CN107194754A - Stock trader's Products Show method based on mixing collaborative filtering - Google Patents
Stock trader's Products Show method based on mixing collaborative filtering Download PDFInfo
- Publication number
- CN107194754A CN107194754A CN201710233443.9A CN201710233443A CN107194754A CN 107194754 A CN107194754 A CN 107194754A CN 201710233443 A CN201710233443 A CN 201710233443A CN 107194754 A CN107194754 A CN 107194754A
- Authority
- CN
- China
- Prior art keywords
- product
- user
- collaborative filtering
- stock trader
- products show
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses stock trader's Products Show method based on mixing collaborative filtering, including：The financial product information inputted during existing stock trader sells and some users purchase relevant financial product record and data, and form initial data base after sale；When promoting old product：Some user's purchase relevant financial product records in initial data base are extracted, the incidence relation between relevant financial product is analyzed with Association Rule Analysis method；And calculate the similar features between relevant financial product with collaborative filtering；The relevance height between the product obtained in Association Rule Analysis method and collaborative filtering and product one product set of composition more than similar features will be used；User is recommended after being screened to the product set.
Description
Technical field
The present invention relates to financial stock trader's Products Show field, and in particular to a kind of stock trader's product based on mixing collaborative filtering
Recommendation method.
Background technology
Application scenarios are widely in actual market for Products Show.(you can for such as Amazon store, Taobao, Jingdone district etc.
Can books interested, phonotape and videotape, clothes, electronic equipment ...) in the news site correlation subject matter such as dispensings of electric business article, today's tops
Hold information displaying.Enterprise can be strengthened Consumer's Experience, be realized more preferable cross-selling, improve the turnover by using Products Show
Degree.For a user, we all suffer from the choice of a large amount of events daily, before ineffective decision-making assistant information, from
The characteristic of something or other is understood in other users feedback information, the process of our choices can be accelerated.Such as：We are not to buying product
Before, choose whether to buy the product and Related product, the information that can be provided by enterprise product commending system is obtained
Well with reference to answer.
Inventor has found that stock trader have accumulated substantial amounts of user (user) to financial product during the present invention is realized
(item) buying behavior and product survey feedback record, but these records are failed to be sufficiently carried out quantitative excavation point
Analysis, hiding potential business characteristic therein and purchase rule are not found, also do not comb to form effective decision assistant according to next
The business sale of market department is supported to expand.
The Products Show difficult point such as other internet electric business, news, videos is contacted, stock trader's Products Show facing challenges are main
Have following：
Scale.Because it is difficult to obtain the real information of user satisfaction, the feasible proposed algorithm of the existing overwhelming majority is at place
Manage stock trader's Products Show just unable to do what one wishes.
Freshness.Stock trader's product is constantly weeded out the old and bring forth the new, and old product is constantly suspended sale of, and new product is omited on the basis of original
Plus change is just released again.Analysis modeling is carried out for the product newly promoted in time during Products Show, while to take into account existing
The balance of old product and new product.
Noise.Because user buys the openness and unobservable influence factor of stock trader's financial product behavior, user's
Historical record is substantially difficult to Accurate Prediction.
The content of the invention
In view of this, this application provides stock trader's Products Show method based on mixing collaborative filtering, existed with strengthening stock trader
Ability in terms of production marketing, expands the product market space and obtains higher operating profit.
To achieve the above object, the present invention provides following technical scheme：
Based on stock trader's Products Show method of mixing collaborative filtering, including：
The financial product information inputted during existing stock trader sells and some users purchase relevant financial product are recorded and sold
Data, and form initial data base afterwards；
When promoting old product：
Some user's purchase relevant financial product records in initial data base are extracted, are analyzed with Association Rule Analysis method
Incidence relation between relevant financial product；And calculate the similar spy between relevant financial product with collaborative filtering
Levy；
Will be high and similar with the relevance between the product obtained in Association Rule Analysis method and collaborative filtering
Product more than feature constitutes a product set；
User is recommended after being screened to the product set.
In a preferred embodiment of the invention, it is described sell in financial product information include fixed income class production
Product, cash class product, rights and interests class product.
In a preferred embodiment of the invention, with also being wrapped before Association Rule Analysis method and collaborative filtering
Include and input feature vector extraction is carried out to initial data base, the input feature vector includes but is not limited to following：Seek advice from number of times, purchase time
Number, Satisfaction Index, whether can buy like product.
In a preferred embodiment of the invention, carrying out screening to the product set includes：For each user, comb
Its product bought is managed, is product and recommendation in its product set for not buying of recommendation.
In a preferred embodiment of the invention, the popularization of new product is included：
The database of new product is set up, the database includes the financial product information and user's purchase correlation that user newly buys
Financial product record and after sale data；
Input feature vector formation analyze data is extracted according to new product information in the database of new product, and to the analyze data
Singular value SVD lowering dimension decompositions are carried out, selection reaches the preceding several elements and data of gross energy threshold percentage；And according to processing after
Data calculate similarity, and find out similar user group；
Similarity is directly calculated according to user profile, similar user group is produced；
Merge similar user group's composition user's set that above-mentioned new product information and user profile are obtained；
User is recommended after being screened to user set.
In a preferred embodiment of the invention, by the product set obtained when promoting old product and popularization new product
When obtain user set merge, formed candidate recommendation list；
According to the degree of accuracy and coverage rate index recommended, proposed algorithm is evaluated and perfect.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the accompanying drawing used required in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
The embodiment of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
The accompanying drawing of offer obtains other accompanying drawings.
Fig. 1 recommends a kind of of method to realize stream for the financial product based on mixing collaborative filtering that the embodiment of the present application is provided
Cheng Tu；
Fig. 2 is based on correlation rule and based on Item collaborative filtering recommendings for what the embodiment of the present application was provided to promoting old product
A kind of implementation process figure；
Fig. 3 is based on commending contents and based on user collaborative filtering recommendings for what the embodiment of the present application was provided to promoting new product
A kind of implementation process figure；
Fig. 4 is specific data analysis list in step S11；
Fig. 5 is the data form in step S21；
Fig. 6 is the rule that is found by association algorithm in step S22；
Fig. 7 is the data form in step S23；
Fig. 8 is the input feature vector table based on commending contents in step S31；
Fig. 9 is the tables of data that will be obtained after step S11 transposition in step S31.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.Based on this
Embodiment in invention, the every other reality that those of ordinary skill in the art are obtained under the premise of creative work is not made
Example is applied, the scope of protection of the invention is belonged to.
The present invention is by the arrangement to stock trader's production marketing data, with the collaborative filtering analysis method in machine learning,
The regularity sold between client characteristics and product and between different product classification has been sought, different business in production marketing are found
Contact between product, finds out the similitude and colony's buying behavior rule between different customer groups, is stock trader's finance product
Cross-selling, the determination of bundle sale strategy provide decision support.
By analyzing stock trader's history finance product sales situation, the regularity of three aspects is found：
One is, the similitude rule between product category；
Two are, the similitude rule between product group；
Three are, the rule of similarity between specific product.
Assuming that currently the consumption hobby of purchase user is similar with the consumption hobby of other certain users, and this hobby is most
Keep stable in nearly a period of time and following a period of time.
On the one hand, liked according to the consumption of active user, find and like other similar clients to it, analyze which they have
A little products are not yet bought for active user, by these Products Shows to active user.
Meanwhile, according to the distribution situation for currently having sold product, analysis is compared similar product with it, these products is pushed away
Recommend to non-buyer.
In general, in stock trader's production marketing, facilitate the first purchase of client, open up a new client, it is paid
Cost be that comparison is high.A frequent customer is kept, and makes them continue to buy more guarantees, relative cost is than relatively low.But
It is when two time selling is carried out to frequent customer, it is impossible to blindly carry out.If can be directed to according to the relevance between product
Property sell, or carry out product bundle sale, be possible to greatly improve the success rate of sale.
In addition, when using mixing collaborative filtering analysis stock trader's finance product sales situation, it must be noted that product
Some features：Invest time limit, risk level, the otherness of income；Meanwhile, stock trader's product is constantly weeded out the old and bring forth the new, old product
Constantly suspend sale of, new product is changed slightly on the basis of original just to be released again, even if the product in the same period sold,
Have is much to belong to of a sort product.
Referring to Fig. 1, Fig. 1 recommends method for the financial product provided in an embodiment of the present invention based on mixing collaborative filtering
A kind of implementation process figure, can include：
Step S11：Input stock trader fixed income class Item (product), cash class Item, rights and interests class Item and User purchase,
The data such as evaluation；
Describe this step in detail.In the embodiment of the present invention, considered based on stock trader's delineation of activities, the product category bag of delimitation
Include：Fixed income class, cash class, rights and interests class, the division of product category are applied to all stock traders；Product group and specific product
(the best preceding 10 sections of products of certain stock trader's sales volume) is selected according to stock trader's specific product situation.
Selected time interval be 2015.09.01 to 2016.03.31, selected data area be finance product hold position,
The data such as transaction journal, OTC product informations.
Based on customer quantity select certain securities broker company to buy 46292 clients and its data such as purchase, evaluation of product
It is used as analysis object.
Behavior user in reference picture 4, wherein matrix, row represent product.Element is evaluation of the user to product in matrix
Fraction, 0 represents not buying the article.
