CN105844501A - Consumption behavior risk control system and method - Google Patents
Consumption behavior risk control system and method Download PDFInfo
- Publication number
- CN105844501A CN105844501A CN201610329807.9A CN201610329807A CN105844501A CN 105844501 A CN105844501 A CN 105844501A CN 201610329807 A CN201610329807 A CN 201610329807A CN 105844501 A CN105844501 A CN 105844501A
- Authority
- CN
- China
- Prior art keywords
- consumer
- consumer record
- risk
- record
- target
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Marketing (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 invention provides a consumption behavior risk control system and method. The system comprises a data preprocessing module, a model training module, a risk prediction module and a risk control module, wherein the data preprocessing module, the model training module, the risk prediction module and the risk control module are connected in sequence. Consumption record data and risk return feedback data are imported. Some consumption records are sampled from the imported consumption record data, and are used as a consumption record training set with the risk feedback data. According to history consumption records of a consumer, a consumption record sequence training set is generated. The consumption record sequence training set is used to train a classifier. The trained classifier is used to predict a target consumption record sequence corresponding to target consumption records. A risk value corresponding to the target consumption records is predicted. According to the predicted risk value, risk control is carried out on the consumer of the target consumption records. According to the system and the method, a model is used to capture the influence of history consumption behaviors on current consumption, so that a risk control effect is great.
Description
Technical field
The invention belongs to Network Monitoring Technology field, particularly relate to the risk control system of a kind of consuming behavior
System and method.
Background technology
The Internet provides easily simultaneously to life, also provides convenient for the network crime.In recent years,
Online fraud constantly occurs, becomes topic of concern.Therefore carry out effectively at line service
Risk control, for consumer provide safety consumption on network environment, to present Internet enterprises particularly
Important.
The risk control of prior art is typically based on the judgement of rule, or uses machine learning to current line
For characteristic, use classification or Clustering Model feature is classified.
The quality of rule is depended in the performance of rule-based method, it usually needs domain expert is by rule of thumb
Editor's fully and suitably rule, needs substantial amounts of manpower, and rule is with subjectivity, and performance can
Can be unstable.
Use machine learning method carry out unusual checking, can according to the transaction data of history,
By supervision or unsupervised method, the behavioral data that user is each is made risk probability prediction,
There is greater flexibility.But existing method the most only considers the current consumption behavior of user, knot
Close Feature Engineering derive some characteristics be predicted, cannot automatically studying history data to ought
Move ahead for impact.But generally consuming behavior is the most abnormal, and the continuous a series of consumption before it
Behavior is relevant.But the target that this kind of method processes is single independent data point, so typically requiring
Extracting, by methods such as time windows, the feature that historical behavior is relevant, the work to Feature Engineering has necessarily
Requirement, need certain experience to select effective feature.
The patent of invention of such as Application No. CN201010001479.2 proposes a kind of risk control system
System and method, tables of data is shown as independent point by the method, trains a Logic Regression Models to predict
The risk probability of data point.But the most abnormal and targeted customer the history consuming behavior of consuming behavior is tight
Close association, in different consumption history contexts, has different unusual determinations.This method exists
When processing the outlier detection problem of band context, need to first pass through Feature Engineering and generate contextual feature,
It is encoded into a point data.The process of this feature engineering needs manually generated effective feature, model
It is the most reasonable that effect is heavily dependent on feature.
Summary of the invention
It is an object of the invention to provide a kind of risk control system and the method for consuming behavior, by using
Grader models, and using the sequence of history consumer record data as the input of model, is caught by model
Catch the impact on current consumption of the history consuming behavior, it is not necessary to extract historical behavior feature by rule of thumb,
Improve the risk control effect of model.
To achieve these goals, technical solution of the present invention is as follows:
A kind of risk control system of consuming behavior, for the consumer record of consumer is detected,
And carrying out risk control, described risk control system includes data preprocessing module, the mould being sequentially connected
Type training module, risk profile module and risk control module, wherein:
Described data preprocessing module, is used for importing consumer record data and risk pays a return visit feedback data,
Sample certain consumer record from the consumer record data imported, and risk feedback data together as
Consumer record training set, and according to the history consumer record of consumer, generate the training of consumer record sequence
Collection；
Described model training module, is used for using consumer record sequence training set to train grader；
Described risk profile module, corresponding to target consumer record for the grader obtained by training
Target consumer records series be predicted, it was predicted that go out the value-at-risk that target consumer record is corresponding；
Described risk control module, for the value-at-risk according to prediction, the consumption to target consumer record
Person carries out risk control.
Wherein, described data preprocessing module, according to the history consumer record of consumer, generates consumption note
Record sequence training set, performs to operate as follows:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
Further, described risk profile module also includes sequence extension submodule, described sequence extension
Submodule, for the history consumer record of the consumer according to target consumer record, remembers target consumer
Record expands to target consumer records series.
Further, the history consumption note of the consumer of described sequence extension submodule target consumer record
Record, expands to target consumer records series by target consumer record, including:
According to the window size set, during the consumption that in the range of selected window, this target consumer record is corresponding
Before between, the history consumer record of this consumer, generates corresponding target together with this target consumer record and disappears
Take records series.
Further, described risk control module also includes feedback capture submodule, for according to wind
Danger value carries out the result paid a return visit less than the consumer of the target consumer record setting threshold value, is disappeared by corresponding
Expense record is put in risk feedback data.
The invention allows for the risk control method of a kind of consuming behavior, for consuming behavior is carried out
Detection, and carry out risk control, described risk control method includes:
Import consumer record data and risk pays a return visit feedback data, adopt from the consumer record data imported
The consumer record that sample is certain, and risk feedback data is together as consumer record training set, and according to disappearing
The history consumer record of the person of expense, generates consumer record sequence training set；
Use consumer record sequence training set training grader；
The target consumer records series that target consumer record is corresponding is entered by the grader obtained by training
Row prediction, it was predicted that go out the value-at-risk that target consumer record is corresponding；
According to the value-at-risk of prediction, the consumer of target consumer record is carried out risk control.
Wherein, the described history consumer record according to consumer, generate consumer record sequence training set,
Including:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
Further, the target that target consumer record is corresponding is disappeared by the described grader obtained by training
Expense records series is predicted, it was predicted that goes out the value-at-risk that target consumer record is corresponding, also includes:
The history consumer record of the consumer according to target consumer record, expands to target consumer record
Target consumer records series.
Further, the history consumer record of the described consumer according to target consumer record, by target
Consumer record expands to target consumer records series, including:
According to the window size set, during the consumption that in the range of selected window, this target consumer record is corresponding
Before between, the history consumer record of this consumer, generates corresponding target together with this target consumer record and disappears
Take records series.
Further, the described value-at-risk according to prediction, the consumer of target consumer record is carried out wind
Danger controls, and also includes:
According to the knot that value-at-risk is paid a return visit less than the consumer of the target consumer record setting threshold value
Really, corresponding consumer record is put in risk feedback data.
The present invention proposes a kind of risk control system and the method for consuming behavior, by using RNN
(recurrent neural network) grader, using the sequence of history consumer record data as the input of model,
The impact on current consumption of the history consuming behavior is caught, it is not necessary to extract by rule of thumb and go through by model
History behavior characteristics, risk control effect is more preferable.Additionally, the method can be caught by multilayer neural network
Catch the non-linear relation between feature and value-at-risk, it is not necessary in Feature Engineering, find rational non-thread
Property function carrys out converting characteristic.
Accompanying drawing explanation
Fig. 1 is the structural representation of risk control system of the present invention；
Fig. 2 is embodiment of the present invention risk control principle schematic；
Fig. 3 is risk control method flow chart of the present invention.
Detailed description of the invention
With embodiment, technical solution of the present invention is described in further details below in conjunction with the accompanying drawings, real below
Execute example and do not constitute limitation of the invention.
The overall thought of the present invention is to combine context to carry out risk control, containing the risk of context
Control unlike the general behavioral value as independent point, process the feature of target behavior self,
It is also contemplated that context residing for the behavior.A same behavior under different contexts, Ke Nengyou
Different risk implications.Concrete, the present invention is it is of concern that this context of history consuming behavior.Existing
Some technology typically by Feature Engineering, generate the feature that the consumption of some history is relevant, problem are converted
Become not comprise the independent digit strong point of context, re-use correlation model and detect.The effect of this method
Really quality is heavily dependent on and how to select and to generate feature.By the present invention in that and use grader mould
Type catches the history consumer record context as target data, can learn automatically context
Dependence with number of targets strong point, it is to avoid artificial feature extraction.
As it is shown in figure 1, the risk control system of a kind of consuming behavior of the present embodiment, locate in advance including data
Reason module, model training module, risk profile module and risk control module.Wherein:
Data preprocessing module, is used for importing consumer record data and risk pays a return visit feedback data, from leading
The consumer record data entered are sampled certain consumer record, and risk feedback data is together as consumption
Record training set, and according to the history consumer record of consumer, generate consumer record sequence training set.
Specifically, data preprocessing module includes that data import submodule and feature extraction submodule.Its
Middle data import submodule and are mainly used in periodically importing consumer record data and wind from Service Database
Feedback data is paid a return visit in danger, sample from the consumer record data imported certain consumer record, and risk
Feedback data is together as consumer record training set.And feature extraction submodule is for according to consumer
History consumer record, generates consumer record sequence training set.
Data preprocessing module is paid a return visit feedback data according to consumer record data and risk and is generated consumption note
Record sequence training set, for carrying out the training of grader.For machine learning, when training data is enough
Time, the sorter model parameter that training obtains is the most accurate, and the effect of classification is the best.Actually business number
Being ever-increasing according to the consumer record in storehouse, increment consumer record can be used to train grader, right
Model parameter is updated.Therefore the present embodiment uses and periodically imports consumption note from Service Database
Record data and risk pay a return visit feedback data, in order to follow-up carry out machine learning update model parameter so that
The effect of grader is more preferable.Comparing the training disposably completing grader, model parameter has obtained more
Newly, effect is more preferable.
The present embodiment is preferably by RNN (recurrent neural network) grader, but is not limited to classification
The concrete kind of device, such as, can use hidden Markov model (HMM), but RNN effect is excellent
In HMM.In RUN grader, it is also possible to use LSTM RNN grader, or use
GRU RNN grader.
It is easily understood that for periodically importing data, when importing first, be introduced into whole
Consumer record data and risk pay a return visit feedback data, and when follow-up importing, only import increment consumption
Record data.The data imported store off-line storehouse, in order to be used for training RNN grader.
Wherein, consumer record data are the consumer record not being identified, and risk return visit feedback data is
Carried out return visit, it has been determined that whether be abnormal consumer record.
One consumer record of the present embodiment, including user name u, spending amount f and consumption time t, institute
State the feature extraction submodule history consumer record according to consumer, generate consumer record sequence training set,
Concrete grammar is as follows:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
Such as consumer record training set includes:
(u1, f1, t1)；
(u2, f2, t2)；
(u1, f3, t3)；
(u2, f4, t4)；
(u3, f5, t5)；
(u1, f6, t6)；
(u2, f7, t7)；
…。
Assume that the window size that sets as current consumption record and front 2 history consumer records thereof, is then given birth to
Become following consumer record sequence:
[(u1, f1, t1), (u1, f3, t3), (u1, f6, t6)]
[(u2, f2, t2), (u2, f4, t4), (u2, f7, t7)]
[(u1, f3, t3), (u1, f6, t6) ...]
…。
The bar number of the record that in the present embodiment, window size comprises in can being consumer record sequence, it is possible to
To be time range corresponding to consumer record, the such as consumption of the history in current time starts a few days ago
Record.
It should be noted that the acquisition of consumer record sequence can be the consumption note recording each consumer
Record, extracts consumer record from the consumer record that consumer is corresponding and generates consumer record sequence.Also may be used
To be to take out all consumer records, according to the window size set, select disappearing in window ranges
Take record and generate consumer record sequence.
Model training module, is used for using consumer record sequence training set to train grader.
The present embodiment uses RNN (recurrent neural network) to model, the consumer record every time obtained
RNN grader can be trained by sequence training set, and training obtains model parameter and preserves, and
It is identified predicting to real-time consumption record according to the model parameter updated.
Risk profile module, the mesh that grader for being obtained by training is corresponding to target consumer record
Mark consumer record sequence is predicted, it was predicted that go out the value-at-risk that target consumer record is corresponding.
The present embodiment target consumer records series, is that corresponding the comprising of real-time consumption record of consumer is gone through
The target consumer records series of history consumer record.Target consumer records series can input after previously generating
To risk profile module, such as directly read in the target consumer record of consumer by data preprocessing module,
Through process generate target consumer records series, be then input to risk profile module predict obtain right
The value-at-risk answered.The target consumer record of consumer can also be input to risk profile module, by wind
The sequence extension submodule that danger prediction module includes, disappears according to the history of the consumer of target consumer record
Take record, target consumer record is expanded to target consumer records series.Sequence extension submodule is by mesh
Mark consumer record expands to the method for target consumer records series and generates target with data preprocessing module
The method of consumer record sequence is identical.
The preparation method of this target consumer records series is also according to the window size set, selected window
In the range of the history consumer record of this consumer before consumption time corresponding to this target consumer record, with this
Target consumer record generates the target consumer records series of correspondence together.
As in figure 2 it is shown, target consumer records the consumption that corresponding history consumer record composition is corresponding
Records series, the consumer record bar number included in window ranges is exactly the size of window, this consumption is remembered
Record sequence inputting is predicted to RNN grader, it was predicted that go out the value-at-risk that target consumer record is corresponding.
Risk control module, for the value-at-risk according to prediction, enters the consumer of target consumer record
Row risk control.
The present embodiment carries out risk control to the consumer of target consumer record, is the risk according to prediction
Value carries out risk control to consumer, includes but not limited to:
Not intervene (value-at-risk is less than 0.5), put into list to be paid a return visit (value-at-risk 0.5～0.8 it
Between), freeze consume (value-at-risk is more than 0.8).
Above-mentioned value-at-risk is only a specific embodiment, and different according to business rule, its value can be carried out
Adjust.
The present embodiment risk control module also includes feedback capture submodule, and this feedback capture submodule is used
According to the result that value-at-risk is paid a return visit less than the consumer of the target consumer record setting threshold value
Corresponding consumer record is put in risk feedback data.
Such as to put into list to be paid a return visit (value-at-risk 0.5～～0.8 between) and freeze consume (risk
Value is more than 0.8) consumer corresponding to consumer record pay a return visit, threshold value is 0.5, according to return visit
As a result, as this consumer's phone does not repeatedly connect, it is unwilling to examine identity, or feedack is incorrect,
Judge that this consumer is as abnormal or normal, for abnormal Frozen Account, for account of thawing normally.
And identify corresponding consumer record, put in risk feedback data, in order to follow-up to RNN grader
It is trained.
By constantly updating risk feedback data, can persistently RNN grader be trained, more
New RNN sorter model parameter, to obtain predicting the outcome more accurately.If the most so carried out
Update, only in accordance with original RNN grader, then can not carry out model parameter according to news
Update, cause RNN grader not adapt to news.
The risk control method of a kind of consuming behavior of the present embodiment, as it is shown on figure 3, for consumption row
For detecting, and carrying out risk control, this risk control method includes:
Import consumer record data and risk pays a return visit feedback data, adopt from the consumer record data imported
The consumer record that sample is certain, and risk feedback data is together as consumer record training set, and according to disappearing
The history consumer record of the person of expense, generates consumer record sequence training set；
Use consumer record sequence training set training grader；
The target consumer records series that target consumer record is corresponding is entered by the grader obtained by training
Row prediction, it was predicted that go out the value-at-risk that target consumer record is corresponding；
According to the value-at-risk of prediction, the consumer of target consumer record is carried out risk control.
Wherein, according to the history consumer record of consumer, generate consumer record sequence training set, with number
Accordingly, method for optimizing is Data preprocess module:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
Similarly, with risk profile module accordingly, the grader obtained by training is to target consumer
The target consumer records series that record is corresponding is predicted, it was predicted that go out the risk that target consumer record is corresponding
Value, also includes:
The history consumer record of the consumer according to target consumer record, expands to target consumer record
Target consumer records series.
Preferably, according to the history consumer record of the consumer of target consumer record, target consumer is remembered
Record expands to target consumer records series, including:
According to the window size set, during the consumption that in the range of selected window, this target consumer record is corresponding
Before between, the history consumer record of this consumer, generates corresponding target together with this target consumer record and disappears
Take records series.
Similarly, with risk control module accordingly, according to the value-at-risk of prediction, target consumer is remembered
The consumer of record carries out risk control, also includes:
According to the knot that value-at-risk is paid a return visit less than the consumer of the target consumer record setting threshold value
Really, corresponding consumer record is put in risk feedback data.
Above example is only limited in order to technical scheme to be described, is not carrying on the back
In the case of present invention spirit and essence thereof, those of ordinary skill in the art work as can be according to the present invention
Make various corresponding change and deformation, but these change accordingly and deformation all should belong to institute of the present invention
Attached scope of the claims.
Claims (10)
1. a risk control system for consuming behavior, for examining the consumer record of consumer
Survey, and carry out risk control, it is characterised in that described risk control system includes the number being sequentially connected
Data preprocess module, model training module, risk profile module and risk control module, wherein:
Described data preprocessing module, is used for importing consumer record data and risk pays a return visit feedback data,
Sample certain consumer record from the consumer record data imported, and risk feedback data together as
Consumer record training set, and according to the history consumer record of consumer, generate the training of consumer record sequence
Collection；
Described model training module, is used for using consumer record sequence training set to train grader；
Described risk profile module, corresponding to target consumer record for the grader obtained by training
Target consumer records series be predicted, it was predicted that go out the value-at-risk that target consumer record is corresponding；
Described risk control module, for the value-at-risk according to prediction, the consumption to target consumer record
Person carries out risk control.
The risk control system of consuming behavior the most according to claim 1, it is characterised in that
Described data preprocessing module, according to the history consumer record of consumer, generates the training of consumer record sequence
Collection, performs to operate as follows:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
The risk control system of consuming behavior the most according to claim 1, it is characterised in that
Described risk profile module also includes sequence extension submodule, and described sequence extension submodule, for root
According to the history consumer record of the consumer of target consumer record, target consumer record is expanded to target and disappears
Take records series.
The risk control system of consuming behavior the most according to claim 3, it is characterised in that
The history consumer record of the consumer of described sequence extension submodule target consumer record, by target consumer
Record expands to target consumer records series, including:
According to the window size set, during the consumption that in the range of selected window, this target consumer record is corresponding
Before between, the history consumer record of this consumer, generates corresponding target together with this target consumer record and disappears
Take records series.
The risk control system of consuming behavior the most according to claim 1, it is characterised in that
Described risk control module also includes feedback capture submodule, for according to value-at-risk is less than setting threshold
The consumer of the target consumer record of value carries out the result paid a return visit, and corresponding consumer record is put into risk
In feedback data.
6. a risk control method for consuming behavior, for detecting consuming behavior, goes forward side by side
Row risk control, it is characterised in that described risk control method includes:
Import consumer record data and risk pays a return visit feedback data, adopt from the consumer record data imported
The consumer record that sample is certain, and risk feedback data is together as consumer record training set, and according to disappearing
The history consumer record of the person of expense, generates consumer record sequence training set；
Use consumer record sequence training set training grader；
The target consumer records series that target consumer record is corresponding is entered by the grader obtained by training
Row prediction, it was predicted that go out the value-at-risk that target consumer record is corresponding；
According to the value-at-risk of prediction, the consumer of target consumer record is carried out risk control.
The risk control method of consuming behavior the most according to claim 6, it is characterised in that
The described history consumer record according to consumer, generates consumer record sequence training set, including:
For the arbitrary consumer record in consumer record training set, according to the window size set, choose
The history consumer record of this consumer before the consumption time that in window ranges, this consumer record is corresponding, with this
Consumer record generates the consumer record sequence of correspondence together；
The consumer record sequence that in consumer record training set, all consumer records are corresponding forms consumer record
Sequence training set.
The risk control method of consuming behavior the most according to claim 6, it is characterised in that
The target consumer records series that target consumer record is corresponding is entered by the described grader obtained by training
Row prediction, it was predicted that go out the value-at-risk that target consumer record is corresponding, also include:
The history consumer record of the consumer according to target consumer record, expands to target consumer record
Target consumer records series.
The risk control method of consuming behavior the most according to claim 8, it is characterised in that
The history consumer record of the described consumer according to target consumer record, expands to target consumer record
Target consumer records series, including:
According to the window size set, during the consumption that in the range of selected window, this target consumer record is corresponding
Before between, the history consumer record of this consumer, generates corresponding target together with this target consumer record and disappears
Take records series.
