CN103399883B - Method and system for performing personalized recommendation according to user interest points/concerns - Google Patents
Method and system for performing personalized recommendation according to user interest points/concerns Download PDFInfo
- Publication number
- CN103399883B CN103399883B CN201310304671.2A CN201310304671A CN103399883B CN 103399883 B CN103399883 B CN 103399883B CN 201310304671 A CN201310304671 A CN 201310304671A CN 103399883 B CN103399883 B CN 103399883B
- Authority
- CN
- China
- Prior art keywords
- user
- block
- session
- focus
- personalized recommendation
- 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
Abstract
The invention provides a method and a system for performing personalized recommendation according to user interest points/concerns. The method includes: (a) integrating user behavior logs of a user on multiple product lines within a scheduled time period, (b) dividing sessions according to the logs of the user within the scheduled time period, (c) performing similar requirement behavior information integration on the user behavior logs of the divided sessions and mining the user interest points/concerns, and (d) displaying personalized recommendations aiming for the user interest points/concerns through a recommendation ranking model according to the mined user interest points/concerns.
Description
Technical field
The present invention relates to network service, more particularly, it is related to a kind of carry out personalized recommendation according to user interest point
Method and system.
Background technology
With the development of electronic information technology, network has changed the life style of people.For example, people can utilize
Network Capture oneself books interested, film, music, commodity etc., therefore Netowrk tape gives the life of people's efficient quick.People
Be accustomed to the equipment using computer, mobile phone etc. with function of surfing the Net, by browsing oneself webpage interested
Practise, entertain, doing shopping etc. and to meet itself multi-faceted demand.
Typically, user inputs keyword by the search engine in network and finds webpage interested.But pass through this
The search of kind of search engine, can present to the very many webpages of user, therefore to user find oneself desired content bring tired
Difficult.
Content of the invention
The present invention seeks to realizing from user the behavior record on multiple product lines, the concern demand point of digging user is come
Make up the openness of single product line user behavior, be fully understood from from multiple dimensions and portray user, thus providing the user with
Preferably personalized search and recommendation service.
It is an object of the invention to provide a kind of method and system carrying out personalized recommendation according to user interest point.
According to an aspect of the present invention, there is provided a kind of side carrying out personalized recommendation according to user interest point/focus
Method, the method includes:A () integrates User action log within a predetermined period of time on multiple product lines for the user;B () is directed to
User's user behaviors log within a predetermined period of time conversates division;C () carries out same to the User action log of the session dividing
Class demand action information integration and digging user focus/point of interest;D () is according to the user interest point/focus excavated
Show the personalized recommendation for user interest point/focus by recommending order models.
Step (b) may include following sub-step:Carry out block division for user's user behaviors log within a predetermined period of time;Logical
Cross the session judging that using grader the set of blocks dividing is located.
Described piece of division can be based at least one of following rule and execute:Rule 1:The time identical Consecutive Days
Will is classified as same piece;Rule 2:If continuous 2 logging time interval is more than certain time, it is classified as different blocks;Rule
Then 3:If the product line of continuous daily record is specific product line, it is classified as same piece;Rule 4:In contrast current log
Whether the text message extracting is identical with the text message extracting in a upper daily record, is if the same classified as same piece, no
It is then the beginning of a new block.
Judge that the step of the session that the set of blocks dividing is located may include:By traveling through the set of blocks dividing, for given
Current block, it is first determined whether belong in a session with a upper block;If it is, this block is included into previous piece of place
Session;Otherwise utilize Backtracking Strategy, whether have in all pieces that judgement looked forward in the scheduled time and belong to one with current block
Current block, if it has, being then included into current block in the session that the block finding is located, is otherwise included into new session by the block of individual session
In.
Step (c) may include following sub-step:Calculate target data by extracting the specific fields in the session dividing;
Calculate the theme of each target using subject classification technology, and each subject classification identical target is grouped together with shape
Become multiple goal sets;Each goal set is carried out with participle to set up term co-occurrence figure;Known by application drawing clustering algorithm
Other user interest point/focus.
Show that the step of the personalized recommendation for user interest point/focus may include:When user in a search engine
When scanning for, the personalized recommendation meeting user interest point/focus is shown in search by the search word based on user input
Before result.
Show that the step of the personalized recommendation for user interest point/focus may include:When user browses webpage,
Except in remaining block of the main blocks of webpage browsing, display meets the personalized recommendation of user interest point/focus.
According to a further aspect in the invention, there is provided a kind of personalized recommendation is carried out according to user interest point/focus
System, this system includes:Integral data source module, for integrating use within a predetermined period of time on multiple product lines for the user
Family user behaviors log;Sessionizing module, conversates division for the user behaviors log within a predetermined period of time for user;Concern
Point/point of interest excavates module, for the User action log of the session dividing is carried out similar demand action information integration and
Digging user focus/point of interest;Personalized recommendation module, for passing through to push away according to the user interest point/focus excavated
Recommend order models and show the personalized recommendation for user interest point/focus.
Sessionizing module may include:Block divides submodule, for the user behaviors log within a predetermined period of time for user
Carry out block division;Sessionizing submodule, the session that the set of blocks for judge by using grader to divide is located.
Block is divided submodule and can be executed block division based at least one of following rule:Rule 1:Time is identical
Continuous daily record be classified as same piece;Rule 2:If continuous 2 logging time interval is more than certain time, it is classified as difference
Block;Rule 3:If the product line of continuous daily record is specific product line, it is classified as same piece;Rule 4:Contrast is current
Whether the text message extracting in daily record is identical with the text message extracting in a upper daily record, is if the same classified as same
Block, otherwise for the beginning of a new block.
Sessionizing submodule can judge the session that the set of blocks dividing is located by following steps:Divided by traversal
Set of blocks, for given current block, it is first determined whether and a upper block belong in a session;If it is, should
Block is included into the session at previous piece of place;Otherwise utilize Backtracking Strategy, in judging look forward in the scheduled time all pieces whether
There is the block belonging to a session with current block, if it has, being then included into current block in the session that the block finding is located, otherwise will
Current block is included in new session.
Focus/point of interest excavates module and may include:Target data calculating sub module, for by extracting the session dividing
In specific fields calculating target data;Goal set forms submodule, for calculating each using subject classification technology
The theme of target, and each subject classification identical target is grouped together to form multiple goal sets;Term co-occurrence figure
Setting up submodule, for carrying out participle to set up term co-occurrence figure to each goal set;User interest point/focus identification
Module, for by application drawing clustering algorithm come identifying user point of interest/focus.
When user scans in a search engine, user interest can will be met by the search word based on user input
The personalized recommendation of point/focus is shown in before Search Results, personalized recommendation module come to show for user interest point/
The personalized recommendation of focus.
When user browses webpage, use can be met by display in remaining block of the main blocks except the webpage browsing
The personalized recommendation of family point of interest/focus, personalized recommendation module is showing the personalization for user interest point/focus
Recommend.
The other aspect of the present invention and/or advantage will partly be illustrated in following description, some passes through to retouch
State and will be apparent from, or can learn through the enforcement of the present invention.
Brief description
By the description carrying out below in conjunction with the accompanying drawings, the above and other purpose of the present invention and feature will become more clear
Chu, wherein:
Fig. 1 is the flow process illustrating the according to embodiments of the present invention method carrying out personalized recommendation according to user interest point
Figure;
Fig. 2 is to illustrate the flow chart that session divides;
Fig. 3 is to illustrate the example that session divides;
Fig. 4 is to illustrate the User action log of the session dividing is carried out similar demand action information integration and digs
The flow chart of the process of pick user's focus/point of interest;
Fig. 5 is the frame of the according to embodiments of the present invention system carrying out personalized recommendation according to user interest point/focus
Figure.
Specific embodiment
Now, embodiments of the invention are described in detail, its example represents in the accompanying drawings, wherein, identical label table all the time
Show identical part.Below by way of embodiment being described with reference to the drawings to explain the present invention.
Fig. 1 is to illustrate the according to embodiments of the present invention method carrying out personalized recommendation according to user interest point/focus
Flow chart.
As shown in figure 1, in step S101, integrating multiple data sources, that is, integrate user making a reservation on multiple product lines
User action log (for example, user behaviors log is clicked in search) in time period.Here product line may include for example big search, patch
, know, library etc..Specifically, by being existed come organizing user according to time sequencing with user identity (ID) for keyword
User behaviors log is clicked in search on each product line.
In step S102, for user's user behaviors log within a predetermined period of time conversate (session) divide.
Specifically, session (session) is a logical meaning, and it represents a user one within certain time
It is intended to, from the point of view of the daily record behavior of user, session specifically can become to have associated one group retrieval with stipulations and click on.For example,
One user has continuously searched for " Beijing fresh flower " " BMW " " fresh flower express delivery " " benz " " Buick " " fresh flower purchase " 6 inquiries.Its
In we can see that<" Beijing fresh flower " " fresh flower express delivery " " fresh flower purchase ">With<" BMW " " benz " " Buick ">This two groups inspections
Mermis is intended in different behaviors, as two different session.Therefore, same session can be defined as same user
It is that behavior is clicked in the search meeting certain single piece of information demand within one section of continuous time.