Step S12：To the old Item of popularization, according to " being based on correlation rule "+" collaborative filtering based on Item " shape
Into recommendation；
Describe this step in detail.In the embodiment of the present invention, it is contemplated that old Item data volume elapsed time is longer and customer
Purchaser record is than sparse, so being taken based on correlation rule：First by point between Item major class, group and specific product
Analysis scope is substantially locked.Simultaneously, it is considered to which the similitude between product is predicted recommendation.Such as know product a and c very phase
Seemingly；
Judge product a and c similarity, implementation has a lot, such as COS distance, Euclidean distance etc., herein according to
Tanimoto coefficients are calculated, and specific formula is shown in S23 because liking a user A while also liking c, when the user only buys
During a, c also recommends to the user A.
Step S13：To the new Item of popularization, according to " being based on commending contents "+" collaborative filtering based on user " shape
Into recommendation；
Describe this step in detail.In the embodiment of the present invention, it is contemplated that newly promotes product records less, therefore is primarily based on
The problem of commending contents are to solve cold start-up.It will recommend to find out search problem, by various labels come marked product, such as low wind
The attributes such as danger, short cycle, income stabilization.The data that these attributes are needed as Similarity Measure are predicted recommendation.Meanwhile,
Consider that user's similitude goes to promote these new products.Such as user A likes article a, user B to like article a, b, c, and user C likes
A, c, then it is assumed that user A, user B, user C are similar, so c is recommended into user A.
Step S14：Above two method is merged according to thinking in parallel, the recommendation list of candidate is formed；
Describe this step in detail.In the embodiment of the present invention, corresponding user can be found for new, old product.As above
State example, the matching relationship of product and user is, product c and user A matchings.
Step S15：According to the degree of accuracy and coverage rate index recommended, proposed algorithm is evaluated and perfect.
Describe this step in detail.In the embodiment of the present invention, all products and corresponding user are matched, and in reality
In verify, these users are to the response condition of recommended products, if responsiveness and coverage rate are very high, then it is assumed that this set of flow
Implementation process is reasonable.Otherwise, tuning is carried out to recommendation process.
If recommendation is improper, tuning is to need Optimized Iterative whole process, it is necessary to adjust each in whole whole process
The division of link, such as product type, product it is new/old define duration, user to the extension adjustment of the data input features of product,
Extension adjustment, the adjustment of Similarity Measure mode etc. of the data characteristic of product itself.
Such as above-mentioned example, the matching relationship of product and user are that product c and user A are matched, in practical business product marketing
Middle party A-subscriber have purchased product c really, then it is assumed that everything is reasonable science.
Present invention research finds the contact between different commodity in production marketing, finds out customer purchasing behavior pattern.Will be mixed
Close collaborative filtering method to be applied in stock trader's sales analysis by product, decision-making branch can be provided for the determination of stock trader's Product Sale Strategy
Hold.
Referring to Fig. 2, Fig. 2 is based on correlation rule and based on Item associations to be provided in an embodiment of the present invention to promoting old product
With a kind of implementation process figure of filtered recommendation, it can include：
Step S21：To the old Item of popularization, extracted and inputted according to the buying behaviors of related user on these Item etc.
Feature.
Reference picture 5, describes this step in detail.In the embodiment of the present invention, the method according to S22 steps is different, and input is deposited
In difference.For the input feature vector of correlation rule, for the format sample of following table, (T represents that client buys this category product, and F represents visitor
This category product is not bought in family)：
Input shown in form (accompanying drawing 4) is seen in S11 for the input feature vector based on Item collaborative filterings.
Step S22：The Item set of user preferences is combed out based on correlation rule；
Describe this step in detail.In the embodiment of the present invention, Data Mining buys the system of situation mainly for client to product
Meter analysis, grasps different classes of (product) data cases.The Data Mining of the present invention mainly includes following several respects：(1) it is single
The purchase situation statistics of classification (product)；(2) the purchase situation statistics of classification (product) combination.Counted by product information table,
Certain securities broker company has 397 sections of fixed income class product, 38 sections of rights and interests class product, 25 sections of cash class product.
The present invention, with Apriori Association Rule Analysis methods, is sought by the arrangement to stock trader's production marketing data
The regularity sold between client characteristics and product and between different product classification.The setting of confidence level and support, a side
Face, if confidence level sets too low, it is possible to can be flooded by a large amount of insecure rules；On the other hand, if by confidence level and
It is too high that support is set, and can be limited by apparent or inevitable rule.Solution in actual applications：
Constantly adjustment confidence level and support arrange parameter, eliminate unreliable rule, while providing rule of reason for targetedly marketing
Space.The rule found by association algorithm is as shown in Figure 6：
Pass through association rule mining, it was found that the incidence relation between some products, and these incidence relations are not for
The product marketing activity come has support and directive function.
Observe accompanying drawing 6 form, the main function of association be exactly analyze product category, product group, specific product it
Between close relation degree, according to strong product is associated, stock trader can targetedly be combined sale and bundle sale, improve
Loyalty and resident contribution of the client to product.
Step S23：Collaborative filtering based on Item combs similar Item set；
Describe this step in detail.In the embodiment of the present invention, the similarity between article, the method calculated here are calculated first
For the Similarity Measure of Tanimoto coefficients：
Wherein x, y represent different product respectively, and i is the input feature vector of the product；
Reference picture 7, citing describes above-mentioned formula implication in detail.If x, y represent in following table product list any 2 not respectively
Same product, such as x are that item1, y are item2, and i is the corresponding input feature vector of product, it is assumed here that only comprising being consulted number of times, quilt
Buy number of times, investigation satisfaction.Therefore the formula of the Similarity Measure according to Tanimoto coefficients,
That is x is item1 and y is that similitude between item2 is 0.64.
Step S24：Merging forms Candidate Recommendation item set；
Describe this step in detail.In the embodiment of the present invention, by the relevance and phase between the product obtained in correlation rule
Item set is constituted like product.
Step S25：For each user, its item bought is combed, is during its item for recommending not buy gathers
Item and recommendation.
Describe this step in detail.In the embodiment of the present invention, in view of in practical business, occurring repeating the same time to push away to user
The same situation for having purchased financial product is recommended, causes user to perplex and complain, the Item that a recommended user does not buy herein.
Referring to Fig. 3, Fig. 3 is based on commending contents and based on user collaborations for popularization new product provided in an embodiment of the present invention
A kind of implementation process figure of filtered recommendation, can include：
Step S31：To the new Item of popularization, extracted and inputted according to the buying behaviors of related user on these Item etc.
Feature；
Describe this step in detail.In the embodiment of the present invention, the input feature vector based on commending contents is shown in accompanying drawing 8：
The form that input feature vector based on user collaborative filterings is shown in step S11, this form of illustrated in greater detail.Herein will
Following table is obtained after S11 transposition, and (table of 6 rows 4 row, can extend row, column, general row>Row)：
Reference picture 9, if m representative products numbers, n represents user's number, and matrix value represents scoring, and 0 represents not score.
Step S32：These data are carried out with singular value SVD lowering dimension decompositions, selection reaches the former of gross energy threshold percentage
Individual element and data；
Describe this step in detail.In the embodiment of the present invention, on the other hand, inputted for " collaborative filtering based on user ",
In view of being limited different from Eigenvalues Decomposition just for square formation, singular value decomposition can be suitably used for Arbitrary Matrix.For in S31 steps
Form carry out singular value decomposition：
Citing describes above-mentioned formula implication in detail.According to the description in step S31, then the every row representative products attribute of U matrixes,
Now by dimensionality reduction U matrixes after, the attribute of each product can be represented (assuming that k is tieed up) with lower dimension.
Assuming thatThen A is carried out after singular value decomposition, obtained：
It is a diagonal matrix after split-matrix, it can be seen that Σ is very special.Each element non-negative, and subtract successively
Small, element value represents the matrix towards the change weight in the direction of the corresponding characteristic vector of some characteristic value.Find energy contribution
The corresponding element of the high singular value of accounting (because singular value declines quickly, owns shared by preceding several or preceding 10% singular value sum
90%) singular value sum gross energy can just be rapidly achieved, you can obtain feature after dimensionality reduction and input, be further to calculate similar
Degree provides support.
Here taken Σ diagonal are gone forward 2 elements (now preceding 2 elements account for gross energy ratio up to 85%), then Σ
(6*4) dimensionality reduction Σ (2*2), U (6*6) dimensionality reduction U (6*2), VT(4*4) dimensionality reduction is VT(2*4), so that A be approx changed intoTool
Body is as follows：
The higher input feature vector of contribution is preceding 2 row of U matrixes, and " title " is unrelated in step S31, based on form be
Second form in S31, as the form in step S11
Assuming that now with individual new user A, and this known user A is to the scoring vector of 6 product lists：[5 5 0
0 0 5]T, (this vector is column vector).Vector approximation after dimensionality reduction by calculating new user A after singular value decomposition：
User
Similarly, vector after user 1 in example table, user 2, user 3, the dimensionality reduction of user 4 can be calculated, it is as follows：
User
User 2=[- 0.4275-0.5172]1*2User 3=[- 0.3846 0.8246]1*2
User 4=[- 0.5859 0.0532]1*2User 2=[- 0.4275-0.5172]1*2
Step S33：Similar user set is combed out based on commending contents；
Describe this step in detail.In the embodiment of the present invention, for form in S31, directly calculate similarity and (be similar to S23
In calculating process, here is omitted), similar user group is produced, it is determined that the neighbour of each user.
Step S34：Collaborative filtering based on user combs similar user set；
Describe this step in detail.In the embodiment of the present invention, the feature in S32 is regard as input.Directly calculate similarity (class
The calculating process in S23 is similar to, here is omitted), similar user group is produced, it is determined that the neighbour of each user.
It is high by priority of the main input feature vector in step S32, the input feature vector priority override removed in dimensionality reduction.
Describe this step in detail.The customer manager of stock trader will make the recommendation of personalization to user A.Can be in existing subscriber
1st, in user 2, user 3, user 4, utilize vectorial [- 0.3775-0.0802] after new user A dimensionality reduction1*2Find out the user A's
Similar users.The similarity of vector is calculated herein, according to the formula of the Similarity Measure of Tanimoto coefficients,
,
T (user 2, user A)=0.511848, T (user 3, user A)=0.088056, T (user 2, user A)=0.77987
Then it can be seen that most like for user 1 in new user A and user 1- user 4, the similarity of the two is
0.801589, it is the recommendation in subsequent step.
Certainly, the present invention can suitably reduce similarity threshold in practice, and user 2 is also included in recommendation user set,
Recommendation is 0.77987.
Step S35：Merging forms Candidate Recommendation user set；
Describe this step in detail.In the embodiment of the present invention, merge the user group in S33 and S34, generate the user of candidate
Group.And the similarity between the user in candidate user group is calculated according to S34.
Step S36：For each Item, its matched user is combed, is not buy its during its matching is gathered
User and matching value；
Describe this step in detail.In the embodiment of the present invention, in the embodiment of the present invention, in view of in practical business, occurring to use
The family same time repeats to recommend the same situation for having purchased financial product, causes user to perplex and complain, only recommends not buy herein
The user of the Item.
Such as in S34 example, the scoring vector of observation user 1 is：[5 5 3 0 5 5]T, contrast commenting for new user A
Divide vector：[5 5 0 0 0 5]T.Then user 1 scored and new user A does not score product and sequence, i.e. { product are found out
5：5, product 3：3}.New user A product is recommended for product 5 and product 3, the product 1 bought for user A,
Production 2, product 6 do not do then and recommended.
This patent considers application of the collaborative filtering in financial product recommendation, takes full advantage of the intelligence of user group
It is intelligent, based on premise be that the close user of interest is interested in same product.Or user compares preference and has bought production with it
Product as category.The subjectivity of stock trader's financial product recommendation is so greatly promoted, hair is more removed from the statistical significance
Goods-saling is dug, strong decision support is provided for stock trader's lifting achievement.
The embodiment of each in this specification is described by the way of progressive, and what each embodiment was stressed is and other
Between the difference of embodiment, each embodiment identical similar portion mutually referring to.For device disclosed in embodiment
For, because it is corresponded to the method disclosed in Example, so description is fairly simple, related part is said referring to method part
It is bright.
Directly it can be held with reference to the step of the method or algorithm that the embodiments described herein is described with hardware, processor
Capable software module, or the two combination are implemented.Software module can be placed in random access memory (RAM), internal memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
The above method, it is preferred that recommend method to include the old Item of popularization：
To the old Item of popularization, input feature vector is extracted according to the buying behaviors of related user on these Item etc..1、
Input feature vector can include but is not limited to following：Seek advice from number of times, purchase number of times, Satisfaction Index, whether can buy like product
Deng；2nd, data extracting mode：Phone+survey under information system and line on line；3rd, it is according to same in being embodied
Under the product classification of rank, recommended；
The Item set of user preferences is combed out based on correlation rule；
Collaborative filtering based on Item combs similar Item set；
Merging forms Candidate Recommendation item set；
For each user, comb its item bought, be its recommendation do not buy item set in item with
And recommendation, wherein recommendation is mainly the numerical value between 0-1, expression be this user and item the goodness of fit, closer to 1
Represent that the goodness of fit is higher, represent that the goodness of fit is lower closer to 0, the index is obtained by item Similarity measures, and calculation formula is shown in
Tanimoto coefficients in step S23.
The above method, it is preferred that recommend method to include the new Item of popularization：
To the new Item of popularization, according to related user these Item (item here and item above are different,
Item herein main is the product newly promoted.Item above is the old product promoted.Difference is new item without too
Many user's purchase data) on buying behavior etc. extract input feature vector；
Singular value SVD lowering dimension decompositions are carried out to these data, SVD dimensionality reduction is a kind of general dimensionality reduction mode, here
Purpose be to simplify computation complexity, improve and recommend efficiency；
Selection reaches the preceding several elements and data of gross energy threshold percentage；
Similar user set is combed out based on commending contents；
Collaborative filtering based on user combs similar user set；
Merging forms Candidate Recommendation user set；
For each Item, comb its matched user, be its matching set in do not buy it user and
With value；
Understood via above-mentioned technical scheme, a kind of financial product for mixing collaborative filtering is pushed away disclosed in the embodiment of the present invention
Method is recommended, existing data accumulation and feature difference when being promoted for new, old product have been respectively adopted based on commending contents+base
In user collaborative filterings, based on correlation recommendation+recommended based on Item, and for the new Item larger situation of quantity, in meter
Dimensionality reduction is carried out first with singular value decomposition before calculating similitude, simplifies data, noise is removed, provided for the follow-up algorithm performance that improves
Base support.
Then these the old and new's Products Show results are merged in the way of in parallel, forms overall recommended candidate row
Table.These candidate lists are evaluated according to evaluation index and perfect, obtain optimal recommendation results.
As can be seen that these recommendations have refined the popularization features of various financial products in practical business, in model performance and
Possess preferable balance in application, the response of stock trader customer manager user when selling product can be greatly promoted
Rate, so that opening up the wider array of market space for stock trader provides powerful mean.
The foregoing description of the disclosed embodiments, enables professional and technical personnel in the field to realize or using the present invention.
A variety of modifications to these embodiments will be apparent for those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, it is of the invention
The embodiments shown herein is not intended to be limited to, and is to fit to and principles disclosed herein and features of novelty phase one
The most wide scope caused.
Term " first ", " second ", " the 3rd " " the 4th " in specification and claims and above-mentioned accompanying drawing etc. (if
In the presence of) it is for distinguishing similar part, without for describing specific order or precedence.It should be appreciated that so using
Data can exchange in the appropriate case, so that embodiments herein described herein can be with except illustrating herein
Order in addition is implemented.
Claims (6)
1. stock trader's Products Show method based on mixing collaborative filtering, it is characterised in that including：
The financial product information inputted during existing stock trader sells and some users purchase relevant financial product are recorded and counted after sale
According to, and form initial data base；
When promoting old product：
Some user's purchase relevant financial product records in initial data base are extracted, analyze related with Association Rule Analysis method
Incidence relation between financial product；And calculate the similar features between relevant financial product with collaborative filtering；
By with the relevance height and similar features between the product obtained in Association Rule Analysis method and collaborative filtering
Many products constitute a product set；
User is recommended after being screened to the product set.
2. stock trader's Products Show method as claimed in claim 1 based on mixing collaborative filtering, it is characterised in that described to sell
In financial product information include fixed income class product, cash class product, rights and interests class product.
3. stock trader's Products Show method as claimed in claim 1 based on mixing collaborative filtering, it is characterised in that with association
Also include carrying out input feature vector extraction, the input feature vector to initial data base before rule analysis method and collaborative filtering
It is including but not limited to following：Seek advice from number of times, purchase number of times, Satisfaction Index, whether can buy like product.
4. stock trader's Products Show method as claimed in claim 1 based on mixing collaborative filtering, it is characterised in that to the production
Product set, which carries out screening, to be included：For each user, its product bought is combed, is the product set that its recommendation is not bought
In product and recommendation.
5. stock trader's Products Show method based on mixing collaborative filtering as described in one of claim 1-4, it is characterised in that also
Popularization including new product：
The database of new product is set up, the database includes financial product information and the user's purchase relevant financial that user newly buys
Product record and after sale data；
Input feature vector formation analyze data is extracted according to new product information in the database of new product, and the analyze data is carried out
Singular value SVD lowering dimension decompositions, selection reaches the preceding several elements and data of gross energy threshold percentage；And according to the data after processing
Similarity is calculated, and finds out similar user group；
Similarity is directly calculated according to user profile, similar user group is produced；
Merge similar user group's composition user's set that above-mentioned new product information and user profile are obtained；
User is recommended after being screened to user set.
6. stock trader's Products Show method as claimed in claim 5 based on mixing collaborative filtering, it is characterised in that old by promoting
The user's set obtained when the product set obtained during product and popularization new product is merged, and forms the recommendation row of candidate
Table；
According to the degree of accuracy and coverage rate index recommended, proposed algorithm is evaluated and perfect.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710233443.9A CN107194754A (en) | 2017-04-11 | 2017-04-11 | Stock trader's Products Show method based on mixing collaborative filtering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710233443.9A CN107194754A (en) | 2017-04-11 | 2017-04-11 | Stock trader's Products Show method based on mixing collaborative filtering |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107194754A true CN107194754A (en) | 2017-09-22 |
Family
ID=59871967
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710233443.9A Pending CN107194754A (en) | 2017-04-11 | 2017-04-11 | Stock trader's Products Show method based on mixing collaborative filtering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107194754A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510384A (en) * | 2018-03-13 | 2018-09-07 | 四川享宇金信金融服务外包有限公司 | Loan product recommends method and device |
| CN108960986A (en) * | 2018-06-26 | 2018-12-07 | 西安交通大学 | A kind of supplier's recommended method based on web crawlers |
| CN108985935A (en) * | 2018-07-06 | 2018-12-11 | 兴业证券股份有限公司 | Financial product recommended method and storage medium |
| CN109300045A (en) * | 2018-10-25 | 2019-02-01 | 平安科技（深圳）有限公司 | Financial product recommended method, device, computer equipment and storage medium |
| CN109447728A (en) * | 2018-09-07 | 2019-03-08 | 平安科技（深圳）有限公司 | Financial product recommended method, device, computer equipment and storage medium |
| WO2019071906A1 (en) * | 2017-10-09 | 2019-04-18 | 平安科技（深圳）有限公司 | Financial product recommendation device and method, and computer-readable storage medium |
| TWI658420B (en) * | 2016-11-15 | 2019-05-01 | 平安科技（深圳）有限公司 | Method, device, server and computer readable storage medium for integrate collaborative filtering with time factor |
| CN110189191A (en) * | 2019-05-05 | 2019-08-30 | 苏宁易购集团股份有限公司 | A kind of financial product recommended method and system based on real-time model |
| CN110874788A (en) * | 2019-11-13 | 2020-03-10 | 合肥美的电冰箱有限公司 | Product recommendation method and device and computer storage medium |
| WO2020048059A1 (en) * | 2018-09-03 | 2020-03-12 | 平安科技（深圳）有限公司 | Fund data association method and system, computer device and storage medium |
| CN110880136A (en) * | 2018-09-06 | 2020-03-13 | 北京京东尚科信息技术有限公司 | Recommendation method, system, equipment and storage medium for matched product |
| CN111046111A (en) * | 2019-11-07 | 2020-04-21 | 上海琢学科技有限公司 | Data processing method and terminal equipment |
| CN111090983A (en) * | 2018-10-23 | 2020-05-01 | 北京好啦科技有限公司 | Questionnaire optimization method, device, computer equipment and storage medium |
| CN111881666A (en) * | 2020-07-09 | 2020-11-03 | 腾讯科技（深圳）有限公司 | Information processing method, device, equipment and storage medium |
| CN112182416A (en) * | 2020-09-10 | 2021-01-05 | 北京市科学技术情报研究所 | Scientific and technological resource dynamic collaborative filtering recommendation method based on user group preference |
| CN113254780A (en) * | 2021-06-08 | 2021-08-13 | 武汉卓尔数字传媒科技有限公司 | Information processing method and device, electronic equipment and computer storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156942A (en) * | 2011-04-08 | 2011-08-17 | 东华大学 | RFID based method for recommending commodities |
| CN102402625A (en) * | 2011-12-28 | 2012-04-04 | 深圳市五巨科技有限公司 | Method and system for recommending music |
| CN103093376A (en) * | 2013-01-16 | 2013-05-08 | 北京邮电大学 | Clustering collaborative filtering recommendation system based on singular value decomposition algorithm |
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
| CN105303444A (en) * | 2015-10-10 | 2016-02-03 | 苏州工业园区凌志软件股份有限公司 | Automatic adaption system for financial products |
| CN105430505A (en) * | 2015-11-13 | 2016-03-23 | 云南大学 | IPTV program recommending method based on combined strategy |
-
2017
- 2017-04-11 CN CN201710233443.9A patent/CN107194754A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102156942A (en) * | 2011-04-08 | 2011-08-17 | 东华大学 | RFID based method for recommending commodities |
| CN102402625A (en) * | 2011-12-28 | 2012-04-04 | 深圳市五巨科技有限公司 | Method and system for recommending music |
| CN103093376A (en) * | 2013-01-16 | 2013-05-08 | 北京邮电大学 | Clustering collaborative filtering recommendation system based on singular value decomposition algorithm |
| CN103678672A (en) * | 2013-12-25 | 2014-03-26 | 北京中兴通软件科技股份有限公司 | Method for recommending information |
| CN105303444A (en) * | 2015-10-10 | 2016-02-03 | 苏州工业园区凌志软件股份有限公司 | Automatic adaption system for financial products |
| CN105430505A (en) * | 2015-11-13 | 2016-03-23 | 云南大学 | IPTV program recommending method based on combined strategy |
Non-Patent Citations (2)
| Title |
|---|
| 吴杰: "综合用户偏好和优先新品推荐的协同过滤算法", 《计算机应用与软件》 * |
| 陈平华: "一种结合关联规则的协同过滤推荐算法", 《小型微型计算机***》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI658420B (en) * | 2016-11-15 | 2019-05-01 | 平安科技（深圳）有限公司 | Method, device, server and computer readable storage medium for integrate collaborative filtering with time factor |
| WO2019071906A1 (en) * | 2017-10-09 | 2019-04-18 | 平安科技（深圳）有限公司 | Financial product recommendation device and method, and computer-readable storage medium |
| CN108510384A (en) * | 2018-03-13 | 2018-09-07 | 四川享宇金信金融服务外包有限公司 | Loan product recommends method and device |
| CN108960986A (en) * | 2018-06-26 | 2018-12-07 | 西安交通大学 | A kind of supplier's recommended method based on web crawlers |
| CN108985935A (en) * | 2018-07-06 | 2018-12-11 | 兴业证券股份有限公司 | Financial product recommended method and storage medium |
| WO2020048059A1 (en) * | 2018-09-03 | 2020-03-12 | 平安科技（深圳）有限公司 | Fund data association method and system, computer device and storage medium |
| CN110880136A (en) * | 2018-09-06 | 2020-03-13 | 北京京东尚科信息技术有限公司 | Recommendation method, system, equipment and storage medium for matched product |
| CN109447728A (en) * | 2018-09-07 | 2019-03-08 | 平安科技（深圳）有限公司 | Financial product recommended method, device, computer equipment and storage medium |
| CN111090983A (en) * | 2018-10-23 | 2020-05-01 | 北京好啦科技有限公司 | Questionnaire optimization method, device, computer equipment and storage medium |
| CN111090983B (en) * | 2018-10-23 | 2022-04-22 | 北京好啦科技有限公司 | Questionnaire optimization method, device, computer equipment and storage medium |
| CN109300045A (en) * | 2018-10-25 | 2019-02-01 | 平安科技（深圳）有限公司 | Financial product recommended method, device, computer equipment and storage medium |
| CN110189191A (en) * | 2019-05-05 | 2019-08-30 | 苏宁易购集团股份有限公司 | A kind of financial product recommended method and system based on real-time model |
| CN111046111A (en) * | 2019-11-07 | 2020-04-21 | 上海琢学科技有限公司 | Data processing method and terminal equipment |
| CN110874788A (en) * | 2019-11-13 | 2020-03-10 | 合肥美的电冰箱有限公司 | Product recommendation method and device and computer storage medium |
| CN111881666A (en) * | 2020-07-09 | 2020-11-03 | 腾讯科技（深圳）有限公司 | Information processing method, device, equipment and storage medium |
| CN111881666B (en) * | 2020-07-09 | 2021-08-17 | 腾讯科技（深圳）有限公司 | Information processing method, device, equipment and storage medium |
| CN112182416A (en) * | 2020-09-10 | 2021-01-05 | 北京市科学技术情报研究所 | Scientific and technological resource dynamic collaborative filtering recommendation method based on user group preference |
| CN112182416B (en) * | 2020-09-10 | 2024-02-02 | 北京市科学技术情报研究所 | Scientific and technological resource dynamic collaborative filtering recommendation method based on user group preference |
| CN113254780A (en) * | 2021-06-08 | 2021-08-13 | 武汉卓尔数字传媒科技有限公司 | Information processing method and device, electronic equipment and computer storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107194754A (en) | Stock trader's Products Show method based on mixing collaborative filtering | |
| Bronnenberg et al. | The emergence of market structure in new repeat-purchase categories: The interplay of market share and retailer distribution | |
| Olson et al. | Advanced data mining techniques | |
| Tsiptsis et al. | Data mining techniques in CRM: inside customer segmentation | |
| Albadvi et al. | A hybrid recommendation technique based on product category attributes | |
| Yoseph et al. | The impact of big data market segmentation using data mining and clustering techniques | |
| Chorianopoulos | Effective CRM using predictive analytics | |
| Sifa et al. | Customer lifetime value prediction in non-contractual freemium settings: Chasing high-value users using deep neural networks and SMOTE | |
| Valletti et al. | Consumer profiling with data requirements: Structure and policy implications | |
| Kehoe et al. | Dynamic competition in the era of big data | |
| Zaim et al. | E-CRM success factors as determinants of customer satisfaction rate in retail website | |
| Ahn et al. | Customer attrition analysis in the securities industry: a large-scale field study in Korea | |
| Salazar et al. | An approach for the identification of cross-sell and up-sell opportunities using a financial services customer database | |
| Collier et al. | A perspective on data mining | |
| UnnisaBegum et al. | Data mining techniques for big data | |
| Raeisi et al. | E-commerce customer churn prediction by gradient boosted trees | |
| Yuping et al. | New methods of customer segmentation and individual credit evaluation based on machine learning | |
| Rezaeinia et al. | An integrated AHP-RFM method to banking customer segmentation | |
| Aldana | Data mining industry: emerging trends and new opportunities | |
| Wu et al. | Application of data mining in customer relationship management | |
| Chen et al. | Assortment optimization for the multinomial logit model with repeated customer interactions | |
| Moslehi et al. | Investigation of effective factors in expanding electronic payment in Iran using datamining techniques | |
| Singh et al. | Data mining: consumer behavior analysis | |
| Chu | A theory of durable dominance | |
| Xuanyuan et al. | Application of C4. 5 Algorithm in Insurance and Financial Services Using Data Mining 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: 20170922 |