The risk control method of consuming behavior the most according to claim 6, it is characterised in that
The described value-at-risk according to prediction, carries out risk control to the consumer of target consumer record, also includes:
According to the knot that value-at-risk is paid a return visit less than the consumer of the target consumer record setting threshold value
Really, corresponding consumer record is put in risk feedback data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610329807.9A CN105844501A (en) | 2016-05-18 | 2016-05-18 | Consumption behavior risk control system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610329807.9A CN105844501A (en) | 2016-05-18 | 2016-05-18 | Consumption behavior risk control system and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105844501A true CN105844501A (en) | 2016-08-10 |
Family
ID=56593540
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610329807.9A Pending CN105844501A (en) | 2016-05-18 | 2016-05-18 | Consumption behavior risk control system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105844501A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106888205A (en) * | 2017-01-04 | 2017-06-23 | 浙江大学 | A kind of non-intrusion type is based on the PLC method for detecting abnormality of power consumption analysis |
| CN107563122A (en) * | 2017-09-20 | 2018-01-09 | 长沙学院 | The method of crime prediction of Recognition with Recurrent Neural Network is locally connected based on interleaving time sequence |
| CN108446978A (en) * | 2018-02-12 | 2018-08-24 | 阿里巴巴集团控股有限公司 | Handle the method and device of transaction data |
| WO2018166457A1 (en) * | 2017-03-15 | 2018-09-20 | 阿里巴巴集团控股有限公司 | Neural network model training method and device, transaction behavior risk identification method and device |
| CN108718296A (en) * | 2018-04-27 | 2018-10-30 | 广州西麦科技股份有限公司 | Network management-control method, device and computer readable storage medium based on SDN network |
| CN109697614A (en) * | 2017-10-23 | 2019-04-30 | 北京京东金融科技控股有限公司 | For detecting the method and device of fraud data |
| CN111461865A (en) * | 2020-03-31 | 2020-07-28 | 中国银行股份有限公司 | Data analysis method and device |
| CN113450116A (en) * | 2020-03-24 | 2021-09-28 | 国家计算机网络与信息安全管理中心 | Transaction risk analysis method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090234683A1 (en) * | 2000-06-30 | 2009-09-17 | Russell Anderson | Detecting and Measuring Risk with Predictive Models Using Content Mining |
| CN102567788A (en) * | 2010-12-28 | 2012-07-11 | ***通信集团重庆有限公司 | Real-time identification system and real-time identification method for fraudulent practice in communication services |
| CN103793484A (en) * | 2014-01-17 | 2014-05-14 | 五八同城信息技术有限公司 | Fraudulent conduct identification system based on machine learning in classified information website |
| US20140310159A1 (en) * | 2013-04-10 | 2014-10-16 | Fair Isaac Corporation | Reduced fraud customer impact through purchase propensity |
| CN104778622A (en) * | 2015-04-29 | 2015-07-15 | 清华大学 | Method and system for predicting TPS transaction event threshold value |
-
2016
- 2016-05-18 CN CN201610329807.9A patent/CN105844501A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090234683A1 (en) * | 2000-06-30 | 2009-09-17 | Russell Anderson | Detecting and Measuring Risk with Predictive Models Using Content Mining |
| CN102567788A (en) * | 2010-12-28 | 2012-07-11 | ***通信集团重庆有限公司 | Real-time identification system and real-time identification method for fraudulent practice in communication services |
| US20140310159A1 (en) * | 2013-04-10 | 2014-10-16 | Fair Isaac Corporation | Reduced fraud customer impact through purchase propensity |
| CN103793484A (en) * | 2014-01-17 | 2014-05-14 | 五八同城信息技术有限公司 | Fraudulent conduct identification system based on machine learning in classified information website |
| CN104778622A (en) * | 2015-04-29 | 2015-07-15 | 清华大学 | Method and system for predicting TPS transaction event threshold value |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106888205B (en) * | 2017-01-04 | 2020-02-18 | 浙江大学 | Non-invasive PLC anomaly detection method based on power consumption analysis |
| CN106888205A (en) * | 2017-01-04 | 2017-06-23 | 浙江大学 | A kind of non-intrusion type is based on the PLC method for detecting abnormality of power consumption analysis |
| CN108629413B (en) * | 2017-03-15 | 2020-06-16 | 创新先进技术有限公司 | Neural network model training and transaction behavior risk identification method and device |
| WO2018166457A1 (en) * | 2017-03-15 | 2018-09-20 | 阿里巴巴集团控股有限公司 | Neural network model training method and device, transaction behavior risk identification method and device |
| CN108629413A (en) * | 2017-03-15 | 2018-10-09 | 阿里巴巴集团控股有限公司 | Neural network model training, trading activity Risk Identification Method and device |
| TWI689874B (en) * | 2017-03-15 | 2020-04-01 | 香港商阿里巴巴集團服務有限公司 | Method and device for neural network model training and transaction behavior risk identification |
| CN107563122B (en) * | 2017-09-20 | 2020-05-19 | 长沙学院 | Crime prediction method based on interleaving time sequence local connection cyclic neural network |
| CN107563122A (en) * | 2017-09-20 | 2018-01-09 | 长沙学院 | The method of crime prediction of Recognition with Recurrent Neural Network is locally connected based on interleaving time sequence |
| CN109697614A (en) * | 2017-10-23 | 2019-04-30 | 北京京东金融科技控股有限公司 | For detecting the method and device of fraud data |
| CN108446978A (en) * | 2018-02-12 | 2018-08-24 | 阿里巴巴集团控股有限公司 | Handle the method and device of transaction data |
| CN108718296A (en) * | 2018-04-27 | 2018-10-30 | 广州西麦科技股份有限公司 | Network management-control method, device and computer readable storage medium based on SDN network |
| CN113450116A (en) * | 2020-03-24 | 2021-09-28 | 国家计算机网络与信息安全管理中心 | Transaction risk analysis method and device, electronic equipment and storage medium |
| CN111461865A (en) * | 2020-03-31 | 2020-07-28 | 中国银行股份有限公司 | Data analysis method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105844501A (en) | Consumption behavior risk control system and method | |
| CN103974311B (en) | Based on the Condition Monitoring Data throat floater detection method for improving Gaussian process regression model | |
| CN103336906B (en) | The sampling Gaussian process regression model that in the image data stream of environmental sensor, continuous abnormal detects | |
| US20190004484A1 (en) | Combined method for detecting anomalies in a water distribution system | |
| CN109034194B (en) | Transaction fraud behavior deep detection method based on feature differentiation | |
| CN106453293A (en) | Network security situation prediction method based on improved BPNN (back propagation neural network) | |
| Clements et al. | Sequential deep learning for credit risk monitoring with tabular financial data | |
| CN108052092B (en) | Subway electromechanical equipment state anomaly detection method based on big data analysis | |
| CN110555477A (en) | municipal facility fault prediction method and device | |
| CN112202726B (en) | System anomaly detection method based on context sensing | |
| Hanga et al. | A graph-based approach to interpreting recurrent neural networks in process mining | |
| CN106951565B (en) | File classification method and the text classifier of acquisition | |
| CN110956309A (en) | Flow activity prediction method based on CRF and LSTM | |
| Suryo et al. | Improved time series prediction using LSTM neural network for smart agriculture application | |
| CN110580213A (en) | Database anomaly detection method based on cyclic marking time point process | |
| CN110737948A (en) | method for predicting residual life of aero-engine based on deep FNN-LSTM hybrid network | |
| CN110716957B (en) | Intelligent mining and analyzing method for class case suspicious objects | |
| CN116051175A (en) | Click rate prediction model and prediction method based on depth multi-interest network | |
| CN111079348A (en) | Method and device for detecting slowly-varying signal | |
| McMillan et al. | Flow forecasting for leakage burst prediction in water distribution systems using long short-term memory neural networks and Kalman filtering | |
| CN116106890A (en) | Radar target track starting method based on quantum particle swarm and LGBM | |
| CN112560252B (en) | Method for predicting residual life of aeroengine | |
| CN113723660A (en) | Specific behavior type prediction method and system based on DNN-LSTM fusion model | |
| Tian et al. | Structure feature extraction for hierarchical alarm flood classification and alarm prediction | |
| Kumari et al. | Remaining useful life prediction using hybrid neural network and genetic algorithm approaches |
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 |
Application publication date: 20160810 |
|
| RJ01 | Rejection of invention patent application after publication |