CN109345272A - One kind is based on the markovian shop credit risk forecast method of improvement - Google Patents
One kind is based on the markovian shop credit risk forecast method of improvement Download PDFInfo
- Publication number
- CN109345272A CN109345272A CN201811430043.8A CN201811430043A CN109345272A CN 109345272 A CN109345272 A CN 109345272A CN 201811430043 A CN201811430043 A CN 201811430043A CN 109345272 A CN109345272 A CN 109345272A
- Authority
- CN
- China
- Prior art keywords
- shop
- credit
- comment
- markovian
- value
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000006872 improvement Effects 0.000 title claims abstract description 7
- 230000008451 emotion Effects 0.000 claims abstract description 13
- 238000004458 analytical method Methods 0.000 claims abstract description 8
- 238000005516 engineering process Methods 0.000 claims abstract description 6
- 238000003058 natural language processing Methods 0.000 claims abstract description 5
- 238000011156 evaluation Methods 0.000 claims abstract description 4
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 239000002245 particle Substances 0.000 claims description 11
- 230000007704 transition Effects 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000013210 evaluation model Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000009472 formulation Methods 0.000 claims description 2
- 238000009415 formwork Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 claims description 2
- 239000000463 material Substances 0.000 claims description 2
- 239000000203 mixture Substances 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005201 scrubbing Methods 0.000 claims description 2
- 238000004140 cleaning Methods 0.000 claims 1
- 210000005036 nerve Anatomy 0.000 claims 1
- 238000013461 design Methods 0.000 abstract description 3
- 230000009193 crawling Effects 0.000 abstract 1
- 238000005457 optimization Methods 0.000 abstract 1
- 238000007781 pre-processing Methods 0.000 abstract 1
- 238000007726 management method Methods 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 5
- 239000000047 product Substances 0.000 description 5
- 238000012546 transfer Methods 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000002996 emotional effect Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012797 qualification Methods 0.000 description 2
- 238000012372 quality testing Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000012550 audit Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006748 scratching Methods 0.000 description 1
- 230000002393 scratching effect Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0609—Buyer or seller confidence or verification
Abstract
The invention proposes one kind based on the markovian shop credit risk forecast method of improvement, specific step is as follows: 1, shop corpus obtains: crawling the commodity specify information of shop details and shop in electric business platform using web crawlers technology and is saved in database；2, Chinese natural language processing: first pre-processing the comment text of shop commodity, then carries out sentiment analysis using Chinese natural language handling implement, obtains emotion score；3, credit risk forecast: it is merged according to multi-dimensional data, design shop credit comprehensive evaluation form, calculate the CREDIT SCORE in each shop, further, improved Markov chain model is constructed, completes to predict the credit value in shop according to this model, finally, obtaining the risk class in shop according to credit value score.This is just optimization e-commerce shop Supervision on Bio-safety, improves supervision quality and e-commerce security creates condition.
Description
Technical field:
The present invention relates to commercial quality management domains, more particularly to one kind is based on the markovian shop credit of improvement
Risk Forecast Method.
Background technique:
In recent years, it is that the e-commerce for carrying out exchange of goods and service is relied on to flourish with internet, greatly improves
The quality and efficiency of economical operation, change the mode of production and life of the mankind.2016, Global EC market scale was more than
25 trillion dollars become the bright spot and new growing point of world economy.The recent statistics of Department of Commerce of China show, China's net in 2017
Network retail sales reaches 7.18 trillion yuans, and wherein the e-tail volume of physical goods reaches 5.48 trillion yuans, and e-tailing is to consumption
Pulling function further enhance.But since electric business platform mass management mode and traditional forms of enterprises's Quality Management Mode exist
Significant difference, quality management experience can not be directlyed adopt directly by electric business quality monitoring under existing line.In recent years, quality inspection is total
The lifeline that office develops in a healthy way commercial quality as e-commerce is tried to explore the prudential supervision measure for containing innovation, is pushed
Establish the e-commerce commodity matter of " standard is led, Risk Monitoring, online selective examination, source traces, possession is investigated and prosecuted, credit management "
Amount supervision new mechanism.
It is reported, it is logical that the e-commerce product quality monitoring standardization effort that China is carrying out at present is related to basis
With fields such as, quality management, quality sincerity, quality monitoring, quality risk prevention and control, the construction of electric business quality standard system is orderly pushed away
Into.In by the end of December, 2017, and it is " electronics quotient respectively that State Administration of Quality Supervision, Inspection and Quarantine and National Standards Commission, which have also issued 3 electronic commerce context standards,
Business platform businessman enters auditing criterion " " e-commerce platform product information, which is shown, to be required " and " e-commerce quality management art
Language ", it will implement in succession in 2018.These standards enter e-commerce platform businessman and have formulated qualification requirement, to audit
Content, result feedback and Business Information update etc. are made stipulations, to the basic principle of product information displaying, displaying in the network platform
Content and mode propose and be distinctly claimed, allow the product information of Web realease closer to consumer entity experience, make to consume
The online shopping experience of person is truer, more feels at ease.
In development process, very important one is exactly credit for e-commerce.Virtual economy has faster flowing speed
Degree and higher risk will largely impact the trading system of entity if credit crisis occurs.Cause
This, it is necessary to sufficiently research electronic commerce credits risk.Nowadays, the country in terms of study electronic commerce credits prediction also in
Step section, usually by case study and qualitative description method connected applications, and unstructured relevant risk forecast model.Consumer
Be difficult to recognize on network whether shop credible and shop inside commercial quality whether as describing, therefore it is very easy
The commodity in these problems shop are bought, if point of public sentiment data and quality testing data to network Shanghai amount can be passed through
Analysis carries out the prediction of credit value risk to shop on net, so that it may effectively avoid being closed by shop using data mining technology
It closes, the various problems that commodity undercarriage causes.Therefore quality testing department is needed to take effective Supervision Measures to supervise enterprise and improve quotient
Quality is horizontal, while also to reinforce the management of e-commerce platform.
Summary of the invention:
In order to quickly and efficiently analyzed from magnanimity, the information in multi-source heterogeneous related shop its there are the problem of
The shop of credit, it is to tradition that the present invention provides one kind based on markovian shop credit risk forecast method is improved
A kind of supplement of e-commerce method for quality control.
The technical solution adopted by the present invention to solve the technical problems such as following the description:
One kind is based on the markovian shop credit risk forecast method of improvement, it is characterised in that: under this method includes
State step:
Step 1: corpus obtains: utilizing web crawlers software, formulation crawls rule, grabs shop relevant to specified shop
The comment text for spreading the dependent merchandise in information and shop, is saved in database with structured form；
Step 2: Chinese natural language processing: data scrubbing operation being carried out to the comment on commodity data of crawl first, then
Using Chinese natural language handling implement comment corpus is segmented for the first time respectively and part-of-speech tagging, keyword and modal particle
The pretreatment such as identification is to obtain the sentiment analysis result of structuring and be saved in database, further, by these keywords and
Modal particle is compared with keyword in dictionary and modal particle, and the true angle value of comment is calculated；
Step 3: credit risk forecast: proposition shop credit comprehensive evaluation form first, and it is each out based on the formwork calculation
The credit value in shop；Further, improved Markov chain model is constructed, is completed according to this model pre- to the credit value in shop
It surveys, finally, obtaining the risk class in shop according to credit value score.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 1
In, due to the opening of network and diversification, the discreteness of network comment, so that containing a large amount of in the comment text of crawl
" noise ", including comment in vain, comment spam and repeat to comment on, these comments all can cause great shadow to subsequent text analyzing
It rings, in order to solve " dirty " text, it is necessary to be pre-processed to comment text.There is short number of words, information according to electric business comment text
Big and clear main body feature can be set number of words threshold value and judge whether it is the invalid comment of pleonasm comment removal；For rubbish
Rubbish comment can decide whether to remove containing certain Chinese number of words；The literal similarity for finally calculating comment text, which removes, to be repeated
Comment.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 2