Fig. 2 is to illustrate the flow chart that session divides.
In step S201, carry out block (block) for each user user behaviors log within a predetermined period of time and divide.Described
Block is divided and is executed based on following rule:
Rule 1:Time, identical continuous daily record was classified as same block;
Rule 2:If continuous 2 logging time interval is more than certain time(For example, 5 minutes), then it is classified as different
block;
Rule 3:If the product line of continuous daily record is for specific product line (such as " news ", " ting ", " map "),
It is classified as same block(Think in general knowledge if reading news, listening song or browsing map in someone's continuous time
Then it is considered same intention, therefore be classified as same block);
Rule 4:The text message whether phase extracting in the text message extracting in contrast current log and a upper daily record
Same, if the same it is classified as same block, otherwise for the beginning of a new block.
It should be understood that to judge whether to belong to for user journal according to the sequencing of above-mentioned rule (rule 1 is to rule 4)
In same block, first judge whether user journal meets rule 1, without meeting rule 1, then continue judgment rule 2,
By that analogy.
In step S202, to judge the session that the block set dividing is located by using grader.Due to classification
Device belongs to the prior art in machine learning, and therefore here is not repeated.
Specifically, by traveling through the block set dividing, for given current block, it is first determined whether and upper one
Individual block belongs in a session.If it is, this block is included into the session that previous block is located;No
Then utilize Backtracking Strategy, judge whether to have in all block of looking forward the scheduled time in (such as one hour) with currently
Block belongs to the block of a session, if there are the session that current block is then included into the block place found
In, otherwise current block is included in new session.More particularly, whether Backtracking Strategy is mainly judging 2 block
After belonging to same session, see the scheduled time further back block within (such as one hour) whether with currently
Block belongs to same session.Without then reopening a new session, otherwise record current block and
Block before is identical session id.
To describe session partition process below with reference to the example that Fig. 3 provides in detail.
As shown in figure 3, this user is in the time 20:19:14 carry out " Man U " inquiry, 20:21:38 carry out " Man U is live "
Inquiry, 22:01:04 carries out " position of long small pox " inquiry, 22:11:51 carry out that " how improving employee's unstability " looks into
Ask, 22:19:11 carry out " long small pox is what is eaten " inquiry, 23:02:44 carry out " how stablizing Staff of Employees " inquiry.
Divided according to above-mentioned block, can get 6 block in the middle of Fig. 3, then travel through this 6 block, for current
Block judged whether using grader and a upper block belongs to session, judge block1 with grader(Graceful
Connection)And block2(Man U is live)Belong in a session.Then consider block3 and block2, true according to grader
This two block fixed are not belonging to same session, then next back consider block3 and block1, due to this two
The time interval of block is more than the scheduled time (i.e. 1 hour), therefore block3 belongs to new session.Should traverse together
When block5, find that itself and block4 are not belonging to same session, toward the block reviewing in a hour, find with
The time interval of block4, block3 is all within a hour, therefore block5 and block4, block3 are judged,
By grader find block5 and block4 not in a session, and and block3 belong to a session together, then
Block5 is included in the session at block3 place.By that analogy, find that block6 and block4 belongs to a session together.When
After having traveled through all block, the result on the right in output Fig. 3, and the left side is not consider that the session of Backtracking Strategy divides to tie
Really.
Go back to Fig. 1, in step S103, similar demand action letter is carried out to the User action log of the session dividing
Breath is integrated and digging user focus/point of interest.
Below with reference to Fig. 4, step S103 is described in detail.
In step S401, calculate target (goal) number by extracting the useful field in the session (session) dividing
According to.
Specifically, for the feature of different data sources, because the daily record of different product lines is different, therefore by using
Different Strategies are extracted to calculate, to calculating the useful field of point of interest, the target data generating user, useful field therefore here
Can refer to inquire about (Query), title (title) etc..At Query Information meeting twice in the general inquiry field when calculating target
Reason, other are all one times of process, and this is due to more can express user interest it is considered that inquiring about, and header field (mark here
The web page title that topic is clicked on after referring to user's search) etc. the non-Query Information field simply useful supplement to inquiry.
In step S402, calculate the theme of each target using subject classification technology, and by each user's subject classification
Identical target is grouped together to form multiple goal sets.Subject classification technology is to user search queries or to click on mark
Inscribe the machine learning techniques classified.Input is an inquiry or title text, is output as this inquiry or title can divide
Some classification arriving, such as " recreational persona ", " traffic ticketing service ", " educational training " etc..This subject classification technology is machine
Routine techniques in study, therefore here are not repeated.
In step S403, each goal set is carried out with participle to set up term (term) co-occurrence figure.
Specifically, the inquiry to the user in each goal set and title carry out cutting word, are set up according to term co-occurrence
Term co-occurrence figure.The node of each term co-occurrence in figure of in figure corresponds to a term, the frequency of the weight corresponding term appearance of node
Secondary.The co-occurrence of corresponding two terms of each edge of co-occurrence figure, the frequency of the corresponding two term co-occurrences of weight of connection.For example, with
As a example inquiry " I likes dog ", after cutting word, become " I ", " liking ", " dog " these words, the figure that is set up has three nodes, often
Individual one word of node, the weight on node is exactly the frequency that these words occur, and the weight on the side of connection is exactly that two words are common
With occur number of times, such as I and like jointly occurring in that 1 time.
In step S404, by application drawing clustering algorithm come identifying user point of interest/focus.
Specifically, hierarchical clustering is condensed to term co-occurrence figure, how many individual to identify that term co-occurrence in figure has
Cluster (cluster), here it is considered that the point of interest of the corresponding user of each cluster.
Thereafter, the point of interest in each cluster result is exported respectively by next step:When extracting user interest point, first
Select the central term as this point of interest for the node of weight maximum in this cluster.Then choose connection weight and be multiplied by node weights
Term that is maximum and being connected with core node assists term as first.Finally assist art by all with central term and first
The term that language is connected all extracts and assists term as other.So the term quantity of each point of interest is uncertain, such as
Fruit is non-orphaned node, then term quantity is certain is more than or equal to 2;If isolated node, then term quantity is 1.
Afterwards, return in original Session and again check whether point of interest can export:If point of interest occurred at many days
Cross or went out more than 3 times then it is assumed that behavior is enough abundant in Dan Tian, can export, otherwise not export.
Additionally, each point of interest of output includes central term, auxiliary term etc..Each point of interest also exports the product in source
Product line, with assistance application side's deployment strategy filtering policy;Also export place name term, thus identifying the region tendency of interest simultaneously;
And export its last the date to occur and number of days always occurs, the ageing etc. of point of interest is judged with assistance application side.
Then, using the dictionary having been built up, the point of interest having exported is crossed dictionary matching module, with to this point of interest
Tag (tag).Dictionary matching strategy only considers central term and the first auxiliary term, connects this two terms first(Bag
Include forward and reverse), carry out dictionary matching.If the match is successful, the label of output matching;Otherwise carried out using central term
Join, if the match is successful, the label of output matching;Otherwise mated using the first auxiliary term, if the match is successful, exported
The label of coupling.
Now turn to Fig. 1, in step S104, pass through to recommend order models according to the user interest point/focus excavated
Show the personalized recommendation for user interest point/focus.
Above-mentioned steps S101, S102 and S103 can execute in the user terminal or execute in the server.Specifically,
User terminal can download from a server corresponding user interest point/focus and excavate software so that digging user point of interest/pass
Note point.As described above, user interest point/focus excavates software and can click on Operation Log constantly according to the inquiry of user
Ground updates user interest point/focus.
Above-mentioned steps S104 can execute on the server.Above-mentioned recommendation order models can be by existing recommendation order models
Realize, for example, existing recommendation order models can bid according to businessman, according to IP address etc. condition, Search Results be entered
Row sequence, therefore here is not described in detail.
According to embodiments of the invention, for example, when user scans in a search engine, recommendation order models can base
It is listed in meeting the website of user interest point/focus, article, video etc. before Search Results in the search word of user input
Face.Or, can show emerging according to user in remaining block of the main blocks except the webpage browsing when user browses webpage
Interesting point/focus and recommend website, article, video etc., for example, when user interest point/focus is stock, can be in user
Recommend the website related to stock, article, video etc. for user in remaining block of the webpage browsing.
Fig. 5 is the frame of the according to embodiments of the present invention system carrying out personalized recommendation according to user interest point/focus
Figure.
As shown in figure 5, being included according to the system that user interest point carries out personalized recommendation according to the present invention:Integral data
Source module 501, sessionizing module 502, focus/point of interest excavates module 503 and personalized recommendation module 504.
Integral data source module 501 is used for integrating user behavior within a predetermined period of time on multiple product lines for the user
Daily record.
Sessionizing module 502 is used for that user behaviors log within a predetermined period of time conversates division for user.
Focus/point of interest excavates module 503 and is used for carrying out similar demand to the User action log of the session dividing
Behavioural information is integrated and digging user focus/point of interest.
Personalized recommendation module 504 is used for passing through to recommend order models exhibition according to the user interest point/focus excavated
Show the personalized recommendation for user interest point/focus.
It should be understood that integral data source module 501, sessionizing module 502, focus/point of interest excavation module 503 can
By independent or integration realization in the user terminal or integration realization on the server.Personalized recommendation module 504 is implemented in
On server.
As shown in figure 5, sessionizing module 502 includes block divides submodule 5021 and sessionizing submodule 5022.
Wherein, block divides submodule 5021 for user behaviors log within a predetermined period of time carries out block for each user
(block) divide.Described block is divided and is executed based on following rule:
Rule 1:Time, identical continuous daily record was classified as same block;
Rule 2:If continuous 2 logging time interval is more than certain time(For example, 5 minutes), then it is classified as different
block;
Rule 3:If the product line of continuous daily record is for specific product line (such as " news ", " ting ", " map "),
It is classified as same block(Think in general knowledge if reading news, listening song or browsing map in someone's continuous time
Then it is considered same intention, therefore be classified as same block);
Rule 4:The text message whether phase extracting in the text message extracting in contrast current log and a upper daily record
Same, if the same it is classified as same block, otherwise for the beginning of a new block.
Sessionizing submodule 5022 is used for judging what the block set dividing was located by using grader
session.
Again as shown in figure 5, focus/point of interest excavates module 503 includes target data calculating sub module 5031, object set
Close and form submodule 5032, term co-occurrence figure setting up submodule 5033 and user interest point/focus identification submodule 5034.
Target data calculating sub module 5031 by by extract the useful field in the session (session) that divides come based on
Calculate target (goal) data.
Goal set is formed submodule 5032 and is used for being calculated the theme of each target using subject classification technology, and will be every
Individual user's subject classification identical target is grouped together to form multiple goal sets.
Term co-occurrence figure setting up submodule 5033 is used for carrying out participle to set up term co-occurrence figure to each goal set.
User interest point/focus identification submodule 5034 is used for by application drawing clustering algorithm come identifying user interest
Point/focus.
By carrying out the method and system of personalized recommendation according to user interest point/focus according to the present invention, pass through
Determine the point of interest/focus of user, thus recommending more to meet the product of user interest point/focus to user, thus carrying
The satisfaction of high user.
Although be particularly shown and described the present invention with reference to the embodiment of the present invention, those skilled in the art should
This understanding, in the case of without departing from the spirit and scope of the present invention being defined by the claims, can with it is carried out form and
Various changes in details.
Claims (8)
1. a kind of method carrying out personalized recommendation according to user interest point/focus, the method includes:
A () integrates User action log within a predetermined period of time on multiple product lines for the user, described User action log
Search for for user and click on user behaviors log;
B () conversates division for user's user behaviors log within a predetermined period of time, wherein, same session refers to user one
It is that behavior is clicked in the search meeting certain single piece of information demand in section continuous time;
(c) User action log of the session dividing is carried out similar demand action information integration and digging user focus/
Point of interest;
D () passes through to recommend order models to show for user interest point/focus according to the user interest point/focus excavated
Personalized recommendation,
Wherein, step (b) includes following sub-step:
Carry out block division for user's user behaviors log within a predetermined period of time;
To judge the session that the set of blocks dividing is located by using grader,
Wherein, judge that the step of the session that the set of blocks dividing is located includes:
By traveling through the set of blocks dividing, for given current block, it is first determined whether belonging to a session with a upper block
In;
If it is, this block is included into the session at previous piece of place;Otherwise utilize Backtracking Strategy, judge to look forward the scheduled time
Whether there is the block belonging to a session with current block, if it has, then current block to be included into the block institute finding in interior all pieces
Session in, otherwise current block is included in new session,
Wherein, step (c) includes following sub-step:
Calculate target data by extracting the specific fields in the session dividing, wherein, described specific fields include inquiry word
Section;
Calculate the theme of each target using subject classification technology, and each subject classification identical target is grouped together
To form multiple goal sets;
Each goal set is carried out with participle to set up term co-occurrence figure;
By application drawing clustering algorithm come identifying user point of interest/focus.
2. the method for claim 1, wherein described piece of division is executed based at least one of following rule:
Rule 1:Time, identical continuous daily record was classified as same piece;
Rule 2:If continuous 2 logging time interval is more than certain time, it is classified as different blocks;
Rule 3:If the product line of continuous daily record is specific product line, it is classified as same piece;
Rule 4:Whether the text message extracting in contrast current log is identical with the text message extracting in a upper daily record, such as
Really identical, it is classified as same piece, otherwise for the beginning of a new block.
3. the step the method for claim 1, wherein showing the personalized recommendation for user interest point/focus
Including:
When user scans in a search engine, the search word based on user input will meet user interest point/focus
Personalized recommendation be shown in before Search Results.
4. the step the method for claim 1, wherein showing the personalized recommendation for user interest point/focus
Including:
When user browses webpage, in remaining block except the main blocks of webpage browsing display meet user interest point/
The personalized recommendation of focus.
5. a kind of system carrying out personalized recommendation according to user interest point/focus, this system includes:
Integral data source module, for integrating User action log within a predetermined period of time on multiple product lines for the user,
Described User action log is searched for for user and is clicked on user behaviors log;
Sessionizing module, conversates division for the user behaviors log within a predetermined period of time for user, wherein, with for a moment
Words refer to that user is that behavior is clicked in the search meeting certain single piece of information demand within one section of continuous time;
Focus/point of interest excavates module, and the User action log for the session to division carries out similar demand action information
Integrate and digging user focus/point of interest;
Personalized recommendation module, for passing through to recommend order models displaying to be directed to according to the user interest point/focus excavated
The personalized recommendation of user interest point/focus, wherein, sessionizing module includes:
Block divides submodule, carries out block division for the user behaviors log within a predetermined period of time for user;
Sessionizing submodule, the session that the set of blocks for judge by using grader to divide is located,
Wherein, sessionizing submodule judges the session that the set of blocks dividing is located by following steps:
By traveling through the set of blocks dividing, for given current block, it is first determined whether belonging to a session with a upper block
In;
If it is, this block is included into the session at previous piece of place;Otherwise utilize Backtracking Strategy, judge to look forward the scheduled time
Whether there is the block belonging to a session with current block, if it has, then current block to be included into the block institute finding in interior all pieces
Session in, otherwise current block is included in new session,
Wherein, focus/point of interest excavates module and includes:
Target data calculating sub module, for calculating target data by extracting the specific fields in the session dividing, wherein,
Described specific fields include inquiring about field;
Goal set forms submodule, for calculating the theme of each target using subject classification technology, and by each theme
Classification identical target is grouped together to form multiple goal sets;
Term co-occurrence figure setting up submodule, for carrying out participle to set up term co-occurrence figure to each goal set;
User interest point/focus identification submodule, for by application drawing clustering algorithm come identifying user point of interest/concern
Point.
6. system as claimed in claim 5, wherein, block divides submodule and executes block based at least one of following rule
Divide:
Rule 1:Time, identical continuous daily record was classified as same piece;
Rule 2:If continuous 2 logging time interval is more than certain time, it is classified as different blocks;
Rule 3:If the product line of continuous daily record is specific product line, it is classified as same piece;
Rule 4:Whether the text message extracting in contrast current log is identical with the text message extracting in a upper daily record, such as
Really identical, it is classified as same piece, otherwise for the beginning of a new block.
7. system as claimed in claim 5, wherein, when user scans in a search engine, by defeated based on user
The personalized recommendation meeting user interest point/focus is shown in before Search Results by the search word entering, personalized recommendation
Module is showing the personalized recommendation for user interest point/focus.
8. system as claimed in claim 5, wherein, when user browses webpage, by the main region except the webpage browsing
In remaining block of block, display meets the personalized recommendation of user interest point/focus, and personalized recommendation module is directed to show
The personalized recommendation of user interest point/focus.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310304671.2A CN103399883B (en) | 2013-07-19 | 2013-07-19 | Method and system for performing personalized recommendation according to user interest points/concerns |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310304671.2A CN103399883B (en) | 2013-07-19 | 2013-07-19 | Method and system for performing personalized recommendation according to user interest points/concerns |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103399883A CN103399883A (en) | 2013-11-20 |
CN103399883B true CN103399883B (en) | 2017-02-08 |
Family
ID=49563513
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310304671.2A Active CN103399883B (en) | 2013-07-19 | 2013-07-19 | Method and system for performing personalized recommendation according to user interest points/concerns |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103399883B (en) |
Families Citing this family (37)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103699630B (en) * | 2013-12-20 | 2017-05-03 | 北京奇虎科技有限公司 | Method and device for recommending information in browser |
CN104750760B (en) * | 2013-12-31 | 2018-11-23 | 中国移动通信集团上海有限公司 | A kind of implementation method and device for recommending application software |
CN103870541B (en) * | 2014-02-24 | 2017-05-31 | 微梦创科网络科技(中国)有限公司 | Social network user interest digging method and system |
CN104933049B (en) | 2014-03-17 | 2019-02-19 | 华为技术有限公司 | Generate the method and system of Digital Human |
CN105556514B (en) * | 2014-06-25 | 2020-11-20 | 北京百度网讯科技有限公司 | Method and device for data mining based on user search behavior |
CN104462248B (en) * | 2014-11-19 | 2019-05-24 | 北京奇虎科技有限公司 | The method, apparatus and server of web page contents are provided |
CN105809464A (en) * | 2014-12-31 | 2016-07-27 | 中国电信股份有限公司 | Method and device for information delivery |
CN108197331B (en) * | 2015-01-21 | 2020-02-04 | 北京字节跳动科技有限公司 | User interest exploration method and device |
CN105302879B (en) * | 2015-10-12 | 2019-03-08 | 百度在线网络技术(北京)有限公司 | For determining the method and apparatus of user demand |
CN105224807B (en) * | 2015-10-14 | 2018-09-14 | 东软集团股份有限公司 | Case auditing rule extracting method and device, case checking method and system |
CN105718545A (en) * | 2016-01-18 | 2016-06-29 | 合一网络技术(北京)有限公司 | Recommendation method and device of multimedia resources |
CN105868360A (en) * | 2016-03-29 | 2016-08-17 | 乐视控股(北京)有限公司 | Content recommendation method and device based on voice recognition |
CN106021516A (en) * | 2016-05-24 | 2016-10-12 | 百度在线网络技术(北京)有限公司 | Search method and device |
CN105956204B (en) * | 2016-07-01 | 2019-08-02 | 北京奇虎科技有限公司 | The method and device of session Session satisfaction assessment |
WO2018018196A1 (en) * | 2016-07-24 | 2018-02-01 | 严映军 | Book interest determining method and determining system |
CN106293119A (en) * | 2016-07-29 | 2017-01-04 | 百度在线网络技术(北京)有限公司 | A kind of method and apparatus carrying out information recommendation in input method |
CN107817976A (en) * | 2016-09-12 | 2018-03-20 | 阿里巴巴集团控股有限公司 | Page info processing system and page generation, page info provide method and device |
CN106339502A (en) * | 2016-09-18 | 2017-01-18 | 电子科技大学 | Modeling recommendation method based on user behavior data fragmentation cluster |
CN107169014B (en) * | 2017-03-31 | 2021-03-19 | 百度在线网络技术(北京)有限公司 | POI recommendation method, device, equipment and computer readable storage medium |
CN107220312B (en) * | 2017-05-12 | 2020-08-14 | 华中科技大学 | Interest point recommendation method and system based on co-occurrence graph |
CN107688600B (en) * | 2017-07-12 | 2021-06-25 | 百度在线网络技术(北京)有限公司 | Knowledge point mining method and device |
CN107679079A (en) * | 2017-08-29 | 2018-02-09 | 重庆邮电大学 | Method is recommended in personalized application based on cellphone subscriber's interest threedimensional model |
CN110022484A (en) * | 2018-01-08 | 2019-07-16 | 武汉斗鱼网络科技有限公司 | A kind of methods of exhibiting, device, server and medium |
CN108335147B (en) * | 2018-02-28 | 2021-08-13 | 北京搜狐新媒体信息技术有限公司 | Data analysis method and system based on user behaviors |
CN108427756B (en) * | 2018-03-16 | 2021-02-12 | 中国人民解放军国防科技大学 | Personalized query word completion recommendation method and device based on same-class user model |
CN108345698B (en) * | 2018-03-22 | 2022-03-11 | 北京百度网讯科技有限公司 | Method and device for mining attention points of articles |
CN109214923A (en) * | 2018-08-02 | 2019-01-15 | 阿里巴巴集团控股有限公司 | A kind of financing task creating method and device |
CN109902706B (en) * | 2018-11-09 | 2023-08-22 | 华为技术有限公司 | Recommendation method and device |
CN111475734A (en) * | 2018-12-25 | 2020-07-31 | 杭州铭智云教育科技有限公司 | Method for pushing target document directory |
CN110209937A (en) * | 2019-05-31 | 2019-09-06 | 三角兽(北京)科技有限公司 | Information-pushing method and device based on user behavior |
CN112148960B (en) * | 2019-06-27 | 2024-03-22 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for determining category of attention point |
CN113111251A (en) * | 2020-01-10 | 2021-07-13 | 阿里巴巴集团控股有限公司 | Project recommendation method, device and system |
CN111949865A (en) * | 2020-08-10 | 2020-11-17 | 杭州电子科技大学 | Interest point recommendation method based on graph neural network and user long-term and short-term preference |
CN112765400B (en) * | 2020-12-31 | 2024-04-23 | 上海众源网络有限公司 | Weight updating method, content recommending method, device and equipment for interest labels |
CN113761383B (en) * | 2021-10-13 | 2023-06-20 | 中国人民大学 | Search and recommendation fusion system based on unified user behavior modeling |
CN114818643B (en) * | 2022-06-21 | 2022-10-04 | 北京必示科技有限公司 | Log template extraction method and device for reserving specific service information |
CN114840486B (en) * | 2022-06-28 | 2022-09-16 | 广州趣米网络科技有限公司 | User behavior data acquisition method and system and cloud platform |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6668278B1 (en) * | 1996-12-20 | 2003-12-23 | Liberate Technologies | Information retrieval system using an internet multiplexer to focus user selection |
CN102982042A (en) * | 2011-09-07 | 2013-03-20 | 中国移动通信集团公司 | Personalization content recommendation method and platform and system |
-
2013
- 2013-07-19 CN CN201310304671.2A patent/CN103399883B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6668278B1 (en) * | 1996-12-20 | 2003-12-23 | Liberate Technologies | Information retrieval system using an internet multiplexer to focus user selection |
CN102982042A (en) * | 2011-09-07 | 2013-03-20 | 中国移动通信集团公司 | Personalization content recommendation method and platform and system |
Non-Patent Citations (6)
Title |
---|
Web使用挖掘的个性化推荐系统研究;苏欣;《中国优秀硕士学位论文全文数据库·信息科技辑》;20121015(第10期);参见3.2-3.4节,4.2节 * |
WEB使用挖掘研究;黄浩 等;《计算机系统应用》;20080131(第1期);全文 * |
Web日志预处理中优化的会话识别方法;方元康 等;《计算机工程》;20090430;第35卷(第7期);全文 * |
图聚类算法的评价模型;梅娟 等;《计算机工程与应用》;20121231(第14期);全文 * |
基于词共现图的中文微博新闻话题识别;赵文清 等;《智能系统学报》;20121031;第7卷(第5期);参见摘要,1.2-1.4节 * |
数据挖掘技术在网站个性化推荐中的应用;彭建光;《中国优秀硕士学位论文全文数据库·信息科技辑》;20071015(第04期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN103399883A (en) | 2013-11-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103399883B (en) | Method and system for performing personalized recommendation according to user interest points/concerns | |
Ingwersen et al. | Smart city research 1990–2016 | |
CN103491205B (en) | The method for pushing of a kind of correlated resources address based on video search and device | |
CN104063383B (en) | Information recommendation method and device | |
US8650198B2 (en) | Systems and methods for facilitating the gathering of open source intelligence | |
CN103823844B (en) | Question forwarding system and question forwarding method on the basis of subjective and objective context and in community question-and-answer service | |
CN103116657B (en) | A kind of individuation search method of network teaching resource | |
CN102968465B (en) | Network information service platform and the search service method based on this platform thereof | |
CN107730343A (en) | A kind of user's merchandise news method for pushing and equipment based on picture attribute extraction | |
CN104111941B (en) | The method and apparatus that information is shown | |
CN103577549A (en) | Crowd portrayal system and method based on microblog label | |
CN105677780A (en) | Scalable user intent mining method and system thereof | |
CN103914492A (en) | Method for query term fusion, method for commodity information publish and method and system for searching | |
CN102663022B (en) | Classification recognition method based on URL (uniform resource locator) | |
CN103390044B (en) | Method and device for identifying linkage type POI (Point Of Interest) data | |
CN105794154A (en) | System and method for analyzing and transmitting social communication data | |
CN103955529A (en) | Internet information searching and aggregating presentation method | |
CN104834695A (en) | Activity recommendation method based on user interest degree and geographic position | |
CN103593474A (en) | Image retrieval ranking method based on deep learning | |
CN103902697A (en) | Combinatorial search method, client and server | |
Vysotsky et al. | Online tourism system for proposals formation to user based on data integration from various sources | |
CN103023714A (en) | Activeness and cluster structure analyzing system and method based on network topics | |
CN103399855B (en) | Behavior intention determining method and device based on multiple data sources | |
CN103365904A (en) | Advertising information searching method and system | |
CN104516961A (en) | Topic digging and topic trend analysis method and system based on region |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant |