device, storage medium and server |
| CN112533208A (en) * | 2019-08-27 | 2021-03-19 | 中国移动通信有限公司研究院 | Model training method, false terminal identification method and device, and electronic device |
| CN111259985B (en) * | 2020-02-19 | 2023-06-30 | 腾讯云计算(长沙)有限责任公司 | Classification model training method and device based on business safety and storage medium |
| CN111639681A (en) * | 2020-05-09 | 2020-09-08 | 同济大学 | Early warning method, system, medium and device based on education drive type fraud |
| CN111914645A (en) * | 2020-06-30 | 2020-11-10 | 五八有限公司 | Method and device for identifying false information, electronic equipment and storage medium |
| CN112818142B (en) * | 2021-01-29 | 2023-12-08 | 北京达佳互联信息技术有限公司 | Account behavior information processing method and device, electronic equipment and storage medium |
| CN112887325B (en) * | 2021-02-19 | 2022-04-01 | 浙江警察学院 | Telecommunication network fraud crime fraud identification method based on network flow |
| CN113506084A (en) * | 2021-06-23 | 2021-10-15 | 上海师范大学 | False recruitment position detection method based on deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1649311A (en) * | 2005-03-23 | 2005-08-03 | 北京首信科技有限公司 | Detecting system and method for user behaviour abnormal based on machine study |
| CN102176698A (en) * | 2010-12-20 | 2011-09-07 | 北京邮电大学 | Method for detecting abnormal behaviors of user based on transfer learning |
| CN102238045A (en) * | 2010-04-27 | 2011-11-09 | 广州迈联计算机科技有限公司 | System and method for predicting user behavior in wireless Internet |
| CN102402517A (en) * | 2010-09-09 | 2012-04-04 | 北京启明星辰信息技术股份有限公司 | Method and system for establishing normal database login model and method and system for detecting abnormal login behavior |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2009152465A1 (en) * | 2008-06-12 | 2009-12-17 | Guardian Analytics, Inc. | Modeling users for fraud detection and analysis |
-
2014
- 2014-01-17 CN CN201410022138.1A patent/CN103793484B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1649311A (en) * | 2005-03-23 | 2005-08-03 | 北京首信科技有限公司 | Detecting system and method for user behaviour abnormal based on machine study |
| CN102238045A (en) * | 2010-04-27 | 2011-11-09 | 广州迈联计算机科技有限公司 | System and method for predicting user behavior in wireless Internet |
| CN102402517A (en) * | 2010-09-09 | 2012-04-04 | 北京启明星辰信息技术股份有限公司 | Method and system for establishing normal database login model and method and system for detecting abnormal login behavior |
| CN102176698A (en) * | 2010-12-20 | 2011-09-07 | 北京邮电大学 | Method for detecting abnormal behaviors of user based on transfer learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103793484A (en) | 2014-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103793484B (en) | The fraud identifying system based on machine learning in classification information website | |
| Morstatter et al. | A new approach to bot detection: striking the balance between precision and recall | |
| US10484413B2 (en) | System and a method for detecting anomalous activities in a blockchain network | |
| Yu et al. | Attention-based convolutional approach for misinformation identification from massive and noisy microblog posts | |
| Pozzana et al. | Measuring bot and human behavioral dynamics | |
| CN103795612B (en) | Rubbish and illegal information detecting method in instant messaging | |
| CN107807941B (en) | Information processing method and device | |
| CN111371767B (en) | Malicious account identification method, malicious account identification device, medium and electronic device | |
| CN103530540A (en) | User identity attribute detection method based on man-machine interaction behavior characteristics | |
| CN105574544A (en) | Data processing method and device | |
| CN108833139B (en) | OSSEC alarm data aggregation method based on category attribute division | |
| CN104040963A (en) | System and methods for spam detection using frequency spectra of character strings | |
| KR102124846B1 (en) | Source analysis based news reliability evaluation system and method thereof | |
| CN112329816A (en) | Data classification method and device, electronic equipment and readable storage medium | |
| CN109063736B (en) | Data classification method and device, electronic equipment and computer readable storage medium | |
| CN110598129B (en) | Cross-social network user identity recognition method based on two-stage information entropy | |
| CN106537387B (en) | Retrieval/storage image associated with event | |
| CN106202126B (en) | A kind of data analysing method and device for logistics monitoring | |
| WO2024067387A1 (en) | User portrait generation method based on characteristic variable scoring, device, vehicle, and storage medium | |
| WO2019200739A1 (en) | Data fraud identification method, apparatus, computer device, and storage medium | |
| CN103853744A (en) | Deceptive junk comment detection method oriented to user generated contents | |
| Yamak et al. | Detection of multiple identity manipulation in collaborative projects | |
| CN115545103A (en) | Abnormal data identification method, label identification method and abnormal data identification device | |
| CN113157871B (en) | News public opinion text processing method, server and medium applying artificial intelligence | |
| CN109478219A (en) | For showing the user interface of network analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |