CN105096149B - Business tine method for analyzing product association and device - Google Patents
Business tine method for analyzing product association and device Download PDFInfo
- Publication number
- CN105096149B CN105096149B CN201410197090.8A CN201410197090A CN105096149B CN 105096149 B CN105096149 B CN 105096149B CN 201410197090 A CN201410197090 A CN 201410197090A CN 105096149 B CN105096149 B CN 105096149B
- Authority
- CN
- China
- Prior art keywords
- value
- user
- content
- added product
- preference
- 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.)
- Active
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The present invention provides a kind of business tine method for analyzing product association and device, its method includes: to establish internet content knowledge base and value-added product content library, wherein, internet content knowledge base includes that user accesses internet behavioral data, and value-added product content library includes the value-added product content that operator provides；Internet content knowledge base and value-added product content library are established and mapped；Internet behavioral data, which is accessed, according to user obtains user to the preference information of each internet content；User is calculated to each value-added product content preference information；Determine the prediction preference value of the not used value-added product content of user in value-added product content library to each value-added product content preference information to the preference information of each internet content and user according to user；The user equipment used to user sends the prediction highest top n value-added product content of preference value, and N is the integer greater than zero.To accurately obtain the hobby of each user, the accuracy of prediction user preference is improved.
Description
Technical field
The present invention relates to the communication technology more particularly to a kind of business tine method for analyzing product association and device.
Background technique
Deepening continuously and enrich with mobile Internet application, mobile phone users are expanded rapidly, mobile telephone instant communication,
Mobile phone microblogging, mobile phone music are seen everywhere, this tide band is just being experienced by the telecom operators as mobile Internet key link
The impact come.In general, the referred to as increment industry other than voice service (making a phone call) and short message service (short message, multimedia message)
Business.Such as 12530 CRBT, number book house keeper, Fetion, mobile phone reading, mobile video, Mobile Telephone Gps, information house keeper, cell phone map
Deng.Value-added service has become the most important component part of operator's value chain, wide market, and demand is very big.In it was predicted that
State's mobile value-added service market will be increased with the speed per year over 30%.
How the business service condition according to user, carry out user preference analysis and prediction.It is had by oneself in system in operator
There is no user's usage behavior or have seldom user's usage behavior, how to utilize the internet preference of surfing Internet with cell phone user, push away
It recommends the own value-added service content product of operator and improves value-added service to improve the service level of operator itself
Quantity ordered reaches higher user satisfaction.As industry urgent problem to be solved.
In the prior art, equipment relevant to Products Show from value-added service (Value-added Service, referred to as:
VAS user's subscription content product data) are obtained in system, and user content product preference is calculated, is done according to user preference
Products Show.But the user for not ordering content product, user preference can not be obtained.Therefore prediction user is reduced
The accuracy of preference.
Summary of the invention
The present invention provides a kind of business tine method for analyzing product association and device, for improving the standard of value-added service push
True property.
The first aspect of the invention is to provide a kind of business tine method for analyzing product association, comprising:
Establish internet content knowledge base and value-added product content library, wherein the internet content knowledge base includes to use
Family accesses internet behavioral data, and the value-added product content library includes the value-added product content that operator provides；
The internet content knowledge base and the value-added product content library are established and mapped；
Internet behavioral data, which is accessed, according to user obtains the user to the preference information of each internet content；
The user is calculated to each value-added product content preference information；
According to the user to the preference information of each internet content and the user in each value-added product
Hold the prediction preference value that preference information determines the not used value-added product content of user described in the value-added product content library；
It is sent in value-added product described in the highest top n of the prediction preference value to the user equipment that the user uses
Hold, the N is the integer greater than zero.
It is in the first possible implementation, described to calculate the user to each increasing in conjunction with first aspect
It is worth product content preference information, comprising:
The user is obtained to the behavioral data of each value-added product, according to the behavior number of each value-added product
According to the calculating user to each value-added product content-preference value；
The content similarity in the value-added product content library between the different value-added product is obtained, according to difference
Content similarity between value-added product obtains the similar features predicted value of the user.
It is in the second possible implementation, described to obtain in conjunction with the first possible implementation of first aspect
Take the user to the behavioral data of each value-added product, according to the calculating of the behavioral data of each value-added product
User is to each value-added product content-preference value, comprising:
Collaborative filtering similarity is obtained according to the behavioral data of each value-added product, according to the collaborative filtering
Similarity calculation system filters predicted value；
Correlation rule confidence level is obtained according to the behavioral data of each value-added product, is set according to associated rule
Reliability calculates correlation rule predicted value.
In conjunction with the first possible implementation of first aspect or second of possible realization side of first aspect
Formula, it is in the third possible implementation, described that the internet behavioral data acquisition user is accessed to each according to user
The preference information of internet content, comprising:
According to the keyword search of the access times of each internet content and/or each internet content time
Number calculates the user to the preference value of each internet content.
In conjunction with the possible implementation of any one of the above of first aspect, in the fourth possible implementation,
The user obtains the preference value of each internet content by following formula:
Wherein, the VI, jIt is i-th of user to the preference value of j-th of internet content, the Max_
Threshold is the internet content maximum access times, and the Min_Threshold is internet content minimum visit
Ask number.
In conjunction with the 4th kind of possible implementation of first aspect, in a fifth possible implementation, by such as
Lower formula, which calculates, obtains each user preference prediction preference value:
Wherein, the S is that the user preference of the internet content predicts preference value, the VhFor the internet
The preference value of content, the value-added product content-preference value, the collaborative filtering predicted value between the value-added product, correlation rule
Predicted value or similar features predicted value, the WhFor with the VhCorresponding weighted value, the M are the integer greater than zero.
The second aspect of the invention is to provide a kind of business tine analyzing product association device, comprising:
Module is established, for establishing internet content knowledge base and value-added product content library, wherein the internet content
Knowledge base includes that user accesses internet behavioral data, and the value-added product content library includes in the value-added product that operator provides
Hold；
Mapping block is mapped for establishing the internet content knowledge base and the value-added product content library；
Processing module obtains the user to each internet content for accessing internet behavioral data according to user
Preference information；It is also used to calculate the user to each value-added product content preference information；It is also used to according to the user
Preference information and the user to each internet content determine the increasing to each value-added product content preference information
It is worth the prediction preference value of the not used value-added product content of user described in product content library；
Pushing module, the user equipment for using to the user send the highest top n institute of prediction preference value
Value-added product content is stated, the N is the integer greater than zero.
In conjunction with the second aspect, in the first possible implementation, the processing module is specifically used for described in acquisition
User calculates the user couple according to the behavioral data of each value-added product to the behavioral data of each value-added product
Each value-added product content-preference value；Also particularly useful for the different value-added product in the acquisition value-added product content library
Between content similarity, the similar features for obtaining the user according to the content similarity between the different value-added product are pre-
Measured value.
In conjunction with the first possible implementation of the second aspect, in the second possible implementation, the place
Module is managed, specifically for obtaining collaborative filtering similarity according to the behavioral data of each value-added product, according to described
Collaborative filtering similarity calculation system filters predicted value；It is also used to be obtained according to the behavioral data of each value-added product
Correlation rule confidence level, according to associated regular confidence calculations correlation rule predicted value.
In conjunction with the first possible implementation of the second aspect or second of possible realization side of the second aspect
Formula, in the third possible implementation, the processing module, specifically for the access according to each internet content
The keyword search number of number and/or each internet content calculates the user to each internet content
Preference value.
In conjunction with the possible implementation of any one of the above of the second aspect, in the fourth possible implementation,
The processing module obtains the user to the preference value of each internet content by following formula:
Wherein, the VI, jIt is i-th of user to the preference value of j-th of internet content, the Max_
Threshold is the internet content maximum access times, and the Min_Threshold is internet content minimum visit
Ask number.
In conjunction with the 4th kind of possible implementation of the second aspect, in a fifth possible implementation, the place
It manages module and each user preference prediction preference value of acquisition is calculated by following formula:
Wherein, the S is that the user preference of the internet content predicts preference value, the VhFor the internet
The preference value of content, the value-added product content-preference value, the collaborative filtering predicted value between the value-added product, correlation rule
Predicted value or similar features predicted value, the WhFor with the VhCorresponding weighted value, the M are the integer greater than zero.
Business tine method for analyzing product association provided in this embodiment, by establishing internet content knowledge base and increment
Product content library, wherein the internet content knowledge base includes that user accesses internet behavioral data, value-added product content library
Comprising the value-added product content that operator provides, the internet content knowledge base and value-added product content library foundation are reflected
It penetrates, accesses internet behavioral data further according to user and obtain the user to the preference information of each internet content, and root
The user is calculated according to the corresponding value-added product behavioral data of the user to believe each value-added product content-preference
Breath, according to the user to the preference information of each internet content and the user to each value-added product content-preference
Information determines the prediction preference value of the not used value-added product content of user described in the value-added product content library, finally to
The user equipment that family uses sends the prediction highest top n value-added product content of preference value, and N is the integer greater than zero.To real
Internet behavioral data is now accessed according to user and obtains user's preference information to the business that do not order in a network, further according to
The value-added product behavioral data that user provides operator, it should be noted that value-added product behavioral data can be to order herein
Purchase information, or user browse portal website of operator when for different value-added product browsing time, whether pay close attention to
Information likes phase with it to push to this kind of user to accurately obtain the hobby of the user without subscribing to value-added product
Same or similar value-added product, improves the accuracy of service propelling.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do one simply to introduce, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of flow diagram of business tine method for analyzing product association provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of another business tine method for analyzing product association provided in an embodiment of the present invention；
Fig. 3 is collaborative filtering schematic diagram provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of business tine analyzing product association device provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Fig. 1 be a kind of flow diagram of business tine method for analyzing product association provided in an embodiment of the present invention, when with
When family accesses internet by terminal devices such as mobile phone, PC, tablet computers, executing subject of the server as this method can
The following each steps of this method are executed, difference preference's information of different user is obtained, thus when user accesses internet again,
The server recommends the other content similar with the user preference to user.Referring to Fig. 1, this method comprises the following steps:
Step 100 establishes internet content knowledge base and value-added product content library, wherein internet content knowledge base packet
Internet behavioral data is accessed containing user, value-added product content library includes the value-added product content that operator provides.
Specifically, wherein each operator is corresponded to and generates a value-added product content library, in each value-added product
Hold the value-added product content that library only includes the particular category that corresponding operator releases, wherein value-added product content can be coloured silk
Bell, video, books, using etc..Using web crawlers, the classifying content system and content of multiple third party's professional websites are crawled.
By internal duplicate removal and manual sorting, internet content knowledge base is formed.For interconnecting music knowledge library, table 1 is interconnection sound
Happy knowledge base schematic table, referring to table 1:
Table 1
Internet content knowledge base and value-added product content library are established mapping by step 101；
Specifically, value-added product content library is defined good.Value-added product content library and internet content knowledge base are built
Vertical mapping relations.By taking music libraries as an example, table 2 is internet content knowledge base and value-added product content library mapping relations schematic table
Lattice, referring to table 2, because content is fixed (example: song title=Jiangnan Style), mapping relations are automatic mappings.If internet
Content and value-added product content have that fraction is inconsistent, can be manually adjusted.
Table 2
| Internet content knowledge base | VAS content library | Attribute classification |
| Music attribute | CRBT attribute | Attribute classification |
| Jiangnan Style | Jiangnan Style | Song title |
| Bigbang | Bigbang | Song title |
| Most dazzle national wind | Most dazzle national wind | Song title |
| Psy bird uncle | Psy | Singer |
| Phoenix legend | Phoenix legend | Singer |
| It is hot | It is atypical | Style of song |
| It cures | It is atypical | Style of song |
Step 102 accesses behavioral data acquisition user in internet according to user to the preference information of each internet content.
Step 103 calculates user to each value-added product content preference information.
It should be noted that step 102 and step 103 are that two steps arranged side by side can that is, after executing step 101
To carry out step 102 and step 103 simultaneously.
It is step 104, inclined to each value-added product content to the preference information of each internet content and user according to user
Good information determines the prediction preference value of the not used value-added product content of user in value-added product content library.
Step 105, the user equipment used to user send the prediction highest top n value-added product content of preference value, and N is
Integer greater than zero.
Business tine method for analyzing product association provided in this embodiment, by establishing internet content knowledge base and increment
Product content library, wherein the internet content knowledge base includes that user accesses internet behavioral data, value-added product content library
Comprising the value-added product content that operator provides, the internet content knowledge base and value-added product content library foundation are reflected
It penetrates, accesses internet behavioral data further according to user and obtain the user to the preference information of each internet content, and root
The user is calculated according to the corresponding value-added product behavioral data of the user to believe each value-added product content-preference
Breath, according to the user to the preference information of each internet content and the user to each value-added product content-preference
Information determines the prediction preference value of the not used value-added product content of user described in the value-added product content library, finally to
The user equipment that family uses sends the prediction highest top n value-added product content of preference value, and the N is the integer greater than zero.From
And realize that accessing internet behavioral data according to user obtains user's preference information to the business that do not order in a network, then
The value-added product behavioral data that operator is provided according to user, it should be noted that value-added product behavioral data can be with herein
For ordering information, or user browse portal website of operator when for different value-added product browsing time, whether close
The information such as note are liked to push to this kind of user with it to accurately obtain the hobby of the user without subscribing to value-added product
Good same or similar value-added product, improves the accuracy of prediction user preference.
Further, for Fig. 1 step 103, a kind of feasible implementation are as follows:
Step 103a, user is obtained to the behavioral data of each value-added product, according to each value-added product
Behavioral data calculates the user to each value-added product content-preference value；
Step 103b, the content similarity in the value-added product content library between the different value-added product, root are obtained
The similar features predicted value of the user is obtained according to the content similarity between the different value-added product.
Wherein, the similar features predicted value of user is for the similarity value for being product two-by-two.
Specifically, obtaining user (VAS system user) by value-added service (VAS) system uses content product behavior (point
Hit, buy, order) data, user is calculated to VAS content product preference value (two).Using collaborative filtering data mining algorithm
(Collaborative Filtering), market basket analysis data mining algorithm (Apriori), similar features data mining are calculated
Method these three mining algorithms of (Similar Feature) similar features calculate the association between value-added product content-content product
Degree, confidence level, similarity.
And in the prior art, according to the content product of user preference, generic content is matched into content product library and is produced
Product recommend user.Example: Ring Back Tone service class library
User has subscribed CRBT " Jiangnan Style ", according to similar recommendation, can give user's color bell recommendation
"Bigbang".But matched according to product classification and recommend similar product, matching rule is single.According to user preference
Content product matches similar product in product library and recommends user, and matching rule is single, cannot obtain user really using row
For related product.(such as: 70% user for liking phoenix legend song by analysis likes the song of Li Jia fine jade, this 2 liang of classes
Song is simultaneously not belonging to same class.This product relevance may be interim, once gas, product relevance are crossed in some TV play
It will reduce).
And in the present embodiment, to subscription content product user, recommend to be changed to the big own production of content association of preference
Product.
Further, for step 103a, a kind of feasible implementation are as follows:
Collaborative filtering similarity is obtained according to the behavioral data of each value-added product, according to the collaborative filtering
Similarity calculation system filters predicted value；
Correlation rule confidence level is obtained according to the behavioral data of each value-added product, is set according to associated rule
Reliability calculates correlation rule predicted value.
It can be seen that product direct correlation is not only similar product association in the present embodiment, need to use row according to all users
To calculate the degree of association between product and product.
Further, for step 102 in Fig. 1, a kind of feasible implementation are as follows:
According to the access times of each internet content and/or the keyword search number meter of each internet content
The user is calculated to the preference value of each internet content.
Specifically, for example, table 3 is that user pays close attention to internet music schematic table, according to the table: can analyze use
Family passes through the happy data of Interworking GateWay phonetic notation, and user is calculated to the concern number of music.
| User | Music | Pay close attention to number |
| 13978965412 | Most dazzle national wind | 6 |
| 13978965412 | Fiery ballad | 5 |
| 13978965412 | Phoenix legend | 5 |
| 13945632178 | Jiangnan style | 4 |
| 13945632178 | Chuan Zanglu | 1 |
| …… | …… | …… |
Further, by normalizing formula, user is calculated to the preference value of each internet content, is specifically led to
Cross the preference value that following formula (1) obtains the internet content:
Wherein, VI, jIt is internet content for preference value Max_Threshold of i-th of user to j-th of internet content
Maximum access times, Min_Threshold are internet content minimum access number.It should be noted that internet content is most
Big access times and internet content minimum access number are directed to the maximum access of all internet contents of user's access
Number and minimum access number.Generally, both for certain specific class content, maximum access times and minimum access are counted
Number, by taking music as an example, for the internet content of this type of music, counts highest clicking rate and minimum point in all music
Hit rate.
With the data instance of table 3 above, if there are five records, Min_Threshold=1, Max_Threshold can be obtained
=6.Wherein, preference value V of the user " 13945632178 " to the internet content of " Jiangnan style "I, jUsing formula (1) are as follows:
(6-1)=0.6.
On the basis of table 3 above, result is exported to following based on formula (1):
Table 4
| User | Music | The preference value of internet content |
| 13978965412 | Most dazzle national wind | 1 |
| 13978965412 | Fiery ballad | 0.8 |
| 13978965412 | Phoenix legend | 0.8 |
| 13945632178 | Jiangnan style | 0.6 |
| 13945632178 | Chuan Zanglu | 0 |
| …… | …… | …… |
Preference value is predicted finally, calculating by following formula (2) and obtaining each user preference:
Wherein, the S is that the user preference of the internet content predicts preference value, VhFor the inclined of internet content
Good value, collaborative filtering predicted value, correlation rule predicted value or similar features between value-added product content-preference value, value-added product
Predicted value, WhFor with VhCorresponding weighted value.That is, S is the preference value and the respective weight sum of products of each internet content, M is
Integer greater than zero.It further, include five kinds of preference informations (preference values of internet content, value-added product in the present embodiment
Collaborative filtering predicted value between content-preference value, value-added product, correlation rule prediction and similar features predicted value) acquisition side
Formula, therefore the M in formula (2) takes 5, if adding other modes, the value of M is according to the more of preference information acquisition modes quantity
Few corresponding adjustment, not limits herein.
Fig. 2 is the flow diagram of another business tine method for analyzing product association provided in an embodiment of the present invention, under
Face is referring to Fig. 2, and this method comprises the following steps:
Step 1 obtains user VAS content behavioral data.
Step 2 is user VAS content behavioral data configuration behavior weight.
Step 3 obtains VAS product content preference information.
Specifically, the step 3 can be with further division are as follows: step 3a, obtain collaborative filtering similarity；Step 3b is obtained
Correlation rule confidence level.
Step 4 calculates VAS product content preference value.
Specifically, step 4 can be with further division are as follows: step 4a, obtained collaborative filtering predicted value；Step 4b, it obtains
Correlation rule predicted value；Step 4c, similar features predicted value is obtained.Wherein, the collaborative filtering predicted value in step 4a is to be based on
What the collaborative filtering similarity in step 3a obtained；Correlation rule predicted value in step 4b is based on the association rule in step 3b
Then confidence level obtains；Similar features predicted value in step 4c is obtained by subsequent step 6.
Step 5 obtains VAS product content library data.
Content similarity between step 6, acquisition VAS product.
Specifically, get the content similarity between VAS product, according to the content similarity between VAS product two-by-two,
Similar features predicted value is obtained in step 4c.
Step 7 obtains internet content knowledge base data.
Specifically, the data are that user accesses internet behavioral data.
Step 8 obtains user to the preference value of each internet content.
Specifically, being to access internet behavioral data based on user to obtain user to the preference value of each internet content.
Step 9, placement algorithm weight.
Specifically, due to step 4 and step 8 respectively according to different algorithms obtain preference information (internet content it is inclined
Good value, the prediction of the collaborative filtering predicted value between value-added product content-preference value, value-added product, correlation rule and similar features are pre-
Measured value), therefore since different algorithms influences final prediction preference value different, the important journey based on each algorithm
Degree is the different weight of each algorithm configuration.
Step 10, the prediction preference value for obtaining VAS product content.
Business tine method for analyzing product association provided in an embodiment of the present invention is illustrated by taking music as an example below:
Fig. 3 is collaborative filtering schematic diagram provided in an embodiment of the present invention, by taking film as an example, when user wants to see electricity now
Shadow, but it is not aware that any portion specifically seen, usual situation, most people can inquire the friend of hobby similar with oneself having, and ask friend
Friend recommends the good film seen recently.And collaborative filtering (Collaborative Filtering, abbreviation CF) core concept and this
Similar, referring to Fig. 3, CF is exactly using the hobby for identical group of having similar tastes and interests with user, sample, come to use in simple terms
Person generates him may interested information.
Collaborative filtering is divided into collaborative filtering and the two different methods of project-based collaborative filtering based on user.
Collaborative filtering based on user, as shown in figure 3, probably thought is because user A and user C has purchased article A
With article C, it is believed that user A and user C are the users having similar tastes and interests.Since user C has purchased article D, it is believed that with
Family A also can be interested in article D, then just recommends article D to user A.
Project-based collaborative filtering, general thought are all bought by user A and user B because of article A and article C,
It is believed that article A and article C are similar articles.Since user C has purchased article A, it is believed that user C can also like article
C then just recommends article C to user C.
Merge user's internet content and map VAS content-preference value, user uses VAS content product preference value.Because being
The association similarity for calculating VAS content product, needing to reject non-VAS content product in internet music, (italic indicates user
Using VAS content product, non-italic indicates that user's internet content maps VAS content product).
Further, it is excavated using data mining software configuration Collaborative Filtering collaborative filtering.
Specifically, Collaborative Filtering collaborative filtering may include: the step of excavation
Step 200 imports user preference data using text delivery node.
Specifically, executing step 201 after importing user preference data.
Step 201, type node are used to configure the field for needing to analyze.
Specifically, such as user mobile phone is as major key, CRBT and VAS content-preference value are input field, similarity and are pushed away
List is recommended as system output.Further, it executes after completing step 201, executes step 202 and 203 respectively.
Step 202 is calculated by data mining, generates a similarity matrix, and export the similarity matrix.
Specifically, the similarity matrix can be used for the calculating of real-time online recommendation, which is referred to down
Cultural association is the same as filter result table 2.
Step 203 is calculated by data mining, generates a recommendation results list, and exports the recommendation results list.
Specifically, the recommendation results list is used directly for user's recommendation.The recommendation results list reference hereafter cooperates with
Filter result table 1.
Execute data digging flow, it can be deduced that filtering predicted value (the VAS content product of each user VAS content product
Filtering predicted value=1 indicate directly output recommendation results), shown in following collaborative filtering result table 1.
Collaborative filtering result table 1
It can be seen that 13658741236 users are 1 to the filtering predicted value of " love apartment theme song " VAS content product,
Indicate that user is likely to order this CRBT.
Execute data digging flow, the similarity between available content product-content product.
Collaborative filtering result table 2
| CRBT 1 | CRBT 2 | Similarity |
| It goes home, | Jiangnan Style | 0.8 |
| It goes home, | Love apartment theme song | 0.7 |
| It goes home, | The temptation gone home | 0.6 |
| The temptation gone home | Love apartment theme song | 0.85714287 |
| The temptation gone home | Jiangnan Style | 0.75 |
| The temptation gone home | Most dazzle national wind | 0.6666667 |
| The temptation gone home | It goes home | 0.6 |
| Jiangnan Style | Most dazzle national wind | 0.8888889 |
| Jiangnan Style | Love apartment theme song | 0.875 |
| Jiangnan Style | It goes home | 0.8 |
| Jiangnan Style | The temptation gone home | 0.75 |
| Most dazzle national wind | It goes home | 0.9 |
| Most dazzle national wind | Jiangnan Style | 0.8888889 |
| Most dazzle national wind | Love apartment theme song | 0.7777778 |
| Most dazzle national wind | The temptation gone home | 0.6666667 |
| Love apartment theme song | Jiangnan Style | 0.875 |
| Love apartment theme song | The temptation gone home | 0.85714287 |
| Love apartment theme song | Most dazzle national wind | 0.7777778 |
| Love apartment theme song | It goes home | 0.7 |
| It goes home | Most dazzle national wind | 0.9 |
Like product can be recommended to user by analyzing result.Such as there is a user to pay close attention to song in mobile Internet
" going home " is more, and does not order this CRBT, according to product similarity, then preferentially " most dazzles the people to his color bell recommendation " going home "
Race's wind ".
The purpose of Apriori association rules mining algorithm, which is to concentrate in a data, finds out related object, sometimes
It waits and is also referred to as " shopping basket " analysis (market basketanalysis).For example, the customer of purchase shoes, there is 30% possibility
Socks can be bought, milk can also be bought by buying in the customer of bread 60%, then socks and shoes, bread and milk are exactly related right
As.
Similar Feature similar features mining algorithm is that one on the idea basis of collaborative filtering is special
Using.It is to carry out user's phase in similar Collaborative Recommendation by the weight of dimension according to the description of the different dimensions of an article
It is calculated like degree.Here article is equivalent to the user in the Collaborative Recommendation based on user, and the description of different dimensions is with regard to suitable here
The commodity that user subscribes in Collaborative Recommendation, and weight is equivalent to the similarity to commodity.
For example the CRBT of one entitled " process ", its singer are " Chen Chusheng ", it is " 2.0 that style, which is " sweetness ", price,
Member ", language are " Chinese ".Just be equivalent to the user that one is named as " passing through " have subscribed be named as " singer _ Chen Chusheng ", " style _
It is happy ", " price _ 2.0 ", " language _ Chinese " 4 projects, the similarity between calculating " process " and other users now.
The weight that different attribute classifies for a CRBT is different, the influence meeting of singer and style to CRBT
It greatly a bit, is 50；And the influence of price and language to relationship between CRBT can be a little bit smaller, is 10.This is influenced like cooperateing with
Collaborative filtering predicted value in filter, i.e., " passing through " user is 50 points to the collaborative filtering predicted value of " singer _ Chen Chusheng ", right
The collaborative filtering predicted value of " price _ 2.0 " is 10 points.Then with regard to carrying out similarity calculation.
According to above-mentioned each mining algorithm, the final correlation rule predicted value for obtaining user VAS content product, similar features
Predicted value is no longer illustrated.
User is calculated to content-preference value weighted sum under each classification to five kinds of algorithm configuration weights, then by formula 2 above
Obtain the corresponding prediction preference value of value-added product content.
Calculation formula:
User is obtained to the prediction preference value of VAS content product.The user VAS content product TOP N that is obtained with this as a result,
Data source as user-customized recommended.
The present embodiment only passes through music and the value-added product content of user is analyzed and recommended, and others increment is produced
Product are equally applicable, and operator newly establishes 4G cell phone video platform.Prepare (to like using fortune for the user group of downloading CRBT > 2 time
The user of battalion's business business), plan a mobile video promotion business.Third party website video note is accessed according to such user group
Record, analyzes user to video, video type, star, region preference.Certainly further according to internet video library mapping operator
There is 4G cell phone video library, calculates the preference prediction preference value of each user-association mobile video, TOP N mobile video is sent to
The user equipment of user.
It should be noted that user group is specific for mobile video business, and because it is newly-built business of setting up one's own business, VAS system
No user that unites uses mobile video business record, can only analyze association VAS hand by user's internet access behavioral data
Machine video preference.It can also use Apriori association rules mining algorithm, Similar Feature similar features are excavated and calculated
Method and Collaborative Filtering collaborative filtering mining algorithm are analyzed, and are calculated between mobile video-mobile video
The degree of association, confidence level, similarity.In the present embodiment, in the case where determining user group, can across classification recommendation association it is own
Content product carries out new business convenient for operator, and can accomplish accurate personal marketing.
Fig. 4 is a kind of structural schematic diagram of business tine analyzing product association device provided in an embodiment of the present invention, the dress
Setting can be arranged on corresponding server, can be made of independent processor, memory and bus, interface module,
Also it can integrate on the original device of the server, not limited herein for the device concrete implementation mode, also, should
Device can execute the corresponding steps of Fig. 1 and Fig. 2, can be realized Fig. 1 and Fig. 2 and correspond to the technology effect that scheme can be realized
Fruit, referring to Fig. 4, which includes: to establish module 10, mapping block 11, processing module 12, pushing module 13.
Module 10 is established, for establishing internet content knowledge base and value-added product content library, wherein in the internet
Holding knowledge base includes that user accesses internet behavioral data, and the value-added product content library includes the value-added product that operator provides
Content；
Mapping block 11 is mapped for establishing the internet content knowledge base and the value-added product content library；
Processing module 12 obtains the user to each internet content for accessing internet behavioral data according to user
Preference information；It is also used to calculate the user to each value-added product content preference information；It is also used to according to the use
Described in family is determining to each value-added product content preference information to the preference information of each internet content and the user
The prediction preference value of the not used value-added product content of user described in value-added product content library；
Pushing module 13, the user equipment for using to the user send the highest top n of the prediction preference value
The value-added product content, the N are the integer greater than zero.
Business tine analyzing product association device provided in this embodiment establishes internet content knowledge by establishing module
Library and value-added product content library, wherein the internet content knowledge base includes that user accesses internet behavioral data, and increment produces
Product content library includes the value-added product content that operator provides, and mapping block is by the internet content knowledge base and the increment
Mapping is established in product content library, and processing module accesses internet behavioral data according to user and obtains the user to each internet
The preference information of content, and the user is calculated to each institute according to the corresponding value-added product behavioral data of the user
Value-added product content preference information is stated, according to the user to the preference information of each internet content and the user to each
The value-added product content preference information determines the not used value-added product content of user described in the value-added product content library
Prediction preference value, user equipment that final pushing module is used to the user sends the highest preceding N of prediction preference value
A value-added product content, the N are the integer greater than zero.To realize that accessing internet behavioral data according to user obtains
User's preference information to the business that do not order in a network, the value-added product behavior number that operator is provided further according to user
According to, it should be noted that value-added product behavioral data can be ordering information herein, or user browses operator's portal
When website for the browsing time of different value-added product, the information such as whether pay close attention to, produced to accurately obtain without subscribing to increment
The hobby of the user of product likes same or similar value-added product with it to push to this kind of user, predicts user preference
Accuracy.
Further, the processing module 12, specifically for obtaining behavior of the user to each value-added product
Data calculate the user to each value-added product content-preference value according to the behavioral data of each value-added product；
Also particularly useful for the content similarity obtained in the value-added product content library between the different value-added product, according to different institutes
State the similar features predicted value that the content similarity between value-added product obtains the user.
Further, the processing module 12, specifically for being obtained according to the behavioral data of each value-added product
Collaborative filtering similarity is taken, predicted value is filtered according to the collaborative filtering similarity calculation system；It is also used to according to described each
The behavioral data of the value-added product obtains correlation rule confidence level, is predicted according to associated regular confidence calculations correlation rule
Value.
Further, the processing module 12, specifically for according to the access times of each internet content and/or
The keyword search number of each internet content calculates the user to the preference value of each internet content.
Further, the processing module 12 obtains the user to each internet content by following formula
Preference value:
Wherein, the VI, jFor Max_Threshold described in preference value of i-th of user to j-th of internet content
For the internet content maximum access times, the Min_Threshold is the internet content minimum access number.
Further, the processing module 12 is calculated by following formula obtains each user preference prediction preference
Value:
Wherein, the S is that the user preference of the internet content predicts preference value, the VhFor the internet
The preference value of content, the value-added product content-preference value, the collaborative filtering predicted value between the value-added product, correlation rule
Predicted value or similar features predicted value, the WhFor with the VhCorresponding weighted value, the M are the integer greater than zero.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above-mentioned each method embodiment can lead to
The relevant hardware of program instruction is crossed to complete.Program above-mentioned can be stored in a computer readable storage medium.The journey
When being executed, execution includes the steps that above-mentioned each method embodiment to sequence；And storage medium above-mentioned include: ROM, RAM, magnetic disk or
The various media that can store program code such as person's CD.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (10)
1. a kind of business tine method for analyzing product association characterized by comprising
Establish internet content knowledge base and value-added product content library, wherein the internet content knowledge base is visited comprising user
Ask internet behavioral data, the value-added product content library includes the value-added product content that operator provides；
The internet content knowledge base and the value-added product content library are established and mapped；
Internet behavioral data, which is accessed, according to user obtains the user to the preference information of each internet content；
The user is calculated to each value-added product content preference information；
It is inclined to each value-added product content to the preference information of each internet content and the user according to the user
Good information determines the prediction preference value of the not used value-added product content of user in the value-added product content library；
The user equipment used to the user sends value-added product content described in the highest top n of the prediction preference value, institute
Stating N is the integer greater than zero；
It is described that the user is obtained to the preference information of each internet content, packet according to user's access internet behavioral data
It includes:
According to the access times of each internet content and/or the keyword search number meter of each internet content
The user is calculated to the preference value of each internet content.
2. the method according to claim 1, wherein described calculate the user in each value-added product
Hold preference information, comprising:
The user is obtained to the behavioral data of each value-added product, according to the behavioral data meter of each value-added product
The user is calculated to each value-added product content-preference value；
The content similarity in the value-added product content library between the different value-added product is obtained, according to the different increments
Content similarity between product obtains the similar features predicted value of the user.
3. according to the method described in claim 2, it is characterized in that, described obtain the user to each value-added product
Behavioral data calculates the user to each value-added product content-preference according to the behavioral data of each value-added product
Value, comprising:
Collaborative filtering similarity is obtained according to the behavioral data of each value-added product, it is similar according to the collaborative filtering
It spends computing system and filters predicted value；
Correlation rule confidence level is obtained according to the behavioral data of each value-added product, according to associated regular confidence level
Calculate correlation rule predicted value.
4. method according to claim 1 to 3, which is characterized in that the user is in each internet
The preference value of appearance is obtained by following formula:
Wherein, the VI, jIt is i-th of user to the preference value of j-th of internet content, the Max_Threshold is
The internet content maximum access times, the Min_Threshold are the internet content minimum access number.
5. according to the method described in claim 4, obtaining each user preference it is characterized in that, calculating by following formula
Predict preference value:
Wherein, the S is that the user preference of the internet content predicts preference value, the VhFor the internet content
Preference value, the value-added product content-preference value, the collaborative filtering predicted value between the value-added product, correlation rule prediction
Value or similar features predicted value, the WhFor with the VhCorresponding weighted value, the M are the integer greater than zero.
6. a kind of business tine analyzing product association device characterized by comprising
Module is established, for establishing internet content knowledge base and value-added product content library, wherein the internet content knowledge
Library includes that user accesses internet behavioral data, and the value-added product content library includes the value-added product content that operator provides；
Mapping block is mapped for establishing the internet content knowledge base and the value-added product content library；
Processing module obtains the user to the preference of each internet content for accessing internet behavioral data according to user
Information；It is also used to calculate the user to each value-added product content preference information；It is also used to according to the user to every
The preference information of a internet content and the user determine that the increment produces to each value-added product content preference information
The prediction preference value of the not used value-added product content of user in product content library；
Pushing module, the user equipment for using to the user send increasing described in the highest top n of the prediction preference value
It is worth product content, the N is the integer greater than zero；
The processing module, specifically for according to each internet content access times and/or each internet
The keyword search number of content calculates the user to the preference value of each internet content.
7. device according to claim 6, which is characterized in that the processing module is specifically used for obtaining the user couple
The behavioral data of each value-added product calculates the user to each institute according to the behavioral data of each value-added product
State value-added product content-preference value；Also particularly useful between the different value-added product in the acquisition value-added product content library
Content similarity obtains the similar features predicted value of the user according to the content similarity between the different value-added product.
8. device according to claim 7, which is characterized in that the processing module is specifically used for according to each institute
The behavioral data for stating value-added product obtains collaborative filtering similarity, is filtered and is predicted according to the collaborative filtering similarity calculation system
Value；It is also used to obtain correlation rule confidence level according to the behavioral data of each value-added product, according to associated rule
Confidence calculations correlation rule predicted value.
9. according to device described in claim 6-8 any one, which is characterized in that the processing module is obtained by following formula
The user is obtained to the preference value of each internet content:
Wherein, the VI, jIt is i-th of user to the preference value of j-th of internet content, the Max_Threshold is
The internet content maximum access times, the Min_Threshold are the internet content minimum access number.
10. device according to claim 9, which is characterized in that the processing module is calculated by following formula and obtained often
A user preference predicts preference value:
Wherein, the S is that the user preference of the internet content predicts preference value, the VhFor the internet content
Preference value, the value-added product content-preference value, the collaborative filtering predicted value between the value-added product, correlation rule prediction
Value or similar features predicted value, the WhFor with the VhCorresponding weighted value, the M are the integer greater than zero.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410197090.8A CN105096149B (en) | 2014-05-09 | 2014-05-09 | Business tine method for analyzing product association and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410197090.8A CN105096149B (en) | 2014-05-09 | 2014-05-09 | Business tine method for analyzing product association and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105096149A CN105096149A (en) | 2015-11-25 |
| CN105096149B true CN105096149B (en) | 2019-02-01 |
Family
ID=54576512
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410197090.8A Active CN105096149B (en) | 2014-05-09 | 2014-05-09 | Business tine method for analyzing product association and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105096149B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107644094B (en) * | 2017-09-27 | 2020-07-14 | 广州大学 | Method, device, server and storage medium for constructing boutique resource library |
| CN113326427A (en) * | 2020-11-17 | 2021-08-31 | 崔海燕 | Service push configuration updating method based on big data positioning and cloud computing center |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853463A (en) * | 2009-03-30 | 2010-10-06 | 北京邮电大学 | Collaborative filtering recommending method and system based on client characteristics |
| CN102663626A (en) * | 2012-04-26 | 2012-09-12 | 焦点科技股份有限公司 | Collaborative filtering recommended method based on provincial characteristics |
| CN103473354A (en) * | 2013-09-25 | 2013-12-25 | 焦点科技股份有限公司 | Insurance recommendation system framework and insurance recommendation method based on e-commerce platform |
| CN103514255A (en) * | 2013-07-11 | 2014-01-15 | 江苏谐云智能科技有限公司 | Method for collaborative filtering recommendation based on item level types |
| CN103593417A (en) * | 2013-10-25 | 2014-02-19 | 安徽教育网络出版有限公司 | Collaborative filtering recommendation method based on association rule prediction |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7881984B2 (en) * | 2007-03-30 | 2011-02-01 | Amazon Technologies, Inc. | Service for providing item recommendations |
-
2014
- 2014-05-09 CN CN201410197090.8A patent/CN105096149B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853463A (en) * | 2009-03-30 | 2010-10-06 | 北京邮电大学 | Collaborative filtering recommending method and system based on client characteristics |
| CN102663626A (en) * | 2012-04-26 | 2012-09-12 | 焦点科技股份有限公司 | Collaborative filtering recommended method based on provincial characteristics |
| CN103514255A (en) * | 2013-07-11 | 2014-01-15 | 江苏谐云智能科技有限公司 | Method for collaborative filtering recommendation based on item level types |
| CN103473354A (en) * | 2013-09-25 | 2013-12-25 | 焦点科技股份有限公司 | Insurance recommendation system framework and insurance recommendation method based on e-commerce platform |
| CN103593417A (en) * | 2013-10-25 | 2014-02-19 | 安徽教育网络出版有限公司 | Collaborative filtering recommendation method based on association rule prediction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105096149A (en) | 2015-11-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Nitu et al. | Improvising personalized travel recommendation system with recency effects | |
| TWI690880B (en) | Recommended method and device for marketing products | |
| US10338898B2 (en) | State-specific external functionality for software developers | |
| CN105808685B (en) | Promotion information pushing method and device | |
| US9576313B2 (en) | Recommendation systems and methods using interest correlation | |
| US9146910B2 (en) | Method and system of displaying cross-website information | |
| US20160306889A1 (en) | Structured information about nodes on a social networking system | |
| CN103678518B (en) | Method and device for adjusting recommendation lists | |
| CN104008184A (en) | Method and device for pushing information | |
| US20080294622A1 (en) | Ontology based recommendation systems and methods | |
| CN106557480A (en) | Implementation method and device that inquiry is rewritten | |
| JP2010531626A (en) | Provision of content to mobile communication facilities based on contextual data and behavior data related to a part of mobile content | |
| US20130091130A1 (en) | Systems and methods that utilize preference shields as data filters | |
| US9767204B1 (en) | Category predictions identifying a search frequency | |
| KR101981136B1 (en) | Program, method and system for providing recommendation contents | |
| Wu | Music personalized recommendation system based on hybrid filtration | |
| CN109727047A (en) | A kind of method and apparatus, data recommendation method and the device of determining data correlation degree | |
| Dias et al. | Automating the extraction of static content and dynamic behaviour from e-commerce websites | |
| US10474670B1 (en) | Category predictions with browse node probabilities | |
| US20200081930A1 (en) | Entity-based search system using user engagement | |
| CN106708871A (en) | Method and device for identifying social service characteristics user | |
| CN105095357A (en) | Method and device for processing consultation data | |
| CN102959539A (en) | Method and system for item recommendation in service crossing situation | |
| CN105096149B (en) | Business tine method for analyzing product association and device | |
| US10387934B1 (en) | Method medium and system for category prediction for a changed shopping mission |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |