CN108229542A - A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology - Google Patents
A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology Download PDFInfo
- Publication number
- CN108229542A CN108229542A CN201711395257.1A CN201711395257A CN108229542A CN 108229542 A CN108229542 A CN 108229542A CN 201711395257 A CN201711395257 A CN 201711395257A CN 108229542 A CN108229542 A CN 108229542A
- Authority
- CN
- China
- Prior art keywords
- hidden state
- observation sequence
- borrower
- pen
- debt
- 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
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
Abstract
The present invention relates to a kind of cycles based on Time-Series analysis technology to borrow or lend money credit risk monitoring method, including：The debt-credit data of borrower are extracted from loan application database；It from debt-credit the first observation sequence of extracting data and is handled, obtains the second observation sequence；A GMM HMM model is trained with the second observation sequence, acquires the parameter of GMM HMM models；According to observation sequence and parameter, using forward-backward algorithm algorithm, the probability of the corresponding hidden state of each pen loaning bill comprising borrower is calculated；According to the probability being calculated, the hidden state that each pen of borrower is borrowed money is judged.The present invention is in the case of fine or not label is unknown (or tape label data are limited); it can be according to observation sequence; carry out the unsupervised formula estimation of parameter of model; and it can predict the most probable hidden state (normal refund/promise breaking) of the newest loan of each pen; dynamic updates the risk of completed loan before each pen simultaneously, and collection after borrowing in time can be done for lending mechanism and provides information.
Description
Technical field
The present invention relates to reference fields, and in particular to a kind of cycle debt-credit credit risk monitoring based on Time-Series analysis technology
Method.
Background technology
Risk, which is estimated, to be generally referred to be showed according to the history and current data of debtor, to judge that a certain pen loan occurs
The probability of promise breaking.In current internet financial environment, repeatedly apply, bull application or even altogether debt are a kind of greatly hiding
Risk.If this risk can be captured in time, mechanism progress risk control of making loans will be greatly helped.Existing risk is pre-
The method of estimating belongs to supervised method, needs the training dataset with target variable largely acquired in advance, but because data
The various aspects reason such as business information protection caused by being competed between safety and mechanism, differ surely it is complete be collected into it is each
The each refund details of a debtor can not provide accurate risk for each loan of lending mechanism and estimate.
Invention content
For above-mentioned technical problem, the present invention provides a kind of cycle debt-credit credit risk monitoring based on Time-Series analysis technology
Method.
The technical solution that the present invention solves above-mentioned technical problem is as follows：A kind of cycle debt-credit letter based on Time-Series analysis technology
With risk monitoring method, include the following steps：
The debt-credit data of borrower are extracted from loan application database；
It from debt-credit first observation sequence of extracting data and is handled, obtains the second observation sequence；
A GMM-HMM model is trained with second observation sequence, acquires the parameter of the GMM-HMM models, it is described
Whether the hidden state of GMM-HMM models breaks a contract for the loaning bill of the borrower；
According to the observation sequence and parameter, using forward-backward algorithm algorithm, calculate each pen comprising the borrower and borrow money
The probability of corresponding hidden state；
According to the probability of the corresponding hidden state of each pen loaning bill comprising the borrower, judge the borrower's
The hidden state that each pen is borrowed money, that is, judge whether each pen loaning bill of the borrower breaks a contract.
The beneficial effects of the invention are as follows：It, can be in the case of fine or not label is unknown (or tape label data are limited)
According to observation sequence, the unsupervised formula estimation of parameter of model is carried out, and can predict that the most probable of the newest loan of each pen is hidden
State (normal refund/promise breaking), while dynamic updates the risk of completed loan before each pen, can be done for lending mechanism
Collection provides information after borrowing in time.
Based on the above technical solution, the present invention can also be improved as follows.
Further, it is described the first observation sequence of extracting data and to be handled from the debt-credit, obtain the second observation sequence
Row, specifically include：
The debt-credit data are pre-processed；
Original variable is screened from the debt-credit data by pretreatment；
Variable is derived according to the debt-credit data configuration by pretreatment；
The feature vector of the original variable and derivative variable composition forms the first observation sequence, to the described first observation sequence
Row carry out dimensionality reduction, obtain the second observation sequence.
Further, the original variable includes：Institution Code, product type, industry type, name, identification card number, mobile phone
Number, loan types label, the loan application date, the date of making loans, examine and approve after loan total amount, refund state tag, and/or set
Standby IMEI number.
Further, the derivative variable includes：Age, replace in preset time period mobile phone number, more exchange device number, across
The growth of mechanism number, inter-trade number, debt-credit number, loaning bill total amount, average borrowing balance, demand frequency, current borrowing balance
It is newly opened in rate, current loaning bill difference amount of money growth rate, current interval time growth rate of borrowing money, debt growth rate, default application
Family ratio, present application are opened an account the interval number of days of time and application time, the continuous loaning bill number of current phone number, current phone number
Continuously borrow money number, current industry of accumulative loaning bill number, current facility continuously loaning bill number, and/or is reported as blacklist number.
Further, it is described to train a GMM-HMM model with second observation sequence, acquire the GMM-HMM models
Parameter, specifically include：
Second observation sequence is clustered using K-means algorithms, obtains the number of Gauss model；
According to second observation sequence, using EM algorithms (such as Baum-Welch algorithms), GMM-HMM parameters are obtained.
Further, the probability of the corresponding hidden state of each pen loaning bill comprising the borrower described in the basis, judges
The hidden state that each pen of the borrower is borrowed money, specifically includes：
When the probability of the corresponding hidden state of each pen loaning bill comprising the borrower is more than predetermined threshold value, judge
Corresponding hidden state is promise breaking.
Further, according to the observation sequence and parameter, using viterbi algorithms, it is corresponding to calculate the observation sequence
Hidden state sequence；
According to the corresponding hidden state sequence of the observation sequence, the hiding shape that each pen of the borrower is borrowed money is judged
State, that is, judge whether each pen loaning bill of the borrower breaks a contract.
Further, it is described according to the corresponding hidden state sequence of the observation sequence, judge each pen of the borrower
The hidden state of loaning bill, specifically includes：
By in the corresponding hidden state sequence of the observation sequence, the hidden state sequence of maximum probability is as borrower's
The hidden state that each pen is borrowed money.
Description of the drawings
Fig. 1 is that a kind of cycle based on Time-Series analysis technology provided in an embodiment of the present invention borrows or lends money credit risk monitoring method
Flow chart；
Fig. 2 is the flow chart of step 120 of the embodiment of the present invention；
Fig. 3 is the flow chart of step 130 of the embodiment of the present invention；
Fig. 4 is that another cycle based on Time-Series analysis technology provided in an embodiment of the present invention borrows or lends money credit risk monitoring side
The flow chart of method；
Fig. 5 is that a kind of cycle based on Time-Series analysis technology provided in an embodiment of the present invention borrows or lends money credit risk monitoring method
General frame flow chart.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the present invention.
Fig. 1 is that a kind of cycle based on Time-Series analysis technology provided in an embodiment of the present invention borrows or lends money credit risk monitoring method
Flow chart, as shown in Figure 1, this method includes the following steps：
Step 110, the debt-credit data that borrower is extracted from loan application database；
Step 120 the first observation sequence of extracting data and is handled from the debt-credit, obtains the second observation sequence；
Optionally, in this embodiment, as shown in Fig. 2, step 120 specifically includes：
Step 210 pre-processes the debt-credit data；
Specifically, pretreatment includes carrying out the processes such as denoising, cleaning to debt-credit data.
Step 220 screens original variable from the debt-credit data by pretreatment；
Specifically, original variable is as shown in the table.
Wherein：
(1)termStatus：The refund state tag of a certain pen application needs the serial number refunded by tracking this to obtain
, if the refund record of this application cannot be obtained, which is set as nul l
(2)device.imei：It is described strictly according to the facts if it can obtain equipment I MEI, is otherwise provided as nul l.
Step 230 derives variable according to the debt-credit data configuration by pretreatment；
Specifically, original variable is the essential information of some applications, only by these information, it is not enough to distinguish a certain pen and borrows
Whether money can break a contract, therefore, it is necessary to consider some history loan informations of borrower.In order to keep the stability of feature, need
The history loan information of borrower is carried out regular.Specifically, certain watch window is set, and debtor is in the window for observation
Debt-credit situation in mouthful.In this way, the historical information since the application that both can guarantee different moments is different, help to distinguish it is current this
Pen application is in following situation of honouring an agreement；Can guarantee the application of different moments again has consistent watch window, keeps the steady of feature
It is qualitative.
Construct following derivative variable：
The feature vector of step 240, the original variable and derivative variable composition forms the first observation sequence, to described the
One observation sequence carries out dimensionality reduction, obtains the second observation sequence.
Specifically, above-mentioned original variable and derivative variable totally 39 variables, the feature vectors of 39 dimension of composition.Though each feature
So may be considered continuous variable, but its value remains as natural number, variable-value is still limited, therefore add in it is main into
Analysis PCA processes, the Principle components analysis inside high-dimensional feature is come out.In addition to it is complicated that algorithm can be reduced with dimensionality reduction
Except degree, it can also be floating number by the variables transformations of each dimension, be modeled convenient for mixed Gauss model.
After dimensionality reduction, the feature for selecting 10 dimensions or so is built as the initial data of next step solving model
Mould.
Step 130 trains a GMM-HMM model with second observation sequence, acquires the ginseng of the GMM-HMM models
Whether number, the hidden state of the GMM-HMM models break a contract for the loaning bill of the borrower；
Optionally, in this embodiment, as shown in figure 3, the step 130 specifically includes：
Step 310 clusters second observation sequence using K-means algorithms, obtains the number of Gauss model；
Step 320, according to second observation sequence, using EM algorithms (such as Baum-Welch algorithms), obtain GMM-
HMM parameters.
HMM parameters are λ=(A, B, Π), wherein：
(1) state-transition matrix A：For 2 × 2 matrix, the transition probability of hidden state " promise breaking " and " not breaking a contract " is represented；
(2) initial state probability vector ∏：For the vector of 2 dimensions, the initial of hidden state " promise breaking " and " not breaking a contract " is represented
Probability distribution；
(3) emission probability B：Because arbitrary t moment observation ot, what this feature borrowed money for that can extract was formed
One multi-C vector, value can be discrete or continuous.
A kind of method is, using the method for vector quantization (Vector Quantization), observation discretization to be mapped
To some code book k, the probability b that some state j exports the code book is then calculated againj(k)。
Another method is to choose and construct the feature of continuous value, using continuous probability distribution, usually using mixing
Gauss model approaches, and at this moment emission probability B represents the parameter in mixed Gauss model.Here we use mixed Gauss model
Method.
The method calculated GMM-HMM models is as follows：
(1) to the feature extracted, using the method for principal component analysis (PCA), it is d ' dimensions to carry out the observation after dimensionality reduction
Vector, i.e. ot=(ot 1..., ot d′), t=1 ..., T；
(2) to the feature after dimensionality reduction, using the method for cluster analysis, such as K-means algorithms, K class is obtained；
(3) Gaussian mixtures of construction emission probability B：
Wherein,J=1,2；Nk(x, μk, ∑k) it is d ' dimension Gaussian Profiles.Here B represents unknown parameter
(ω, μ, ∑), ω=(ωjk), j=1,2；K=1 ..., K；μ=(μk), ∑=(∑k), k=1 ..., K.
(4) with parameter lambda=(A, B, the Π) of Baum-Welch algorithms estimation model.
Step 140, according to the observation sequence and parameter, using forward-backward algorithm algorithm, calculate comprising the borrower
Each pen is borrowed money the probability of corresponding hidden state；
The probability of step 150, corresponding hidden state of being borrowed money according to each pen comprising the borrower, described in judgement
The hidden state that each pen of borrower is borrowed money, that is, judge whether each pen loaning bill of the borrower breaks a contract.
Optionally, in this embodiment, the step 150 specifically includes：
When the probability of the corresponding hidden state of each pen loaning bill comprising the borrower is more than predetermined threshold value, judge
Corresponding hidden state is promise breaking.
Specifically, as shown in figure 5, for someone newly-gained loan, if prediction finds its status switch end shape
State is the probability P (h of promise breakingT=1 | O, λ) it is larger and more than threshold value, it is meant that risk that this loan generates promise breaking is high, then right
Current loan application provides alert.
For someone newly-gained loan, if prediction finds certain moment, the probability P (h of loan defaultst=1 | O, λ) compare
Greatly and more than threshold value, it is meant that the risk of this loan defaults is high, then borrowing money to this pen proposes collection early warning.
Optionally, as one embodiment of the present of invention, as shown in figure 4, this method includes：
Step 410, the debt-credit data that borrower is extracted from loan application database；
Step 420 the first observation sequence of extracting data and is handled from the debt-credit, obtains the second observation sequence；
Step 430 trains a GMM-HMM model with second observation sequence, acquires the ginseng of the GMM-HMM models
Whether number, the hidden state of the GMM-HMM models break a contract for the loaning bill of the borrower；
Step 440, according to the observation sequence and parameter, using viterbi algorithms, it is corresponding to calculate the observation sequence
Hidden state sequence；
Step 450, according to the corresponding hidden state sequence of the observation sequence, judge that each pen of the borrower is borrowed money
Hidden state, that is, judge whether each pen loaning bill of the borrower breaks a contract.
Optionally, in this embodiment, the step 450 includes：
By in the corresponding hidden state sequence of the observation sequence, the hidden state sequence of maximum probability is as borrower's
The hidden state that each pen is borrowed money.
Specifically, as shown in figure 5, for someone newly-gained loan, under conditions of known observation sequence, if adopted
Predict that the hidden state sequence end state of its maximum probability of discovery is promise breaking (i.e. h with viterbi algorithmsT=1), then to current
Loan application provides alert.
For someone newly-gained loan, if prediction finds certain moment, hidden state switchs to (i.e. state of breaking a contract by not breaking a contract
htSwitched to 1) by 0, it is meant that the risk of this promise breaking of borrowing money is high, then borrowing money to masking proposes collection early warning.
In addition, can assessment models parameter by the following method validity and precision.
(1) sample known to hidden state is chosen, using their virtual condition as actual value.
(2) assume that the hidden state of these samples is unknown, according to the model parameter acquired, predict the implicit of maximum possible
Status switch, as predicted value.
(3) it is compared by predicted value and with actual value, computation model predictablity rate.
The beneficial effects of the invention are as follows：It, can be in the case of fine or not label is unknown (or tape label data are limited)
According to observation sequence, the unsupervised formula estimation of parameter of model is carried out, and can predict that the most probable of the newest loan of each pen is hidden
State (normal refund/promise breaking), while dynamic updates the risk of completed loan before each pen, can be done for lending mechanism
Collection provides information after borrowing in time.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology, which is characterized in that include the following steps：
The debt-credit data of borrower are extracted from loan application database；
It from debt-credit first observation sequence of extracting data and is handled, obtains the second observation sequence；
A GMM-HMM model is trained with second observation sequence, acquires the parameter of the GMM-HMM models, the GMM-
Whether the hidden state of HMM model breaks a contract for the loaning bill of the borrower；
According to the observation sequence and parameter, using forward-backward algorithm algorithm, calculate each pen loaning bill comprising the borrower and correspond to
Hidden state probability；
According to the probability of the corresponding hidden state of each pen loaning bill comprising the borrower, each pen of the borrower is judged
The hidden state of loaning bill, that is, judge whether each pen loaning bill of the borrower breaks a contract.
It is 2. according to the method described in claim 1, it is characterized in that, described from debt-credit first observation sequence of extracting data
And handled, the second observation sequence is obtained, is specifically included：
The debt-credit data are pre-processed；
Original variable is screened from the debt-credit data by pretreatment；
Variable is derived according to the debt-credit data configuration by pretreatment；
The feature vector of the original variable and derivative variable composition forms the first observation sequence, to first observation sequence into
Row dimensionality reduction obtains the second observation sequence.
3. according to the method described in claim 2, it is characterized in that, the original variable includes：Institution Code, product type,
Industry type, name, identification card number, cell-phone number, loan types label, the loan application date, the date of making loans, examine and approve after loan
The IMEI number of total amount, refund state tag, and/or equipment.
4. according to the method described in claim 2, it is characterized in that, the derivative variable includes：In age, preset time period more
Hand-off machine number, more exchange device number, across mechanism number, inter-trade number, debt-credit number, loaning bill total amount, average borrowing balance, Shen
It please the frequency, growth rate of current borrowing balance, current loaning bill difference amount of money growth rate, current interval time growth rate of borrowing money, negative
New account ratio, present application are opened an account interval number of days, the current phone of time and application time in debt growth rate, default application
Continuously borrow money number, current industry of number continuous loaning bill number, the accumulative loaning bill number of current phone number, current facility is continuously borrowed money time
It counts, and/or is reported as blacklist number.
5. according to claim 1-4 any one of them methods, which is characterized in that described to train one with second observation sequence
A GMM-HMM models acquire the parameter of the GMM-HMM models, specifically include：
Second observation sequence is clustered using K-means algorithms, obtains the number of Gauss model；
According to second observation sequence, using EM algorithms, the parameter of GMM-HMM is obtained.
6. according to the method described in claim 1, it is characterized in that, each pen comprising the borrower is borrowed money described in the basis
The probability of corresponding hidden state judges the hidden state that each pen of the borrower is borrowed money, specifically includes：
When the probability of the corresponding hidden state of each pen loaning bill comprising the borrower is more than predetermined threshold value, judge to correspond to
Hidden state for promise breaking.
7. according to the method described in claim 1, it is characterized in that, according to the observation sequence and parameter, using vi terbi
Algorithm calculates the corresponding hidden state sequence of the observation sequence；
According to the corresponding hidden state sequence of the observation sequence, the hidden state that each pen of the borrower is borrowed money is judged, i.e.,
Judge whether each pen loaning bill of the borrower breaks a contract.
It is 8. the method according to the description of claim 7 is characterized in that described according to the corresponding hidden state sequence of the observation sequence
Row judge the hidden state that each pen of the borrower is borrowed money, specifically include：
By in the corresponding hidden state sequence of the observation sequence, each pen of the hidden state sequence of maximum probability as borrower
The hidden state of loaning bill.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711395257.1A CN108229542A (en) | 2017-12-21 | 2017-12-21 | A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711395257.1A CN108229542A (en) | 2017-12-21 | 2017-12-21 | A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108229542A true CN108229542A (en) | 2018-06-29 |
Family
ID=62647592
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711395257.1A Pending CN108229542A (en) | 2017-12-21 | 2017-12-21 | A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108229542A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110619434A (en) * | 2019-09-17 | 2019-12-27 | 北谷电子有限公司上海分公司 | Method and device for predicting repayment probability, electronic equipment and storage medium |
| CN112927071A (en) * | 2021-04-21 | 2021-06-08 | 顶象科技有限公司 | Post-loan behavior feature processing method and device |
| CN115147203A (en) * | 2022-06-08 | 2022-10-04 | 南京金威诚融科技开发有限公司 | Financial risk intelligent analysis method based on big data |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102270451A (en) * | 2011-08-18 | 2011-12-07 | 安徽科大讯飞信息科技股份有限公司 | Method and system for identifying speaker |
| CN102622535A (en) * | 2012-02-27 | 2012-08-01 | 上海电机学院 | Processing method and processing device based on multiple sequence alignment genetic algorithm |
| CN105512935A (en) * | 2015-12-01 | 2016-04-20 | 郑东东 | Borrowing and lending system based on mobile terminal |
| US20160301893A1 (en) * | 2002-06-25 | 2016-10-13 | International Business Machines Corporation | Personal video recording with messaging |
| CN106779755A (en) * | 2016-12-31 | 2017-05-31 | 湖南文沥征信数据服务有限公司 | A kind of network electric business borrows or lends money methods of risk assessment and model |
| CN107016342A (en) * | 2017-03-06 | 2017-08-04 | 武汉拓扑图智能科技有限公司 | A kind of action identification method and system |
-
2017
- 2017-12-21 CN CN201711395257.1A patent/CN108229542A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160301893A1 (en) * | 2002-06-25 | 2016-10-13 | International Business Machines Corporation | Personal video recording with messaging |
| CN102270451A (en) * | 2011-08-18 | 2011-12-07 | 安徽科大讯飞信息科技股份有限公司 | Method and system for identifying speaker |
| CN102622535A (en) * | 2012-02-27 | 2012-08-01 | 上海电机学院 | Processing method and processing device based on multiple sequence alignment genetic algorithm |
| CN105512935A (en) * | 2015-12-01 | 2016-04-20 | 郑东东 | Borrowing and lending system based on mobile terminal |
| CN106779755A (en) * | 2016-12-31 | 2017-05-31 | 湖南文沥征信数据服务有限公司 | A kind of network electric business borrows or lends money methods of risk assessment and model |
| CN107016342A (en) * | 2017-03-06 | 2017-08-04 | 武汉拓扑图智能科技有限公司 | A kind of action identification method and system |
Non-Patent Citations (2)
| Title |
|---|
| 中国人民银行湘西州中心支行金融生态办: "对湘西州民间借贷风险监测与分析", 《金融经济（理论版）》 * |
| 张婷: "P2P信贷助农过程中农户信用风险评估研究", 《中国优秀硕士学位论文全文数据库 (经济与管理科学辑)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110619434A (en) * | 2019-09-17 | 2019-12-27 | 北谷电子有限公司上海分公司 | Method and device for predicting repayment probability, electronic equipment and storage medium |
| CN112927071A (en) * | 2021-04-21 | 2021-06-08 | 顶象科技有限公司 | Post-loan behavior feature processing method and device |
| CN115147203A (en) * | 2022-06-08 | 2022-10-04 | 南京金威诚融科技开发有限公司 | Financial risk intelligent analysis method based on big data |
| CN115147203B (en) * | 2022-06-08 | 2024-03-15 | 阿尔法时刻科技(深圳)有限公司 | Financial risk analysis method based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rangapuram et al. | Deep state space models for time series forecasting | |
| CN110751557B (en) | Abnormal fund transaction behavior analysis method and system based on sequence model | |
| CN108492194A (en) | Products Show method, apparatus and storage medium | |
| CN107766929B (en) | Model analysis method and device | |
| CN107358247B (en) | Method and device for determining lost user | |
| CN106408592A (en) | Target tracking method based on target template updating | |
| Salehinejad et al. | Customer shopping pattern prediction: A recurrent neural network approach | |
| CN104751363B (en) | Stock Forecasting of Middle And Long Period Trends method and system based on Bayes classifier | |
| CN111932269B (en) | Equipment information processing method and device | |
| WO2015187372A1 (en) | Digital event profile filters | |
| CN108229542A (en) | A kind of cycle debt-credit credit risk monitoring method based on Time-Series analysis technology | |
| CN109840595B (en) | Knowledge tracking method based on group learning behavior characteristics | |
| CN106709588B (en) | Prediction model construction method and device and real-time prediction method and device | |
| Rathnayaka et al. | Geometric Brownian motion with Ito's lemma approach to evaluate market fluctuations: A case study on Colombo Stock Exchange | |
| CN111325248A (en) | Method and system for reducing pre-loan business risk | |
| CN110288459A (en) | Loan prediction technique, device, equipment and storage medium | |
| CN116029395B (en) | Pedestrian flow early warning method and device for business area, electronic equipment and storage medium | |
| CN108268872B (en) | Robust nonnegative matrix factorization method based on incremental learning | |
| CN107729469A (en) | Usage mining method, apparatus, electronic equipment and computer-readable recording medium | |
| Zhang et al. | Additive requirement ratio prediction using trend distribution features for hydrometallurgical purification processes | |
| CN110909775A (en) | Data processing method and device and electronic equipment | |
| CN111160959A (en) | User click conversion estimation method and device | |
| CN109345373A (en) | Check and write off method for prewarning risk, device, electronic equipment and computer-readable medium | |
| CN109447129A (en) | A kind of multi-mode Emotion identification method, apparatus and computer readable storage medium | |
| CN112884569A (en) | Credit assessment model training method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180629 |