CN107590733A - Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks - Google Patents
Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks Download PDFInfo
- Publication number
- CN107590733A CN107590733A CN201710670225.1A CN201710670225A CN107590733A CN 107590733 A CN107590733 A CN 107590733A CN 201710670225 A CN201710670225 A CN 201710670225A CN 107590733 A CN107590733 A CN 107590733A
- Authority
- CN
- China
- Prior art keywords
- platform
- platforms
- social networks
- risk
- information
- 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
Abstract
The present invention relates to internet financial risks assessment technology, it is desirable to provide borrows platform methods of risk assessment based on the net of geographical economy and social networks.This kind borrows platform methods of risk assessment based on the net of geographical economy and social networks includes step：P2P platforms relevant information is extracted from the whole network, data are pre-processed, basic risk assessment index is obtained, obtains social networks risks and assumptions, obtains platform ultimate risk assessment index.The present invention can effectively assess the risk that net borrows platform, have the advantages that calculating is simple, calculating speed is fast, the degree of accuracy is high；The present invention has given up the methods of traditional artificial setting weight, proposes new thinking, and on the premise of risk profile effect is not influenceed, the result of P2P platform risk assessment is provided for the user of internet financial field.
Description
Technical field
The present invention is on internet financial risks assessment technology field, more particularly to based on geographical economy and social networks
Net borrow platform methods of risk assessment.
Background technology
For the P2P bloomed everywhere as emerging rapidly in large numbersBamboo shoots after a spring rain, both without threshold, government also lacks unified reference platform.Now
P2P just as tradition debt-credit industry bank, direct financing and indirect financing can also be divided into.In terms of direct financing, P2P platforms
Information announcing role is functioned only as, helps supply and demand of fund both sides to carry out more efficient matching, without regard to Cash Flow, is also not involved in
Guarantee；But in terms of indirect financing, P2P platforms act as the function of conventional financial intermediary of business bank, it is responsible for connecing from a side
Enter and provide substantive fund to credit requirement side, P2P just act as the role of fund transfer and risk intermediary in this case.
By platform take on be small amount lending mechanism role, or even with illegal fund collection only have a line every.And fund provider is now
Investment like being offered a loan to a lending mechanism, therefore would have to consider the creditworthiness and risk of P2P platforms
Grade.
The risk assessment of traditional financial, the ecological data being mainly based upon in the collage-credit data and banking system of Central Bank rely on
Manual examination and verification are completed.At home reference service it is far from enough improve in the case of, the real core of internet amount of money risk assessment
The heart is the big data that internet can be relied on to obtain, and such as BAT companies possess substantial amounts of user profile, and these data can be used
More comprehensively to predict the risk of petty load.And machine learning will be that big data epoch internet financial company structure is automatic
The sharp weapon of wind transmission control system.
One general risk evaluating system usually requires point following steps and carried out：(1) finger of wind direction grade is determined
Mark system, according to the characteristics of the attribute of evaluated object, evaluated object and purpose of appraisals goes to determine evaluation index system,
The determination of the specifying information, evaluation criteria of classification, evaluation index including evaluation index；(2) data are collected, according to be evaluated
Index specifying information go collect assess object related data；(3) weight is determined, index is assigned according to certain rule
Different weights；(4) assessment result is drawn, according to the calculation formula of index, finally draws each risk class for assessing object.
Financial risks assessment system is diversified, has plenty of and carries out risk assessment for personal information, has plenty of pin
Risk assessment is carried out to process of exchange, has plenty of and carries out risk assessment for financial platform.The information can be caused to cover face of finance is very
The length and breadth of land, therefore in huge information data, it is necessary for how excavating useful information.Data mining is a kind of new business
The industry information processing technology, it is mainly characterized by extracting a large number of services data in business database, changed, analyzing and it
He is handled modelling, therefrom extracts the critical information finance wind direction assessment system of auxiliary commerce decision-making.
Internet financial company and main strategies company are in the more commonly used framework of credit evaluation this respect at present
Regulation engine adds credit scoring card.Credit scoring card is mentioned, the most frequently used algorithm is exactly Logistic Regression, this
It is the algorithm to be looked upon as a magic weapon by bank card center or Financial Engineering aspect.Really, Logistic Regression are because of its letter
It is single, be easy to explain, develop and O＆M cost is relatively low and pursued.But the data dimension of the user obtained in internet compared with
It is more, it is in the majority with discrete or categorical attribute variable, and missing data is more, in this case, Logistic Regression's
Adaptability can be poor.And regulation engine and the separated pattern of credit scoring snap gauge type, sometimes because inside regulation engine certain
A little rules are too strong and refuse to fall many top-tier customers.In addition, in existing risk assessment algorithm, it is redundancy to have many information
, few people are handled these redundancies.It is typically artificial defined simultaneously for the weight of different information,
Thus there is subjectivity.
And it is most commonly used that sorting algorithm for risk assessment, people.And sorting algorithm has many kinds, wherein main someone
Artificial neural networks (Artificial Neural Networks, ANN) and SVMs (SVM, Support Vector
Machine).Artificial neural network is a kind of mathematics for the structure progress information processing that application is similar to cerebral nerve cynapse connection
Model.In this model, composition network is coupled to each other between substantial amounts of node (or " neuron ", or " unit "), i.e. " god
Through network ", to reach the purpose of processing information.Neutral net usually requires to be trained, and the process of training is exactly that network is carried out
The process of study.The value for the connection weight for changing network node is trained to make it have the function of classification, trained network is just
Identification available for object.At present, hundreds of existing different model of neutral net, common are BP networks, radial direction base RBF nets
Network, Hopfield networks, stochastic neural net (Boltzmann machines), Competitive ANN (Hamming networks, Self-organizing Maps
Network) etc..But current neutral net still generally existing convergence rate is slow, computationally intensive, the training time is long and can not explain
The shortcomings of.SVMs (SVM, Support Vector Machine) is one that Vapnik proposes according to Statistical Learning Theory
Kind new learning method, its maximum feature be according to empirical risk minimization, it is optimal to maximize class interval construction
Optimal Separating Hyperplane improves the generalization ability of learning machine, preferably solves the problems such as non-linear, high dimension, local minimum point.
For classification problem, sample of the algorithm of support vector machine in region calculates the decision-making curved surface in the region, thereby determines that the area
The classification of unknown sample in domain.SVM proposes for two classification problems, needs the other sample of two species as training sample.
The implementation most close with the present invention has following several, Chinese invention patent application：It is " a kind of to internet gold
Melt the risk evaluation model that net borrows platform " (application number：201510001103.4), " a kind of lending and borrowing business monitoring and pre-alarming method and
Device " (201610274038.7), " finance debt-credit risk control system and operation method " (201610272171.9), " car
Financial Risk Control system and method " (201610510673.0), " financial risk early-warning method based on multiple agent and
Its system " (201210494653.0).
Invention 1 (a kind of risk evaluation model that platform is borrowed to internet Network and Finance Network) is a kind of to be used to calculate internet Network and Finance Network
The model and method of platform risk is borrowed, including：Mathematical modeling, data mining, data operation.Wherein mathematical modeling is golden by internet
Melt each index that net is borrowed involved by platform and carry out quantification treatment, risk can be calculated by the model of standardization, and lead to
Cross data mining and obtain industry basic data.But this model is not rejected for useless information.And it is when starting
The weight of each information is configured by way of artificial, subjectivity is too strong.
2 (a kind of lending and borrowing business monitoring and pre-alarming method and device) wherein methods of invention include：Obtain lending and borrowing business information, obtain
Take Monitoring Data and judge whether Monitoring Data meets early-warning conditions, sends warning information if meeting.Wherein, Monitoring Data is come
From third party's data system or this end system.The loan information that this method obtains both had not accounted for the residing of loan platform
The influence for the geography economy that environment is brought, does not account for the influence that the social networks of platform in itself are brought yet.
Invention 3 (finance debt-credit risk control system and operation method) includes a server, and the server passes through one
Data analysis control connects prime risk controller and two level risk controller；The server also passes through internet and China
The People's Bank's credit investigation system, main strategies train of mechanism, court system and public security system connection.This method does not have logarithm
According to being handled, cause data inaccurate.
Invent 4 (vehicle Financial Risk Control system and methods) and vehicle Financial Risk Control system is provided, including：It is vehicle-mounted
Terminal；Teleprocessing terminal；Teleprocessing terminal includes；Memory cell；Data analysis unit；Risk assessment unit；User believes
Use taxon；Decision package.But in actual applications, connected each other again between each user, there is strong relation
User, it is consistent in some behaviors, but this method does not consider this relation.
5 (financial risk early-warning method and its system based on multiple agent) this method of invention include information acquiring step：
Data obtaining module search obtains the enterprise external information relevant with finance；Analysis and processing step：Text mining module is to institute
State information and carry out text mining；Analysis and processing module is further analysed in depth to described information and knowledge；Reasoning module root
An optimal counte-rplan are chosen from described multiple counte-rplan and further assess according to the relevant knowledge in knowledge base
It is determined that.
Although five patents can carry out risk control to internet financial field to a certain extent above, they
Still the Shortcomings in following some problems：
1st, without the data for removing redundancy, the data of internet financial field are huge, but are not each item data
Final prediction result will be impacted.If not rejecting these data, we would not obtain accurate result, therefore
The data for removing redundancy are very necessary.
2nd, too subjectivity is set to the weight of each information, when analyzing each index, many people like
That thinks first goes setting weight, so does not only subjectivity, can also influence final evaluation result, therefore the power of artificial setting target
It is worthless again.
3rd, sorting algorithm is performed poor on existing data set, and for existing net borrows platform, we can only determine
Which runs away, but not can determine which must be normal operation.So the processing of most of grader can only rely on
In the information of problematic platform, but various sorting algorithms are not support such data processing in itself, so final
Evaluation result in have very big error.
4th, the influence of geographical economy is not accounted for, it is geographical because economic development is subjected to the influence of geographical environment
The economic development to internet financial industry has a great impact.Therefore in different geographical position, information identical P2P platforms
Different risks can be shown.Discounting for the influence of geographic factor, evaluation and test also has error.
5th, the influence of social networks is not accounted for, the risk for finance of networking often has very big relation with social activity, such as together
One shareholder borrows platform to different two nets and invested, and the risk of this two platforms is similar to a certain extent.Therefore
Introducing social networks can be significantly improved to evaluation result.
The content of the invention
It is a primary object of the present invention to overcome deficiency of the prior art, there is provided a kind of more flexible, perfect is directed to
The P2P platform methods of risk assessments of internet financial field.In order to solve the above technical problems, the solution of the present invention is：
There is provided it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, for entering to P2P platforms
Row risk assessment, it is described that following step is specifically included based on the net of geographical economy and social networks loan platform methods of risk assessment：
First, P2P platform relevant informations are extracted from the whole network；
The list of websites of (artificial) extraction P2P platforms from the whole network, data then are carried out to the webpage in these websites and climbed
Take and carry out data scrubbing (making it format as far as possible), extract industrial and commercial information, the corporate business letter of all P2P platforms
Breath, shareholder senior executive's information, target information, secured information, technical guarantee information；And P2P is put down according to running away and not running away
Platform is classified, and is divided into problem platform (running away) and normal platform (not running away)；
(artificial) selects news website from the whole network, is crawled and the P2P platform phases according to the platform names of each P2P platforms
The news of pass, LDA (Latent Dirichlet Allocation) topic model is carried out to every news of each P2P platforms
Study, obtains the theme of every news, to extract the news public feelings information of all P2P platforms；
2nd, data are pre-processed；
All P2P platforms are classified according to geographical position, are divided into the geographical module of regional geography module and the whole nation, region
Geographical module is divided into East China, West China, south China, North China, the geographical module in five, Central China；
Then frequent item set detection is carried out：With Aprior algorithms, (Apriori algorithm is a kind of the frequent of Mining Association Rules
Item set algorithm, its core concept are to detect two stages by the downward closing of candidate generation and plot come Mining Frequent item
Collection) to the industrial and commercial information of step 1 acquisition, corporate business information, shareholder senior executive's information, target information, secured information, skill
Art ensures that the particular content in information, news public feelings information carries out frequent episode detection, obtains testing result；
3rd, basic risk assessment index is obtained；
With One Class SVM algorithm, (in machine learning field, one-class support vector machines One Class SVM are one
The individual learning model for having supervision, commonly used to carry out pattern-recognition, classification and regression analysis) six ground being obtained to step 2
The data of reason module are modeled, and are respectively obtained each as training set with P2P platform informations the problem of being obtained in step 1
Risk assessment index of the P2P platforms in the geographical module of regional geography module and the whole nation；
The model of the risk assessment index of the geographical module in the whole nation is：
Wherein, the i refers to P2P platforms i；The w refers to regular terms；The F represents canonical item collection；The R is represented just
Real number；The t refers to any arithmetic number；The ρ refers to problem platform risk factor；The ω refers to all platform risk factors
Variance；The G refers to the testing result that the frequent episode obtained by step 2 detects；The ξ refers to the risk factor of platform；Institute
StateRefer to the standard deviation of platform i risk factors；
For P2P platform i, ξ is obtainedi1(ξi1With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the whole nation of i platforms
Risk index), 0≤ξi1≤ 1, ξi1Value it is bigger, illustrate that the risk of the P2P platforms under national environment is bigger；
The model phase of the risk assessment index of the geographical module of model and the whole nation of the risk assessment index of regional geography module
Together, the problem of training set is in region P2P platform informations；
For P2P platform i, ξ is obtainedi2(ξi2With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the region of i platforms
Risk index), 0≤ξi2≤ 1, ξi2Value it is bigger, illustrate the P2P platforms in East China, West China, south China, North China or Central China region
Bigger (the ξ of risk under environmenti1And ξi2It is different just illustrate, the geographical economic risk factor is existing)；
4th, social networks risks and assumptions are obtained；
For each P2P platforms, P2P platforms, shareholder and the social networks of senior executive are excavated, draw the social activity of the P2P platforms
The influence of relation pair platform risk, specifically includes following step：
Step 4a)：Draw social networks figure；
All shareholders, senior executive and P2P platforms are regarded as a little, if some shareholder invested some P2P platform,
Add a line between the shareholder and the P2P platforms, if some senior executive held a post in some P2P platform, in the senior executive and
A line is added between the P2P platforms, if friends be present between some shareholder and some senior executive, in the shareholder and is somebody's turn to do
A line is added between senior executive, obtains social networks figure；
Step 4b)：Determine social networks risks and assumptions；
If one P2P platform of setting can be associated with another P2P platform in social networks figure, within 3 steps, then recognize
It is related platform for the two P2P platforms (relevance of related platform is very big, and risk can be closely similar)；With depth-first
Method, find the platform related to problematic P2P platforms (the problem of classification obtains in step 1 platform)；
Run away provided with P2P platforms i, the P2P platforms number that it can be reached in figure by 3 steps is M, is put down in M P2P
In platform, the P2P platforms number for having run away or having gone wrong is N, then the social networks risk of P2P platforms i and M related platform
The factor is：
When a P2P platform possesses social risk by multiple mark, the maximum in its social networks risks and assumptions is taken；
5th, platform ultimate risk assessment index is obtained；
The social networks of P2P platforms are introduced in the range of regional geography module to be influenceed, the P2P platform areas introduced after social activity
Property risk assessment index in domain is：
ξi2×(1+Ci)；
Wherein, the symbol ξi2Defined with step 3 it is identical, Ci definition defined with step 4 it is identical；
Ultimate risk assessment index of the last COMPREHENSIVE CALCULATING P2P platforms under the influence of the whole nation and under regional environment：
λ×ξi1+(1-λ)ξi2×(1+Ci)；
Wherein, λ represents the influence (setting λ herein as 0.5) that national environment is controlled platform wind direction；The ξi1 ξi2Symbol
Defined with step 3 it is identical, Ci definition defined with step 4 it is identical.
In the present invention, P2P platform relevant informations acquired in the step 1：
The industrial and commercial information includes company incorporated's title, company's type, legal person, registered capital, capital subscribed, registered place
Location, hour of log-on, check and approve time, registration body, whether have annual report, industrial and commercial number, the tax number, mechanism number；
The corporate business information include platform names, on-line time, place city, place province, operational department's quantity, plus
The financial association that enters, affiliate, business business, the background of company of company；
Title of the shareholder senior executive's information including shareholder and senior executive, post, educational background, sex, graduation universities and colleges, working year
Limit, professional qualification certificate, management level experience；
The target information includes title, tender type, issue date, loaning bill Annual Percentage Rate, borrowing balance, the loaning bill of target
In the time limit, rise and throw the amount of money, bid quantity, overdue safeguard, correlative charges, use of the loan, mode of repayment, borrower's name, borrow
Money people's passport NO.；
The secured information includes whether automatic insurance application mode, debenture transfer mode, fund deposit pipe and deposit pipe public affairs
Department, fund whether trustship and hosting company, submit a tender ensure, Support Mode, the provision of risk, guarantee agency；
The technical guarantee information include technical support mechanism, website whether have safety and safety certification it is specific in
Hold, with the presence or absence of APP, whether HTTPS, log in when whether there is identifying code, user to forget Password when give the flow of password for change.
In the present invention, in the step 1, LDA (Latent are carried out to every news of each P2P platforms
Dirichlet Allocation) topic model learns to obtain the theme of every news, to extract the news of all P2P platforms
Public feelings information, specifically include following step：
Step 1a)：To each news, a theme is extracted from all themes of news distribution (crawling) of the whole network；
Step 1b)：A word is extracted from all words distribution corresponding to the above-mentioned theme being pumped to；
Step 1c)：Said process is repeated until traveling through each word in news；
The new probability formula of each word inside news：P (w | D)=∑ p (w | z) × p (z | D)；
The joint probability density of the theme distribution of news, theme vector and word vector：P (θ, z, w | α, β)；
To joint probability density in θ upper integrals, summed on z, obtain the marginal probability of news：P (w | α, β)；
Then to marginal probability quadrature, news probability is obtained：P (D | α, β)；
Finally, by training the maximum probability of α and β news, that is, the distribution of each word and theme on theme are obtained
Distribution in news；Distribution of the theme in every news：p(θ|α)；
Wherein, z is theme set, and w is set of words；The θ refers to single theme, and α and β are language material collection variable (the whole networks
It is unified)；The D is all set of words of each news.
In the present invention, in the step 2, the geographical module in the whole nation include all provinces in the whole nation, autonomous region, municipality directly under the Central Government,
Special administrative region；East China geography module includes Shanghai, Zhejiang, Anhui, Jiangsu；West China geography module includes Chongqing, Qinghai, sweet
Respectful, Ningxia, Shaanxi, Sichuan, Tibet, Xinjiang；South China geography module includes Guangdong, Guangxi, Fujian, Hainan, Taiwan；North China is geographical
Module includes Beijing, Tianjin, Hebei, Shanxi, Heilungkiang, Jilin, Liaoning, the Inner Mongol；Central China geography module includes Henan, lake
North, Hunan, Jiangxi.
In the present invention, the frequent item set detection in the step 2, specifically includes following step：
Step 2a)：List all information of each P2P platforms respectively in six geographical modules, and be each geographical mould
Block sets support min_conf；
Step 2b)：Training data is done with the information of problematic P2P platforms, finds out " frequent 1 item collection ", the set is denoted as
L1, L1 are used for the set L2 for looking for " frequent 2 item collection ", and L2 is used to look for L3；So on, until that can not find " frequent K item collections "；
Wherein " frequent J item collections " refers to the set of J at most shared identical information of P2P platforms of all the problems；
Wherein, the K refers to natural number；The J refers to natural number；
Step 2c)：Iterative process：Frequent k-1 item collections generate 2 subsets, and the son of 2 generations referred to here is concentrated with two
Individual k-1 item collections；If there is two k-1 item collections, each item collection is ranked up according to the lexicographic order of " attribute-value " (typically according to value)；
If the preceding k-2 item of two k-1 item collections is identical, and last difference, then it is attachable, i.e. this k- to prove them
1 item collection can be with marriage, you can connection generation k item collections；If but the subset of an item collection is not frequent item set, the item collection is not yet
It is frequent item set, removes the item collection；
Wherein, the k refers to natural number；
Step 2c)：Iteration finds the K item collections of maximum, and this K item is the useful item of risk assessment.
In the present invention, the step 4a of the step 4) in, social networks figure takes the mode of poll to build figure, i.e., in turn
State, shareholder and the senior executive of P2P platforms are scanned, for the information that can be operated to social networks figure, is entered into society successively
Hand on graph of a relation.
In the present invention, the step 4b of the step 4) in, using the method for depth-first, come it is fast find with it is problematic
P2P platforms related platforms comprise the following steps that：
Step i：Access the vertex v of social networks figure；
Step ii：Successively from v not accessed abutment points, depth-first traversal is carried out to social networks figure；Directly
Into social networks figure and v has that the summit that path communicates is all accessed, or the step-length of traversal has equalized to 3；
Step iii：If now still having summit not to be accessed in social networks figure, the summit not being accessed from one goes out
Hair, re-starts depth-first traversal, untill all summits are accessed in social networks figure；
In this way, we can obtain the social networks between the platform.
Compared with prior art, the beneficial effects of the invention are as follows：
Of the invention first extraction net borrows related website from the whole network, then by information classification in these websites, therefrom extracts
Go out effective information, finally carry out analysis and evaluation.
The present invention can effectively assess the risk that net borrows platform, have and calculate simple, calculating speed is fast, the degree of accuracy is high etc.
Advantage.
The present invention has given up the methods of traditional artificial setting weight, proposes new thinking, is not influenceing risk profile effect
On the premise of fruit, the result of P2P platform risk assessment is provided for the user of internet financial field.
Compared to similar appraisal procedure, algorithm of the invention can propose redundancy, and consider ground
The influence of economy and social networks is managed, there is high accuracy.
Brief description of the drawings
Fig. 1 is inventive algorithm overall flow figure.
Fig. 2 is acquisition effective information flow chart.
Fig. 3 is acquisition final assessment result flow chart.
Embodiment
The present invention is described in further detail with embodiment below in conjunction with the accompanying drawings：
As shown in Figure 1 borrows platform methods of risk assessment based on the net of geographical economy and social networks, first to the number of platform
Divided according to being obtained, then by geographical position, then data are carried out with cleaning and ensures validity, utilizes the side of machine learning
Method carries out basic prediction to platform, then introduces social networks, finally calculates the risk assessment index of P2P platforms, specific bag
Include following step：
First, P2P platform relevant informations are extracted from the whole network
(1) website of P2P platforms is manually extracted from the whole network, then carrying out data to the webpage in these websites crawls simultaneously
Data scrubbing is carried out, it is formatted as far as possible, extracts the industrial and commercial information, corporate business information, shareholder of all P2P platforms
Senior executive's information, target information, secured information, technical guarantee information.
Industrial and commercial information includes company incorporated's title, company's type, legal person, registered capital, capital subscribed, registered address, note
The volume time, check and approve time, registration body, whether have annual report, industrial and commercial number, the tax number, mechanism number.
Corporate business information include platform names, on-line time, place city, place province, operational department's quantity, add
Financial association, affiliate, business business, the background of company of company.
Title of shareholder senior executive's information including shareholder and senior executive, post, educational background, sex, graduation universities and colleges, length of service, duty
Industry credentials, management level experience.
Title of the target information including target, tender type, issue date, loaning bill Annual Percentage Rate, borrowing balance, loaning bill phase
Limit, rise and throw the amount of money, bid quantity, overdue safeguard, correlative charges, use of the loan, mode of repayment, borrower's name, borrow money
People's passport NO..
Secured information includes whether automatic insurance application mode, debenture transfer mode, fund deposit pipe and Cun Guan companies, money
Gold whether trustship and hosting company, submit a tender ensure, Support Mode, the provision of risk, guarantee agency.
Technical guarantee information includes technical support mechanism, whether website has the particular content of safety and safety certification, is
It is no exist APP, whether HTTPS, give the flow of password when logging in when whether thering is identifying code, user to forget Password for change.
(2) news website is manually selected from the whole network, is crawled and the P2P platforms according to the platform names of each P2P platforms
Related news, LDA (Latent Dirichlet Allocation) theme mould is carried out to every news of each P2P platforms
Type learns, and obtains the theme of every news, to extract the news public feelings information of all P2P platforms, specifically includes following step：
Step 1a)：To each news, one theme of middle extraction is crawled from all themes of news distribution of the whole network；
Step 1b)：A word is extracted from all words distribution corresponding to the above-mentioned theme being pumped to；
Step 1c)：Said process is repeated until traveling through each word in news；
The new probability formula of each word inside news：P (w | D)=∑ p (w | z) × p (z | D)；
The joint probability density of the theme distribution of news, theme vector and word vector：P (θ, z, w | α, β)；
To joint probability density in θ upper integrals, summed on z, obtain the marginal probability of news：P (w | α, β)；
Then to marginal probability quadrature, news probability is obtained：P (D | α, β)；
Finally, by training the maximum probability of α and β documents, that is, the distribution of each word and theme on theme are obtained
Distribution of the distribution theme in every news in news：p(θ|α)；
Wherein, z is theme set, and w is set of words；The θ refers to single theme；α and β is language material collection variable, the whole network
It is unified.
2nd, data are pre-processed
(1) all P2P platforms are classified according to geographical position, are divided into the geographical module of regional geography module and the whole nation,
Regional geography module is divided into East China, West China, south China, North China, the geographical module in five, Central China.
The geographical module in the whole nation includes all provinces in the whole nation, autonomous region, municipality directly under the Central Government, special administrative region.
East China geography module includes Shanghai, Zhejiang, Anhui, Jiangsu.
West China geography module includes Chongqing, Qinghai, Gansu, Ningxia, Shaanxi, Sichuan, Tibet, Xinjiang.
South China geography module includes Guangdong, Guangxi, Fujian, Hainan, Taiwan.
North China geography module includes Beijing, Tianjin, Hebei, Shanxi, Heilungkiang, Jilin, Liaoning, the Inner Mongol.
Central China geography module includes Henan, Hubei, Hunan, Jiangxi.
(2) frequent item set detection is carried out：With Aprior algorithms, (Apriori algorithm is a kind of the frequent of Mining Association Rules
Item set algorithm, its core concept are to detect two stages by the downward closing of candidate generation and plot come Mining Frequent item
Collection) to the industrial and commercial information of step 1 acquisition, corporate business information, shareholder senior executive's information, target information, secured information, skill
Art ensures that the particular content in information, news public feelings information carries out frequent episode detection, obtains testing result.
Comprising the following steps that for frequent item set detection is described：
Step 2a)：List all information of each P2P platforms respectively in six geographical modules, and be each geographical mould
Block sets support min_conf；
Step 2b)：Training data is done with the information of problematic P2P platforms, finds out " frequent 1 item collection ", the set is denoted as
L1, L1 are used for the set L2 for looking for " frequent 2 item collection ", and L2 is used to look for L3；So on, until that can not find " frequent K item collections "；
Wherein " frequent J item collections " refers to the set of J at most shared identical information of P2P platforms of all the problems；
Wherein, the K refers to natural number；The J refers to natural number；
Step 2c)：Iterative process：Frequent k-1 item collections generate 2 subsets, and the son of 2 generations referred to here is concentrated with two
Individual k-1 item collections；If there is two k-1 item collections, each item collection is ranked up according to the lexicographic order of " attribute-value " (typically according to value)；
If the preceding k-2 item of two k-1 item collections is identical, and last difference, then it is attachable, i.e. this k- to prove them
1 item collection can be with marriage, you can connection generation k item collections；If but the subset of an item collection is not frequent item set, the item collection is not yet
It is frequent item set, removes the item collection；
Wherein, the k refers to natural number；
Step 2c)：Iteration finds the K item collections of maximum, and this K item is the useful item of risk assessment.
3rd, basic risk assessment index is obtained
(1) with One Class SVM algorithm, (in machine learning field, one-class support vector machines One Class SVM are
One learning model for having supervision, commonly used to carry out pattern-recognition, classification and regression analysis) step 2 is obtained six
The data of geographical module are modeled, and are respectively obtained each as training set with P2P platform informations the problem of being obtained by one
Risk assessment index of the P2P platforms in the geographical module of regional geography module and the whole nation.
Because our training samples only have the problematic platform data collection in the whole nation, training one is needed for training sample
This compact classification boundaries.We select One Class SVM as basic model herein.Input data is：Passed through in national
Cross the platform information of data cleansing.Training data is the problematic platform data in the whole nation.Thus we model in N-dimensional space,
Wherein N refers to the number of effective information.
The model of the risk assessment index of the geographical module in the whole nation is：
Wherein, the i refers to P2P platforms i；The w refers to regular terms；The F represents canonical item collection；The R is represented just
Real number；The t refers to any arithmetic number；The ρ refers to problem platform risk factor；The ω refers to all platform risk factors
Variance；The G refers to the testing result that the frequent episode obtained by step 2 detects；The ξ refers to the risk factor of platform；Institute
StateRefer to the standard deviation of platform i risk factors；
For P2P platform i, ξ is obtainedi1, ξi1With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the whole nation of i platforms
Risk index, 0≤ξi1≤ 1, ξi1Value it is bigger, illustrate that the risk of the P2P platforms under national environment is bigger.
(2) the same problematic platform with the region is distinguished to existing platform in East China, West China, south China, North China, Central China
It is modeled.Input data is East China, West China, south China, North China, the respective platform information by data cleansing in Central China, is trained
Data are problematic platform data in region, and thus we model in Ni dimension spaces, and wherein Ni refers to effectively believe in region
The number of breath.
The model of the risk assessment index of regional geography module, model here is identical with the model in the whole nation, and training set is
The problem of in region P2P platform informations.
For P2P platform i, ξ is obtainedi2, ξi2With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the region of i platforms
Risk index, 0≤ξi2≤ 1, ξi2Value it is bigger, illustrate the P2P platforms in East China, West China, south China, North China or Central China region
Risk under environment is bigger.
ξi1And ξi2It is different just illustrate, the geographical economic risk factor is existing.
4th, social networks risks and assumptions are obtained
For each P2P platforms, P2P platforms, shareholder and the social networks of senior executive are excavated, draw the social activity of the P2P platforms
The influence of relation pair platform risk, specifically includes following step：
Step 4a)：Draw social networks figure；
All shareholders, senior executive and P2P platforms are regarded as a little, if some shareholder invested some P2P platform,
Add a line between the shareholder and the P2P platforms, if some senior executive held a post in some P2P platform, in the senior executive and
A line is added between the P2P platforms, if friends be present between some shareholder and some senior executive, in the shareholder and is somebody's turn to do
A line is added between senior executive, obtains social networks figure.
Social networks figure takes the mode of poll to build figure, i.e., scans state, shareholder and the senior executive of P2P platforms in turn, for
The information that can be operated to social networks figure, is entered on social networks figure successively.
Step 4b)：Determine social networks risks and assumptions；
If one P2P platform of setting can be associated with another P2P platform in social networks figure, within 3 steps, then recognize
It is related platform for the two P2P platforms, the relevance of related platform is very big, and risk can be closely similar；With depth-first
Method, the platform related to problematic P2P platforms is found, is comprised the following steps that：
Step i：Access the vertex v of social networks figure；
Step ii：Successively from v not accessed abutment points, depth-first traversal is carried out to social networks figure；Directly
Into social networks figure and v has that the summit that path communicates is all accessed, or the step-length of traversal has equalized to 3；
Step iii：If now still having summit not to be accessed in social networks figure, the summit not being accessed from one goes out
Hair, re-starts depth-first traversal, untill all summits are accessed in social networks figure；
In this way, we can obtain the social networks between the platform.
Run away provided with P2P platforms i, the P2P platforms number that it can be reached in figure by 3 steps is M, is put down in M P2P
In platform, the P2P platforms number for having run away or having gone wrong is N, then the social networks risk of P2P platforms i and M related platform
The factor is：
When a P2P platform possesses social risk by multiple mark, the maximum in its social networks risks and assumptions is taken.
5th, platform ultimate risk assessment index is obtained
(1) theoretically analyze, shareholder and senior executive are very sparse in trans-regional social activity, if calculated complete
In the range of state, the speed of calculating can be had a strong impact on, therefore ignores the social influence in gamut herein.In regional geography module
In the range of introduce the social networks of P2P platforms and influence, introduce it is social after P2P land regions risk assessment indexes be：
ξi2×(1+Ci)；
Wherein, the symbol ξi2Defined with step 3 it is identical, Ci definition defined with step 4 it is identical.
(2) ultimate risk assessment index of the last COMPREHENSIVE CALCULATING P2P platforms under the influence of the whole nation and under regional environment：
λ×ξi1+(1-λ)ξi2×(1+Ci)；
Wherein, λ represents the influence that national environment is controlled platform wind direction, and it is 0.5 to set λ herein；The ξi1 ξi2Symbol
Defined with step 3 it is identical, Ci definition defined with step 4 it is identical.
It is explained as follows below for term mentioned in the present invention：
1st, frequent item set (risk item collection)：
Frequent episode：In multiple set, the element entry that frequently occurs is exactly frequent episode.Frequent item set：There are a series of collection
Close, these gather some identical elements, and the high element of the frequency of occurrences forms a subset simultaneously in set, meets certain threshold value
Condition, it is exactly frequent item set.In P2P fields, a platform has industrial and commercial information, target information etc., is included again per category information
More seed items.Wherein, all subitems that can be impacted to platform risk are exactly the frequent episode in P2P fields, can be to platform risk
The set of all subitems impacted is exactly the frequent item set in P2P fields.In this invention, frequent episode and risk item implication phase
Together, frequent item set is identical with risk item collection implication.
2nd, abnormality detection：
Abnormality detection refers to be monitored abnormal data or system.In this invention, abnormality detection refers to, to net
Borrow platform whether the detection that can be run away.
3rd, it is geographical economical：
Geographical environment refers to the natural environment of social development, natural conditions, is the base that human society is depended on for existence with development
Plinth, and the source of economic development.Undoubtedly, economic development is subjected to the influence of geographical environment, is particularly led in resource
Into the traditional economy growth pattern of type, the situation of geographical environment largely determines the level of economic development of a state.Cause
This geographical economy just refers to the economic traits having because of geographic area.
4th, the geographical economic risk factor：
Because economic development is subjected to the influence of geographical environment, the geographical economic development to internet financial industry
Have a great impact.Therefore in different geographical position, information identical P2P platforms can show different risks, so ground
Reason is economical and weighs an important references element of P2P platform risks.In this invention, the geographical economic risk factor just refers to
The geographical economic factor that platform risk assessment is borrowed on net and is influenceed, we geographical position be divided into the whole nation, East China, West China, south China,
North China, six, Central China module.
5th, social networks risks and assumptions：
Social networks in this invention refer to, the investment and management that shareholder and senior executive are done in financial field.Internet gold
The risk melted often has very big relation with social activity, such as same shareholder borrows platform to different two nets and invested, and this two
The risk of family's platform is similar to a certain extent.In this invention, social networks risks and assumptions just refer to social networks pair
Net borrows the factor that platform risk assessment influences.
Finally it should be noted that listed above is only specific embodiment of the invention.It is clear that the invention is not restricted to
Above example, there can also be many variations.One of ordinary skill in the art can directly lead from present disclosure
All deformations for going out or associating, are considered as protection scope of the present invention.
Claims (7)
1. a kind of borrow platform methods of risk assessment based on the net of geographical economy and social networks, for carrying out risk to P2P platforms
Assess, it is characterised in that it is described based on the net of geographical economy and social networks borrow platform methods of risk assessment specifically include it is following
Step：
First, P2P platform relevant informations are extracted from the whole network；
The list of websites of P2P platforms is extracted from the whole network, then carrying out data to the webpage in these websites crawls line number of going forward side by side
According to cleaning, industrial and commercial information, corporate business information, shareholder senior executive's information, target information, the safety for extracting all P2P platforms are protected
Hinder information, technical guarantee information；And P2P platforms are classified according to running away and not running away, it is divided into problem platform and normally puts down
Platform；
News website is selected from the whole network, the news related to the P2P platforms is crawled according to the platform names of each P2P platforms,
LDA topic model study is carried out to every news of each P2P platforms, the theme of every news is obtained, to extract all P2P
The news public feelings information of platform；
2nd, data are pre-processed；
All P2P platforms are classified according to geographical position, are divided into the geographical module of regional geography module and the whole nation, regional geography
Module is divided into East China, West China, south China, North China, the geographical module in five, Central China；
Then frequent item set detection is carried out：The industrial and commercial information that is obtained with Aprior algorithms to step 1, corporate business information, shareholder
Particular content in senior executive's information, target information, secured information, technical guarantee information, news public feelings information carries out frequent
Item detection, obtains testing result；
3rd, basic risk assessment index is obtained；
The data of the six geographical modules obtained with One Class SVM algorithm to step 2 are modeled, and use step 1
In the problem of obtaining P2P platform informations as training set, it is geographical in regional geography module and the whole nation to respectively obtain each P2P platforms
The risk assessment index of module；
The model of the risk assessment index of the geographical module in the whole nation is：
<mrow>
<munder>
<mi>min</mi>
<mrow>
<mi>w</mi>
<mo>&Element;</mo>
<mi>F</mi>
<mo>,</mo>
<mi>&xi;</mi>
<mo>&Element;</mo>
<msup>
<mi>R</mi>
<mi>t</mi>
</msup>
<mo>,</mo>
<mi>&rho;</mi>
<mo>&Element;</mo>
<mi>R</mi>
</mrow>
</munder>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mo>|</mo>
<mo>|</mo>
<mi>&omega;</mi>
<mo>|</mo>
<msup>
<mo>|</mo>
<mn>2</mn>
</msup>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mi>G</mi>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mi>i</mi>
</munder>
<msub>
<mi>&xi;</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<mi>&rho;</mi>
</mrow>
Wherein, the i refers to P2P platforms i；The w refers to regular terms；The F represents canonical item collection；The R represents arithmetic number；
The t refers to any arithmetic number；The ρ refers to problem platform risk factor；The ω refers to the side of all platform risk factors
Difference；The G refers to the testing result that the frequent episode obtained by step 2 detects；The ξ refers to the risk factor of platform；It is describedRefer to the standard deviation of platform i risk factors；
For P2P platform i, ξ is obtainedi1, ξi1With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the national risk of i platforms
Index, 0≤ξi1≤ 1, ξi1Value it is bigger, illustrate that the risk of the P2P platforms under national environment is bigger；
The model of the risk assessment index of regional geography module is identical with the model of the risk assessment index of the geographical module in the whole nation, instruction
Practice the problem of integrating as in region P2P platform informations；
For P2P platform i, ξ is obtainedi2, ξi2With the ξ of above-mentioned formulaiIt is same variable, refers here particularly to the Regional Risk of i platforms
Index, 0≤ξi2≤ 1, ξi2Value it is bigger, illustrate the P2P platforms in East China, West China, south China, North China or Central China regional environment
Under risk it is bigger；
4th, social networks risks and assumptions are obtained；
For each P2P platforms, P2P platforms, shareholder and the social networks of senior executive are excavated, draw the social networks of the P2P platforms
Influence to platform risk, specifically includes following step：
Step 4a)：Draw social networks figure；
All shareholders, senior executive and P2P platforms are regarded as a little, if some shareholder invested some P2P platform, in the stock
A line is added between east and the P2P platforms, if some senior executive held a post in some P2P platform, in the senior executive and the P2P
A line is added between platform, if friends be present between some shareholder and some senior executive, in the shareholder and the senior executive
Between add a line, obtain social networks figure；
Step 4b)：Determine social networks risks and assumptions；
If one P2P platform of setting can be associated with another P2P platform in social networks figure, within 3 steps, then it is assumed that this
Two P2P platforms are related platform；With the method for depth-first, the platform related to problematic P2P platforms is found；
Run away provided with P2P platforms i, the P2P platforms number that it can be reached in figure by 3 steps is M, in M P2P platform
In, the P2P platforms number for having run away or having gone wrong is N, then the social networks risk of P2P platforms i and M related platform because
Son is：
<mrow>
<msub>
<mi>C</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfrac>
<mi>N</mi>
<mi>M</mi>
</mfrac>
<mo>;</mo>
</mrow>
When a P2P platform possesses social risk by multiple mark, the maximum in its social networks risks and assumptions is taken；
5th, platform ultimate risk assessment index is obtained；
The social networks of P2P platforms are introduced in the range of regional geography module to be influenceed, the P2P land regions introduced after social activity
Risk assessment index is：
ξi2×(1+Ci)；
Wherein, the symbol ξi2Defined with step 3 it is identical, Ci definition defined with step 4 it is identical；
Ultimate risk assessment index of the last COMPREHENSIVE CALCULATING P2P platforms under the influence of the whole nation and under regional environment：
λ×ξi1+(1-λ)ξi2×(1+Ci)；
Wherein, λ represents the influence that national environment is controlled platform wind direction；The ξi1 ξi2Rapid three definition of sign synchronization is identical, and Ci is fixed
Justice defines identical with step 4.
2. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, acquired P2P platform relevant informations in the step 1：
The industrial and commercial information includes company incorporated's title, company's type, legal person, registered capital, capital subscribed, registered address, note
The volume time, check and approve time, registration body, whether have annual report, industrial and commercial number, the tax number, mechanism number；
The corporate business information include platform names, on-line time, place city, place province, operational department's quantity, add
Financial association, affiliate, business business, the background of company of company；
Title of the shareholder senior executive's information including shareholder and senior executive, post, educational background, sex, graduation universities and colleges, length of service, duty
Industry credentials, management level experience；
Title of the target information including target, tender type, issue date, loaning bill Annual Percentage Rate, borrowing balance, loaning bill phase
Limit, rise and throw the amount of money, bid quantity, overdue safeguard, correlative charges, use of the loan, mode of repayment, borrower's name, borrow money
People's passport NO.；
The secured information includes whether automatic insurance application mode, debenture transfer mode, fund deposit pipe and Cun Guan companies, money
Gold whether trustship and hosting company, submit a tender ensure, Support Mode, the provision of risk, guarantee agency；
The technical guarantee information includes technical support mechanism, whether website has the particular content of safety and safety certification, is
It is no exist APP, whether HTTPS, give the flow of password when logging in when whether thering is identifying code, user to forget Password for change.
3. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, in the step 1, every news progress LDA topic models of each P2P platforms is learnt to obtain every news
Theme, to extract the news public feelings information of all P2P platforms, specifically include following step：
Step 1a)：To each news, a theme is extracted from all themes of news distribution of the whole network；
Step 1b)：A word is extracted from all words distribution corresponding to the above-mentioned theme being pumped to；
Step 1c)：Said process is repeated until traveling through each word in news；
The new probability formula of each word inside news：P (w | D)=∑ p (w | z) × p (z | D)；
The joint probability density of the theme distribution of news, theme vector and word vector：P (θ, z, w | α, β)；
To joint probability density in θ upper integrals, summed on z, obtain the marginal probability of news：P (w | α, β)；
Then to marginal probability quadrature, news probability is obtained：P (D | α, β)；
Finally, by training the maximum probability of α and β news, that is, the distribution of each word and theme are obtained on theme new
Distribution in news；Distribution of the theme in every news：p(θ|α)；
Wherein, z is theme set, and w is set of words；The θ refers to single theme, and α and β are language material collection variables；The D is every
All set of words of one news.
4. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, in the step 2, the geographical module in the whole nation includes all provinces in the whole nation, autonomous region, especially municipality directly under the Central Government, administration
Area；East China geography module includes Shanghai, Zhejiang, Anhui, Jiangsu；West China geography module includes Chongqing, Qinghai, Gansu, Ningxia, Shan
West, Sichuan, Tibet, Xinjiang；South China geography module includes Guangdong, Guangxi, Fujian, Hainan, Taiwan；North China geography module includes north
Capital, Tianjin, Hebei, Shanxi, Heilungkiang, Jilin, Liaoning, the Inner Mongol；Central China geography module includes Henan, Hubei, Hunan, river
West.
5. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, the frequent item set detection in the step 2, specifically includes following step：
Step 2a)：List all information of each P2P platforms respectively in six geographical modules, and set for each geographical module
Put support min_conf；
Step 2b)：Training data is done with the information of problematic P2P platforms, finds out " frequent 1 item collection ", the set is denoted as L1, L1
For looking for the set L2 of " frequent 2 item collection ", and L2 is used to look for L3；So on, until that can not find " frequent K item collections "；Wherein
" frequent J item collections " refers to the set of J at most shared identical information of P2P platforms of all the problems；
Wherein, the K refers to natural number；The J refers to natural number；
Step 2c)：Iterative process：Frequent k-1 item collections generate 2 subsets, and the son of 2 generations referred to here is concentrated with two k-
1 item collection；If there is two k-1 item collections, each item collection is ranked up according to the lexicographic order of " attribute-value "；If two k-1 item collections
Preceding k-2 item it is identical, and last difference, then prove they be it is attachable, i.e., this k-1 item collection can with marriage,
Generation k item collections can be connected；If but the subset of an item collection is not frequent item set, the item collection is gone nor frequent item set
Fall the item collection；
Wherein, the k refers to natural number；
Step 2c)：Iteration finds the K item collections of maximum, and this K item is the useful item of risk assessment.
6. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, the step 4a of the step 4) in, social networks figure takes the mode of poll to build figure, i.e., scans P2P platforms in turn
State, shareholder and senior executive, for the information that can be operated to social networks figure, be entered into successively on social networks figure.
7. it is according to claim 1 it is a kind of based on the net of geographical economy and social networks borrow platform methods of risk assessment, its
It is characterised by, the step 4b of the step 4) in, using the method for depth-first, carry out fast find and problematic P2P platforms phase
The platform of pass comprises the following steps that：
Step i：Access the vertex v of social networks figure；
Step ii：Successively from v not accessed abutment points, depth-first traversal is carried out to social networks figure；Until society
Hand over the summit that in graph of a relation and v has path to communicate all accessed, or the step-length of traversal has equalized to 3；
Step iii：If now still thering is summit not to be accessed in social networks figure, the summit not accessed from one, weight
It is new to carry out depth-first traversal, untill all summits are accessed in social networks figure；
In this way, we can obtain the social networks between the platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710670225.1A CN107590733A (en) | 2017-08-08 | 2017-08-08 | Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710670225.1A CN107590733A (en) | 2017-08-08 | 2017-08-08 | Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107590733A true CN107590733A (en) | 2018-01-16 |
Family
ID=61043131
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710670225.1A Pending CN107590733A (en) | 2017-08-08 | 2017-08-08 | Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107590733A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360089A (en) * | 2018-11-20 | 2019-02-19 | 四川大学 | Credit risk prediction technique and device |
| CN110400080A (en) * | 2019-07-26 | 2019-11-01 | 浙江大搜车软件技术有限公司 | Examination data monitoring method, device, computer equipment and storage medium |
| CN111768285A (en) * | 2019-04-01 | 2020-10-13 | 杭州金智塔科技有限公司 | Credit wind control model construction system and method, wind control system and storage medium |
| CN113449060A (en) * | 2021-06-29 | 2021-09-28 | 金陵科技学院 | Geographic big data security risk assessment method based on mixed gene expression programming |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080243719A1 (en) * | 2005-08-10 | 2008-10-02 | Axcessnet Innovations | Risk profiles in networked loan market and lending management system |
| CN102708096A (en) * | 2012-05-29 | 2012-10-03 | 代松 | Network intelligence public sentiment monitoring system based on semantics and work method thereof |
| CN103020256A (en) * | 2012-12-21 | 2013-04-03 | 电子科技大学 | Association rule mining method of large-scale data |
| CN103455705A (en) * | 2013-05-24 | 2013-12-18 | 中国科学院自动化研究所 | Analysis and prediction system for cooperative correlative tracking and global situation of network social events |
| CN105740245A (en) * | 2014-12-08 | 2016-07-06 | 北京邮电大学 | Frequent item set mining method |
| CN105809360A (en) * | 2016-03-16 | 2016-07-27 | 北京银讯财富信息技术有限公司 | P2P industry risk assessment method |
| CN105894369A (en) * | 2015-01-04 | 2016-08-24 | 必普(上海)金融信息服务有限公司 | Risk assessment model for internet financial network loan platform |
| CN105931115A (en) * | 2016-04-19 | 2016-09-07 | 上海闻玺投资咨询有限公司 | Network loan platform development index rating system and working method thereof |
-
2017
- 2017-08-08 CN CN201710670225.1A patent/CN107590733A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080243719A1 (en) * | 2005-08-10 | 2008-10-02 | Axcessnet Innovations | Risk profiles in networked loan market and lending management system |
| CN102708096A (en) * | 2012-05-29 | 2012-10-03 | 代松 | Network intelligence public sentiment monitoring system based on semantics and work method thereof |
| CN103020256A (en) * | 2012-12-21 | 2013-04-03 | 电子科技大学 | Association rule mining method of large-scale data |
| CN103455705A (en) * | 2013-05-24 | 2013-12-18 | 中国科学院自动化研究所 | Analysis and prediction system for cooperative correlative tracking and global situation of network social events |
| CN105740245A (en) * | 2014-12-08 | 2016-07-06 | 北京邮电大学 | Frequent item set mining method |
| CN105894369A (en) * | 2015-01-04 | 2016-08-24 | 必普(上海)金融信息服务有限公司 | Risk assessment model for internet financial network loan platform |
| CN105809360A (en) * | 2016-03-16 | 2016-07-27 | 北京银讯财富信息技术有限公司 | P2P industry risk assessment method |
| CN105931115A (en) * | 2016-04-19 | 2016-09-07 | 上海闻玺投资咨询有限公司 | Network loan platform development index rating system and working method thereof |
Non-Patent Citations (2)
| Title |
|---|
| 林春雨等: "基于大数据技术的P2P网贷平台风险预警模型", 《大数据》 * |
| 董小轶: "基于层次分析法的P2P网贷平台综合评估", 《时代金融》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109360089A (en) * | 2018-11-20 | 2019-02-19 | 四川大学 | Credit risk prediction technique and device |
| CN109360089B (en) * | 2018-11-20 | 2020-07-17 | 四川大学 | Loan risk prediction method and device |
| CN111768285A (en) * | 2019-04-01 | 2020-10-13 | 杭州金智塔科技有限公司 | Credit wind control model construction system and method, wind control system and storage medium |
| CN110400080A (en) * | 2019-07-26 | 2019-11-01 | 浙江大搜车软件技术有限公司 | Examination data monitoring method, device, computer equipment and storage medium |
| CN113449060A (en) * | 2021-06-29 | 2021-09-28 | 金陵科技学院 | Geographic big data security risk assessment method based on mixed gene expression programming |
| CN113449060B (en) * | 2021-06-29 | 2022-07-01 | 金陵科技学院 | Geographic big data security risk assessment method based on mixed gene expression programming |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | Design of deep belief networks for short-term prediction of drought index using data in the Huaihe river basin | |
| CN106779755A (en) | A kind of network electric business borrows or lends money methods of risk assessment and model | |
| CN107590733A (en) | Platform methods of risk assessment is borrowed based on the net of geographical economy and social networks | |
| CN112132233A (en) | Criminal personnel dangerous behavior prediction method and system based on effective influence factors | |
| CN107092929A (en) | Criminal offense case association string and method and system based on clustering technique | |
| CN104239485A (en) | Statistical machine learning-based internet hidden link detection method | |
| CN107341183A (en) | A kind of Website classification method based on darknet website comprehensive characteristics | |
| Agarwal et al. | Big data and predictive analytics in fire risk using weather data | |
| CN108492001A (en) | A method of being used for guaranteed loan network risk management | |
| CN104463601A (en) | Method for detecting users who score maliciously in online social media system | |
| CN106295332A (en) | Based on interval number and the Information Security Risk Assessment Methods of ideal solution | |
| CN109146611A (en) | A kind of electric business product quality credit index analysis method and system | |
| Mangal et al. | Fake news detection with integration of embedded text cues and image features | |
| CN107944761A (en) | Early warning and monitoring analysis method is complained based on artificial intelligence protection of consumers' rights index enterprise | |
| Arora et al. | PREDICTION & WARNING: a method to improve student's performance | |
| CN108564465A (en) | A kind of enterprise credit management method | |
| Kurniawan et al. | Decision support model using FIM Sugeno for assessing the academic performance | |
| CN106778252A (en) | Intrusion detection method based on rough set theory Yu WAODE algorithms | |
| TWM583089U (en) | Smart credit risk assessment system | |
| Zhang | Research of default risk of commercial bank's personal loan based on rough sets and neural network | |
| Wang et al. | Innovative risk early warning model based on internet of things under big data technology | |
| CN106326472B (en) | One kind investigation information integrity verification method | |
| Yasser Sahib Nassar | Building and analyzing a crisis management model using Fuzzy DEMATEL technique | |
| CN108629507A (en) | A kind of enterprise credit management system | |
| Kelana et al. | Challenge handling flood disaster by recommendation system with Analytical Hierarchy Process Methods |
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: 20180116 |