In, it is characterised in that: with arrange emotion word dictionary to inside database comment on commodity extract keyword and modal particle into
Row comparison, judges the gap of practical scoring commented on and provided, and given metric 3 indicates that comment is consistent with practical scoring, measures
Value 1 indicates that comment is not inconsistent with practical scoring, and the score of this month all comments is averaged to the true value for comment on commodity in this month.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, choose some key indexes in the store information grabbed, including seller's credit, buyer's credit, license information, after sale rate,
Validity is commented on, BP neural network training then is carried out to these indexs and obtains their weight, calculates obtaining for shop credit value
Point.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, the calculation formula of total credit value score in shop are as follows:
R=Rn+Sn
Wherein, RnIndicate the existing Credit Evaluation Model of C2C e-commerce, SnIndicate new Credit Evaluation Model.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, the total credit value in shop obtains in calculation formula, RnCalculation formula are as follows:
Rn=Rn-1+rn,…rn∈{-1,0,1}
Wherein, RnIndicate the current credit of user, Rn-1Indicate the recent credit rating of user, rnIndicate that user obtains n-th
Credit response scoring, rn∈{-1,0,1}。
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, the total credit value in shop obtains in calculation formula, SnCalculation formula are as follows:
Sn=α * S1+β*S2+γ*S3+δ*S4+ε*S5
Wherein, S1,S2,S3,S4,S5It is true to be expressed as seller's credit, buyer's credit, license information, rate, comment after sale
Degree；α, β, γ, δ, ε respectively indicate the corresponding weight of each factor.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, in the Markov prediction model of computed improved, state transition probability matrix are as follows:
Wherein, m is the status number by the first step point；For EiDesignated state E is transferred to through 1 stepjThe frequency occurred
Number；For state EiThe total frequency occurred.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, in the Markov prediction model of computed improved, risk transfer probability are as follows:
Wherein, pn1+pn2+…+pnk=1.
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, in the Markov prediction model of computed improved, if a certain variable original state EiInitial vector be P0, then its k walks state
Transition probability matrix is
Pk=(P1)k
Then the state vector after k step transfer is
Pk=P0×Pk=Po(P1)k
It is based on improving in markovian shop credit risk forecast method in above-mentioned one kind, in the step 3
In, in improved Markov prediction model, BP neural network model is initially set up, normalizes original data；Then with original
First three data in beginning data predicts fourth data, and predicted value is found out residual values compared with actual value；Then normalizing
Change and carries out state demarcation；Further find out state transition probability.Finally with markoff process amendment BP neural network prediction
As a result, calculating the credit value of shop next month
Beneficial effects of the present invention: magnanimity, multi-source heterogeneous is obtained from electric business platform website using web crawlers tool
Commodity use comment text inside shop details and shop, by shallow-layer, the Chinese text information processing technology of deep layer,
So that non-structured data become the data of structuring, shop credit overall merit is obtained, and construct the pre- of shop credit value
Model is surveyed, and then risk profile is carried out to shop credit value.This has just helped the buyer of electronic goods and monitoring department to determine
Hotel owner's credit in shop determines the transaction risk of the commodity of the following purchase, and effectively avoid being caused by shop credit various asks
Topic, it is established that Risk-prove infrastructure improves the safety of e-commerce transaction.
Detailed description of the invention:
Fig. 1 is that overall flow figure of the invention is also Figure of abstract of the invention.
Fig. 2 is overall technology route map of the invention.
Fig. 3 is preconditioning technique route map of the invention.
Fig. 4 is its design drawing for constructing shop CREDIT SCORE.
Fig. 5 is building BP neural network broad flow diagram.
Fig. 6 is that it constructs markoff process broad flow diagram.
Fig. 7 is its shop credit risk grade classification design drawing.
Specific embodiment:
Below with reference to specific attached drawing, the present invention is further illustrated.
The present invention is to carry out information scratching to large-scale electric business platform by web crawlers tool, obtains magnanimity, multi-source heterogeneous
Chinese network store information and commodity user comment text, and Chinese natural language processing is carried out to it, construction one is complete
Whole shop credit risk forecast model.
One kind is based on the markovian shop credit risk forecast method of improvement, including corpus obtains, Chinese is natural
Language Processing and credit risk forecast these three steps, as shown in Figure 1.These three steps are described in detail respectively below.
Step 1, corpus obtains: utilizing web crawlers tool, it is relevant that specified shop is acquired from large-scale electric business platform
The comment text of the dependent merchandise in store information and shop, and be saved in local data base, then to the comment information of preservation into
Row pretreatment, reduces the noise in data, obtains true, reliable, non-structured comment corpus.
Comment on the following process of process that corpus obtains.The rule that crawls of web crawlers tool is formulated, crawl shop is detailed
The comment information two stage content of information and commodity stores the data of crawl into local data base, and a part is the warp in shop
Statistical information is sought, another part is that the comment information of commodity becomes original comment text；Data are carried out to its original comment text
Pretreatment generates comment corpus, is also stored into database.
Wherein, due to diversification, the discreteness of the opening of network and network comment, so that grabbed from electric business platform
Contain a large amount of noises in network comment text, if directly carrying out text mining to it, acquired results may with it is practical generate compared with
Large deviation.So meeting actual as a result, need to original comment set is filtered and be cleaned to obtain, noise be reduced.In advance
Processing includes deleting blank, useless comment, deletes punctuation mark extra in comment, deletes the word of redundancy in comment, deletes
Comment, modification wrong word except number of words less than 4 words, simplified Chinese character replace the complex form of Chinese characters, the comment for deleting redundancy etc..
Step 2, Chinese natural language is handled: being carried out respectively for the first time using Chinese natural language handling implement to comment corpus
The pretreatment such as identification of participle and part-of-speech tagging, keyword and modal particle is to obtain the sentiment analysis result of structuring and be saved in
In database, further, the emotion score of this comment is calculated, as shown in Figure 3.
2.1) participle and part-of-speech tagging
Client feedback is the unstructured of textual form for the purpose of the comment on electric business platform is to exchange and share
Natural language to therefrom excavate valuable information then needs that it is converted into structural data by participle technique.To commenting
The tool that The Analects of Confucius material carries out participle use is ICTCLAS, and the tool of part-of-speech tagging use is carried out to the comment corpus after participle
It is ICTCLAS, in order to improve the precision ratio of product features extraction, the part-of-speech tagging method of selection is to mark out more specific situation
Second level mark.
2.2) sentiment analysis
By analyzing the Chinese network comment text of the homologous isomery of magnanimity it is found that the comment of user feedback is user to purchase
Commodity in-service evaluation, the viewpoint of oneself is usually expressed with adjective, noun or verb.Sentiment dictionary is the collection of emotion word
It closes.In broad terms, refer to comprising the tendentious phrase of emotion or sentence；In the narrow sense, refer to comprising passionate tendentious
Set of words.Sentiment dictionary generally comprises two parts, a positive emotional word dictionary and a negative emotion word dictionary.Emotion
Dictionary is the basic resource of text emotion analysis.Emotion is divided into 7 basic class and 21 small by Chinese emotion vocabulary ontology library
Class is respectively labeled word emotional category and intensity, calculates the emotion score of every comment.A such as comment text
For " well last year just bought carried out safety this year not only but also it is good I like ", extract keyword and modal particle be " safety well
It is good " the emotion score of this comment is calculated.
Step 3, quality risk is evaluated: being utilized P2P technology, is assessed the credit index of businessman, in combination with having adopted
The details (such as seller's credit, qualification authentication) in the shop collected propose the numeralization model R for being directed to business trust；Its
Middle data are divided into dynamic credit evaluating system index and static credit scoring model.Dynamic indicator includes that the positive rating of businessman, businessman sell
Afterwards rate, buyer's credit, commodity degree of being consistent, seller's attitude, logistics service quality, favorable comment number, in comment number, difference comment number and comment
By degree of being consistent.Static State Index includes the commodity price of businessman and the license information of businessman.According to the data of extraction and related text
Investigation is offered, the data such as Fig. 4 that may influence business trust is established.
The present invention, which comprehensively considers, assigns above-mentioned each factor of key index different weights, here using BP neural network to it
It is trained, starts the weight of setting data item by hand, then weight is trained such as Fig. 5 using BP neural network.Weight S
Calculating have very big practical significance to the foundation of shop credit risk forecast model.
The calculation formula of related significance coefficient are as follows:
X=Wjk
The calculation formula of the index of correlation are as follows:
Rij=| (1-e-x)/(1+e-x)|
Y=rij
The calculation formula of absolute effect coefficient are as follows:
I is neural network input unit, i=1 ... m；J is neural network output unit, j=1 ... n；K is neural network
Implicit unit, k=1 ... P, WkiFor the weight coefficient between input layer i and hidden layer neuron k；WjkFor output layer
Weight coefficient between neuron j and hidden layer neuron k.
The prediction model such as Fig. 6 Markov is finally constructed, according to obtained shop credit value, with adjacent first trimester
Data are the credit value that training sample predicts 4th month, are divided by the risk class for establishing as shown in Figure 7.
The calculation formula of total credit value score in shop are as follows:
R=Rn+Sn
Wherein, RnIndicate the existing Credit Evaluation Model of C2C e-commerce, SnIndicate new Credit Evaluation Model.
RnCalculation formula are as follows:
Rn=Rn-1+rn,…rn∈{-1,0,1}
Wherein, RnIndicate the current credit of user, Rn-1Indicate the recent credit rating of user, rnIndicate that user obtains n-th
Credit response scoring, rn∈{-1,0,1}。
SnCalculation formula are as follows:
Sn=α * S1+β*S2+γ*S3+δ*S4+ε*S5
Wherein, S1,S2,S3,S4,S5It is true to be expressed as seller's credit, buyer's credit, license information, rate, comment after sale
Degree；α, β, γ, δ, ε respectively indicate the corresponding weight of each factor.
The building of BP neural network-Markov prediction model is carried out after obtaining shop credit value in above-mentioned steps, is built
Retail shop's credit value obtained above is mainly input in BP neural network prediction model by vertical step, calculates predicted value and reality
The residual error of actual value；After several states will be divided into respect to residual values using level status partitioning, according to Markov property to pre-
Measured value is modified.Specific steps are for example following: (1) establishing BP neural network model, normalize raw sample data；(2) with former
The data that preceding 3 months data in beginning sample data predict 4th month, and so on, after obtaining recombination data, it is carried out
Training, verifying and prediction；(3) by predicted value compared with actual value, opposite residual values is found out, carry out state demarcation after normalization；
(4) frequency is shifted according to 1 step state and finds out 1 step state transition probability matrix；(5) transfer of k step state is found out generally according to C-K equation
Rate matrix；(6) BP neural network prediction result is corrected with markoff process
Shop credit can represent the whole shop operation situation of businessman, can intuitively reflect the consumption of e-commerce
Environment, constructing this improved Markov prediction model has important role to shop credit risk.Markov Chain can be with
Reflect the transition probability of risk status:
Wherein, pn1+pn2+…+pnk=1.
State transition probability matrix are as follows:
Wherein, m is the status number by the first step point；For EiDesignated state E is transferred to through 1 stepjThe frequency occurred
Number；For state EiThe total frequency occurred.
In improved Markov prediction model, if a certain variable original state EiInitial vector be P0, then its k is walked
State transition probability matrix are as follows:
Pk=(P1)k
Then the state vector after k step transfer is
Pk=P0×Pk=Po(P1)k
The present invention can be grabbed using web crawlers tool on large-scale electric business platform with specified shop associated detailed information and
The user comment text of commodity inside shop, and a series of processing are carried out to it, establish the credit rationally with operable shop
The prediction model of value, and then risk profile is carried out to shop credit value.Using method of the invention, e-commerce can be optimized
Shop Supervision on Bio-safety improves supervision quality, can before consumption be preferably minimized risk, provide for e-commerce security
Sound assurance.
Claims (10)
1. one kind is based on the markovian shop credit risk forecast method of improvement, it is characterised in that:
Step 1: shop corpus obtains
Using web crawlers software, formulation crawls rule, grabs the related production in store information relevant to specified shop and shop
The comment text of product, is saved in database with structured form；
Step 2: Chinese natural language processing
Data scrubbing operation is carried out to the comment on commodity data of crawl first, then using Chinese natural language handling implement to commenting
The Analects of Confucius material is segmented for the first time respectively and the pretreatment such as the identification of part-of-speech tagging, keyword and modal particle is to obtain the feelings of structuring
Sense analysis result is simultaneously saved in database, further, by keyword and modal particle in these keywords and modal particle and dictionary
It compares, the true angle value of comment is calculated；
Step 3: credit risk forecast
Shop credit comprehensive evaluation form is proposed first, and goes out the credit value in each shop based on the formwork calculation；Further, it constructs
Improved Markov chain model out is completed to predict the credit value in shop, finally, must be got according to credit value according to this model
To the risk class in shop.
2. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 1, crawler technology is that the regular expression made is used to acquire electric business by the http protocol in webpage
The comment information in all comment on commodity areas in the details in certain shop of website and shop.
3. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 2, (1) is mainly repeated to comment text data default value, text to the cleaning of text data and comment number of words
The pretreatment of limitation；(2) to the sentiment analysis of the text data of the keyword of extraction and modal particle.
4. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: it is compared, is judged with keyword and modal particle of the emotion word dictionary arranged to the comment on commodity extraction inside database
The gap for the practical scoring that buyer comments on and provides, given metric 3 indicate that comment is consistent with practical scoring, and metric 1 indicates
Comment is not inconsistent with practical scoring, and the score of this month all comments is averaged to the true value for comment on commodity in this month.
5. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: choose some key indexes in the store information grabbed, including seller's credit, buyer's credit, license information, after sale rate,
Validity is commented on, BP neural network training then is carried out to these indexs and obtains the weight of these factors, calculates each shop letter
With the score of value.
6. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 3, the calculation formula of total credit value score in shop are as follows:
R=Rn+Sn
Wherein, RnIndicate the existing Credit Evaluation Model of C2C e-commerce, SnIndicate new Credit Evaluation Model.
7. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 3, the total credit value in shop obtains in calculation formula, RnCalculation formula are as follows:
Rn=Rn-1+rn,…rn∈{-1,0,1}
Wherein, RnIndicate the current credit of user, Rn-1Indicate the recent credit rating of user, rnIndicate that user obtains n-th credit
Feedback score, rn∈{-1,0,1}。
8. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 3, the total credit value in shop obtains in calculation formula, SnCalculation formula are as follows:
Sn=α * S1+β*S2+γ*S3+δ*S4+ε*S5
Wherein, S1, S2, S3, S4, S5It is expressed as seller's credit, buyer's credit, license information, after sale rate, comment validity；
α, β, γ, δ, ε respectively indicate the corresponding weight of each factor.
9. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in the Markov prediction model of computed improved, state transition probability matrix are as follows:
Wherein, m is the status number by the first step point；For EiDesignated state E is transferred to through 1 stepjThe frequency occurred；For state EiThe total frequency occurred.
10. as described in claim 1 a kind of based on markovian shop credit risk forecast method is improved, feature exists
In: in step 3, in improved Markov prediction model, BP neural network model is initially set up, normalizes script number
According to；Then fourth data is predicted with the first three data in initial data, predicted value is found out into residual error compared with actual value
Value；Then normalization carries out state demarcation；Further find out state transition probability；Finally with markoff process amendment BP nerve
Neural network forecast as a result, calculating the prediction credit value in shop lower January.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811430043.8A CN109345272A (en) | 2018-11-28 | 2018-11-28 | One kind is based on the markovian shop credit risk forecast method of improvement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811430043.8A CN109345272A (en) | 2018-11-28 | 2018-11-28 | One kind is based on the markovian shop credit risk forecast method of improvement |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109345272A true CN109345272A (en) | 2019-02-15 |
Family
ID=65318362
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811430043.8A Pending CN109345272A (en) | 2018-11-28 | 2018-11-28 | One kind is based on the markovian shop credit risk forecast method of improvement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109345272A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110020942A (en) * | 2019-04-12 | 2019-07-16 | 中电科大数据研究院有限公司 | Method for early warning, device, equipment and the storage medium of credit risk |
| CN111461876A (en) * | 2020-05-07 | 2020-07-28 | 赵玉洁 | E-commerce credit system management system and method based on big data |
| CN111861507A (en) * | 2020-06-30 | 2020-10-30 | 成都数之联科技有限公司 | Identification method and system for analyzing risks of online catering stores in real time |
| CN113158082A (en) * | 2021-05-13 | 2021-07-23 | 聂佼颖 | Artificial intelligence-based media content reality degree analysis method |
| WO2021232856A1 (en) * | 2020-05-21 | 2021-11-25 | 中国标准化研究院 | Big data-based online sales commodity sampling and testing method |
| CN114693188A (en) * | 2022-05-31 | 2022-07-01 | 四川骏逸富顿科技有限公司 | Risk supervision method, system and equipment for drug retail industry |
-
2018
- 2018-11-28 CN CN201811430043.8A patent/CN109345272A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110020942A (en) * | 2019-04-12 | 2019-07-16 | 中电科大数据研究院有限公司 | Method for early warning, device, equipment and the storage medium of credit risk |
| CN111461876A (en) * | 2020-05-07 | 2020-07-28 | 赵玉洁 | E-commerce credit system management system and method based on big data |
| WO2021232856A1 (en) * | 2020-05-21 | 2021-11-25 | 中国标准化研究院 | Big data-based online sales commodity sampling and testing method |
| CN111861507A (en) * | 2020-06-30 | 2020-10-30 | 成都数之联科技有限公司 | Identification method and system for analyzing risks of online catering stores in real time |
| CN111861507B (en) * | 2020-06-30 | 2023-10-24 | 成都数之联科技股份有限公司 | Identification method and system for real-time analysis of risks of network restaurant shops |
| CN113158082A (en) * | 2021-05-13 | 2021-07-23 | 聂佼颖 | Artificial intelligence-based media content reality degree analysis method |
| CN114693188A (en) * | 2022-05-31 | 2022-07-01 | 四川骏逸富顿科技有限公司 | Risk supervision method, system and equipment for drug retail industry |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109345272A (en) | One kind is based on the markovian shop credit risk forecast method of improvement | |
| KR101741509B1 (en) | Device and method for analyzing corporate reputation by data mining of news, recording medium for performing the method | |
| CN107977798B (en) | Risk assessment method for quality of electronic commerce product | |
| Qiu et al. | Automatic spatiotemporal and semantic information extraction from unstructured geoscience reports using text mining techniques | |
| CN105512687A (en) | Emotion classification model training and textual emotion polarity analysis method and system | |
| CN108733748B (en) | Cross-border product quality risk fuzzy prediction method based on commodity comment public sentiment | |
| Shafin et al. | Product review sentiment analysis by using nlp and machine learning in bangla language | |
| CN108984775B (en) | Public opinion monitoring method and system based on commodity comments | |
| CN103886501B (en) | Post-loan risk early warning system based on semantic sentiment analysis | |
| CN111353044B (en) | Comment-based emotion analysis method and system | |
| CN107886240B (en) | Rule-based cross-border e-commerce commodity quality risk identification method | |
| CN104899229A (en) | Swarm intelligence based behavior clustering system | |
| Banić et al. | Using big data and sentiment analysis in product evaluation | |
| CN114254201A (en) | Recommendation method for science and technology project review experts | |
| Anh et al. | Extracting customer reviews from online shopping and its perspective on product design | |
| Desai et al. | Business intelligence visualization using deep learning based sentiment analysis on amazon review data | |
| Zhang et al. | A hybrid neural network approach for fine-grained emotion classification and computing | |
| Liu et al. | Age inference using a hierarchical attention neural network | |
| Devasia et al. | Feature extracted sentiment analysis of customer product reviews | |
| CN112632377B (en) | Recommendation method based on user comment emotion analysis and matrix decomposition | |
| Rana et al. | A conceptual model for decision support systems using aspect based sentiment analysis | |
| Jagadeesan et al. | Twitter Sentiment Analysis with Machine Learning | |
| Xie et al. | The twenty-first century of structural engineering research: A topic modeling approach | |
| Maylawati et al. | Implicit aspect extraction in product reviews using FIN algorithm | |
| Wang et al. | Research on opinion spam detection by time series anomaly detection |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190215 |
|
| WD01 | Invention patent application deemed withdrawn after publication |