CN108389066A - Content distribution method based on Dynamic Programming and system - Google Patents
Content distribution method based on Dynamic Programming and system Download PDFInfo
- Publication number
- CN108389066A CN108389066A CN201710063920.1A CN201710063920A CN108389066A CN 108389066 A CN108389066 A CN 108389066A CN 201710063920 A CN201710063920 A CN 201710063920A CN 108389066 A CN108389066 A CN 108389066A
- Authority
- CN
- China
- Prior art keywords
- content
- user group
- preference
- allocated
- label
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Transfer Between Computers (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention provides a kind of content distribution method and systems.This method includes：For each content to be allocated, content tab and respective confidence are determined；For each user group, user group label and respective confidence are determined；According to the user group label of each user group, the relevance with the content tab of each content to be allocated and respective respective confidence, determine preference of each user group to each content to be allocated；And according to the preference, content is distributed to each user group so that total preference reaches maximum.
Description
Technical field
The present invention relates to content push technologies, and in particular, to a kind of content distribution method based on Dynamic Programming and is
System.
Background technology
Personalized content has become the important competitiveness of internet product with service at present.Either music, news production
The commending contents of product or the operation of electric business website, all can't do without the application of personalized recommendation technology.In the operation of electric business website
Field has several different content (for example, activity, commodity, advertisement etc.) can be with for being illustrated in the resource-niche of user at the moment
It selects.The method of generally use is to show the highest content of quality to full dose user.Although this method most directly, most general do
Method, and overall effect is fine, however this method does not account for user's feature, cannot be targetedly that user's displaying is most closed
Suitable content, and the waste of a large amount of not selected contents can be caused.It therefore, should be according to their inclined for different users
It is good, it distributes and is most suitable for their content, so that whole bandwagon effect is (for example, clicking rate/conversion ratio/introducing order amount of money
Etc. indexs) reach best.
It is targetedly that user shows appropriate content to realize, another method is that user is divided into several users
Group, and fancy grade score of each user group to each content is assessed, then content is distributed to each user group by hand by operation.
Although this method can realize the purpose for targetedly distributing content for user group, it consumes operation manpower, often
It is required that operation is quite familiar with the division of each content and user group.In the situation that inner capacities is big, user group is more, renewal frequency is high
Under, this method is less efficient.In addition, there is also the method that another can be targetedly user's distribution content, institute
Stating method introduces heuritic approach to carry out reasonable distribution to content so that can handle a large amount of contents and user group.However, opening
Hairdo algorithm is only capable of finding locally optimal solution, and non-optimal allocation plan, is unable to reach optimal distribution effects.
Therefore, it is necessary to a kind of content distribution method and system based on Dynamic Programming, enabling so that whole displaying effect
Fruit reaches best.
Invention content
It is to solve at least the above and/or disadvantage in terms of the disclosure and provides at least the following advantages.
According to the first aspect of the invention, a kind of content distribution method is provided, including：For each content to be allocated,
Determine content tab and respective confidence；For each user group, user group label and respective confidence are determined；According to each use
Relevance and respective respective confidence of the user group label of family group with the content tab of each content to be allocated, determine each
Preference of the user group to each content to be allocated；And according to the preference, content is distributed to each user group so that total inclined
Good degree reaches maximum.
Preferably, the determining content tab and respective confidence include：Contents attribute is determined as content tab.
Preferably, the contents attribute includes being in commodity set, Taxonomy Information, store information and brand message
It is one or more.
Preferably, the method further includes：Content tab and respective confidence with the content is stored as hive tables.
Preferably, the method further includes the maximum allocated user group quantity that the extraction content allows.
Preferably, the determining user group label and respective confidence include：It is recorded according to the historical behavior of user group, really
Determine user group label.
Preferably, the method further includes：User group label and respective confidence with the user group is stored as
Hive tables.
Preferably, the user group label of each user group of the basis and the content tab of each content to be allocated are associated with
Property and respective respective confidence include to the preference of each content to be allocated to determine each user group：It is waited for point if described
One or more content tabs with content are associated with one or more user group labels of the user group, then distinguish
For each associated content tab and user group set of tags, the respective confidence of the content tab and the user are calculated
The product of the respective confidence of group's label, and each associated content tab will be directed to and user group set of tags is calculated multiplies
Product is added summation, will be described and be determined as basic preference score of the user group to the content；And if described wait for point
Any content tab with content is not associated with any user group's label of the user group, then by the user group to described
The basic preference score of content is denoted as 0；And based on the basic preference score, determine each user group to each to be allocated interior
The preference of appearance.
Preferably, the method further includes being stored as each user group to the basic preference score of each content to be allocated
Hive tables.
Preferably, the method further includes assessing the quality point of the content, wherein the use of each user group of the basis
The relevance and respective respective confidence of family group's label and the content tab of each content to be allocated determines each user group
Preference to each content to be allocated includes：The product of basic preference score and the quality point of the content is calculated, it will be described
Product is determined as final preference score of the user group to the content；Based on the final preference score, each use is determined
Preference of the family group to each content to be allocated.
Preferably, described to distribute content to each user group according to the preference so that total preference reaches maximum can be with
Including：Acquisition can distributing user group's quantity for each content；Relative to each user group in each user group, it is based on
It is acquired can distributing user group's quantity, traversal for the user group each content assignment scheme and determine in each
The total preference for holding allocation plan is directed to the user group to be determined and stored in the case of total preference reaches maximum
Content assignment scheme；Each user group is reversed；And each user group of the poll through reversing, and it is directed to each user
Group obtains the corresponding content assignment scheme stored, to distribute content to each user group according to preference so that total
Preference reaches maximum.
According to the second aspect of the invention, a kind of content distribution device is provided, may include：Content tab module, matches
It is set to：For each content to be allocated, content tab and respective confidence are determined；User group label model, is configured to：For every
A user group determines user group label and respective confidence；Preference computing module, is configured to：According to user group label model
The content tab of the user group label of identified each user group and each content to be allocated determined by content tab module
Relevance and respective respective confidence, determine preference of each user group to each content to be allocated；And content point
With module, it is configured to：According to the preference determined by preference computing module, content is distributed to each user group so that total preference
Degree reaches maximum.
Preferably, the preference computing module is further configured to：If the content to be allocated is one or more
A content tab is associated with one or more user group labels of the user group, then is directed to respectively each associated interior
Hold label and user group label, calculates the respective confidence of the content tab and the respective confidence of the user group label
Product, and each associated content tab will be directed to and summed with the calculated product addition of user group label, by described in and really
It is set for the basic preference score to the content for the user group；And if any content tab of the content to be allocated not
It is associated with any user group's label of the user group, then the user group is denoted as the basic preference score of the content
0；And based on the basic preference score, determine preference of each user group to each content to be allocated.
Preferably, the content distribution module is further configured to：Acquisition can distributing user group's number for each content
Amount；Relative to each user group in each user group, based on it is acquired can distributing user group's quantity, traversal is for described
Each content assignment scheme of user group simultaneously determines the total preference for being directed to each content assignment scheme, to be determined and stored in
The content assignment scheme for the user group in the case of always preference reaches maximum；Each user group is reversed；With
And each user group of the poll through reversing, and the corresponding content assignment side stored is obtained for each user group
Case, to distribute content to each user group according to preference so that total preference reaches maximum.
According to the third aspect of the invention we, a kind of content allocation system is provided, including：Memory is configured to storage needle
Content tab and respective confidence to content to be allocated and the user group label and respective confidence for each user group；
And processor, it is connected with memory via wired or wireless way, and be configured to：For each content to be allocated, determine in
Hold label and respective confidence；For each user group, user group label and respective confidence are determined；According to each user group
Relevance and respective respective confidence of the user group label with the content tab of each content to be allocated, determine each user group
To the preference of each content to be allocated；And according to the preference, content is distributed to each user group so that total preference reaches
To maximum.
Description of the drawings
Below in conjunction with attached drawing, above and other aspect, feature and the advantage of the example embodiment of the disclosure will be become apparent from,
In attached drawing：
Fig. 1 shows the block diagram of the exemplary hardware arrangement of content allocation system according to example embodiment of the present invention.
Fig. 2 shows the operation flow sheets of content distribution method according to example embodiment of the present invention.
Fig. 3 shows the flow chart of content distribution method according to example embodiment of the present invention.
Fig. 4 shows the specific calculation that content assignment scheme is determined for each user group according to example embodiment of the present invention
The flow chart of method.
Fig. 5 shows the block diagram of content distribution device according to example embodiment of the present invention.
Specific implementation mode
The example that the present invention is described below with reference to attached drawing is implemented.The present invention provides a kind of contents based on Dynamic Programming
Distribution method and system, the content distribution method and system can distribute for user in most suitable one from several contents
Hold, obtains the distribution effects of theoretically global optimum.
Fig. 1 shows the block diagram of the exemplary hardware arrangement 100 of content allocation system according to example embodiment of the present invention.
Fig. 1 is the block diagram shown according to the exemplary hardware of the content allocation system of embodiment of the present disclosure arrangement 100.Hardware
Arrangement 100 includes memory 110 and processor 120.
Memory 110 may include having non-volatile or form of volatile memory memory, e.g. electrically erasable
Except programmable read only memory (EEPROM), flash memory, and/or hard disk drive.Memory 110 is configurable to store to be directed to and wait for
The information of content and each user group is distributed, for example, content tab for content to be allocated and respective confidence and being directed to
The user group label of each user group and respective confidence etc..
In addition, memory 110 can also include computer program 111, which includes code/computer
Readable instruction makes hardware layout 100 and/or including hardware layout 100 when being executed by the processor 120 in arrangement 100
Equipment inside can execute the flow of content distribution method for example described in the invention and its any deformation.In addition, calculating
Machine program 111 can be configured with the computer program code of such as computer program module 111A~111C frameworks.
Processor 120 (for example, microprocessor, digital signal processor (DSP) etc.).Processor 120 can be for holding
Single treatment unit either multiple processing units of the different actions of row flow described herein.Processor 120 passes through load
One or more instruction codes on memory determine content tab and respective confidence to be directed to each content to be allocated；
For each user group, user group label and respective confidence are determined；According to the user group label of each user group with each wait for
The relevance of the content tab of content and respective respective confidence are distributed, determines each user group to each content to be allocated
Preference；And according to the preference, content is distributed to each user group so that total preference reaches maximum.
Processor 120 can be single cpu (central processing unit), but can also include that two or more processing are single
Member.For example, processor 120 may include general purpose microprocessor, instruction set processor and/or related chip group and/or special micro-
Processor (for example, application-specific integrated circuit (ASIC)).Processor can also include the onboard storage device for caching purposes.It calculates
Machine program can be carried by being connected to the computer program product of processor.Computer program product may include storing thereon
There is the computer-readable medium of computer program.For example, computer program product can be flash memory, random access memory
(RAM), read-only memory (ROM), EEPROM, and above computer program module can use depositing in UE in an alternative embodiment
The form of reservoir is distributed in different computer program products.
In addition, the arrangement 100 can also include input unit 102, the Yi Jiyong for receiving signal from other entities
In the output unit 104 for providing signal to other entities.Input unit 102 and output unit 104 can be arranged to single reality
The entity that body either detaches.In an exemplary embodiment of the invention, input unit and output unit can be achieved aobvious to touch
Show that device, content allocation system show the order of webpage in response to being had input by touch display, executes according to the present invention shows
The content distribution method of example embodiment.After targetedly distributing content to each user, pass through the touch display
Output.
Although being implemented as computer program module above in conjunction with the code means in Fig. 1 the disclosed embodiments,
Hardware layout 100 is made to execute the operation of content distribution method according to the present invention when being executed in processor 120, however alternative
In embodiment, at least one in the code means can at least be implemented partly as hardware circuit.
Fig. 2 shows the operation flow sheets of content distribution method according to example embodiment of the present invention.Specifically, described
Content distribution method includes four parts, and first in operation S1, mark is carried out to content.Secondly operation S3, to user group into
Row mark.Then preference of each user group to each content is calculated in S5.Finally in S7, according to content assignment logic, needle
Content is distributed to each user group so that global total preference realizes highest.That is, to have fully to content to be allocated
Understanding, be content it is tagged.In conjunction with user group label, preference of the user group to each content is calculated.Then
Using the dynamic programming method in operational research, content the most matched is distributed for each user group.Alternatively, this method can be with
Including recruitment evaluation step (S9), to compare this method and former scheme, Evaluated effect.
Below with reference to the content distribution method of Fig. 3 detailed descriptions according to example embodiment of the present invention.For example, the present invention carries
The technical solution of confession can be based on hive databases and hadoop frames, and by using redis external caches, to calculate distribution
As a result, and in memory by result preservation, being used for service end interface.Hadoop is a kind of distributed system architecture.
User can develop distributed program in the case where not knowing about distributed low-level details.The power of cluster is made full use of to carry out
High-speed computation and storage.Hive databases are a data base tools based on Hadoop, can be by the data file of structuring
Be mapped as a database table, and simple sql query functions be provided, sql sentences can be converted to MapReduce tasks into
Row operation.Its advantage is that learning cost is low, simple MapReduce statistics can be fast implemented by class SQL statement, it is not necessary to open
The MapReduce applications for sending out special, are very suitable for the statistical analysis of database.Redis is a use ANSI C language increased income
Speech writes, support network, can based on memory also can persistence log type, Key-Value databases, and provide multilingual
API.It should be noted that implementation above frame is merely exemplary, the invention is not limited thereto.
Fig. 3 shows the flow chart of content distribution method 300 according to example embodiment of the present invention.
Content distribution method according to example embodiment of the present invention, in S301, for each content to be allocated, determine in
Hold label and respective confidence.In one embodiment, content mark can be generated by the contents attribute of conclusion content to be allocated
Label.In electric business website operation, content to be allocated is usually advertising campaign, exclusive commodity etc., middle from the discussion above can be extracted
Go out the contents attribute such as commodity set, Taxonomy Information, store information, brand message, as content tab.In one
It includes multiple labels to hold, and a label can correspond to multiple contents.The respective confidence of content tab can indicate institute
The correlation degree of content and corresponding contents label is stated, for example, can be indicated by 0 or 1, i.e., if the content is with described
Its confidence level can be then considered as 1 by content tab, its confidence level is otherwise considered as zero.It should be noted that can also come by other means
Confidence level is indicated, to represent the degree of correlation of the content and the content tab.Furthermore, it is possible to by in the content
Hold label and respective confidence storage in memory, such as be stored as hive tables, for example, shown in the following table 1：
1 content tab hive literary name sections of table
In addition, except the content tab and respective confidence for extracting content to be allocated, it can also determine and be tolerated in described
Perhaps maximum allocated user group quantity, and can equally store it in memory, such as it is stored as hive as shown in Table 2
Table：
The maximum allocated user group number hive literary name sections of 2 content of table
It on the other hand, can also be to content quality except the content tab and respective confidence for extracting content to be allocated
Make assessment, that is, assess the quality point of the content.For example, by inquiring the history sales volume of commodity, movable history shows,
Score is formed with the mode of logistic regression.The score can also store in memory, such as be stored as hive shown in table 3
Table：
3 content quality hive literary name sections of table
In S303 user group label and respective confidence are determined for each user group.The respective confidence of user tag
It is similar with the respective confidence of content tab, the correlation degree for indicating the user group and corresponding user group label.For example,
It can be recorded according to the historical behavior of user group, determine user group label and respective confidence, wherein user group label can be with
Content tab is associated, and its incidence relation can store in memory.The incidence relation may include it is identical or
Cover.For example, if user tag is fresh and the content tab of a content includes fruit, then it is assumed that the user tag
It is associated with content tab.An implementation according to the present invention, on the basis of existing subscriber's group's label, for new
The content tab of increasing improves content respective labels for user group.It should be clear that a user group can include multiple labels, and one
A label can correspond to multiple user groups.Furthermore, it is possible in memory by user group label and respective confidence storage, such as
Hive tables are stored as, as shown in table 4：
4 user group label hive literary name sections of table
Then, in S305, according to the pass of the user group label of each user group and the content tab of each content to be allocated
Connection property and respective respective confidence, determine preference of each user group to each content to be allocated.Specifically, if it is described
One or more content tabs of content to be allocated are associated with one or more user group labels of the user group, then
Be directed to each associated content tab and user group set of tags respectively, calculate the respective confidence of the content tab with it is described
The product of the respective confidence of user group label, and will be calculated with user group set of tags for each associated content tab
Product addition summation, will be described and be determined as basic preference score of the user group to the content.If described wait for
Any content tab for distributing content is not associated with any user group's label of the user group, then by the user group to institute
The basic preference score for stating content is denoted as 0.It is then based on the basic preference score, determines each user group to each to be allocated
The preference of content.In addition, after basic preference score is calculated, basic preference score and the content can also be calculated
Quality point product, and the product is determined as final preference score of the user group to the content.Finally, it is based on
The final preference score determines preference of each user group to each content to be allocated.Furthermore, it is possible to by each user group
Hive tables are equally stored as to the basic preference score or final preference score of each content to be allocated, as shown in table 5：
5 user group of table is to content-preference score literary name section
In S307, according to the preference, content is distributed to each user group so that total preference reaches maximum.For example, being
Guarantee finally obtains maximum total preference in the overall situation, and allocation algorithm can design as follows：
First, it is assumed that contents list is N and user's group-list is M, content maximum allocated user group number is obtained from table 2,
For example, the maximum allocated user group number of content j is c (j).
Then, preference score matrix is obtained from table 5, for example, user group i can be expressed as w to the preference score of content j
(i, contentj)。
Then, define n dimension matrix f (g1, g2 ... gn), indicate to g1 user group of the 1st content assignment, to the 2nd
In the case that g2 user group ... of a content assignment is to gn user group of n-th of content assignment, the highest that can be obtained is always inclined
Good grades.
So, for i-th of user group, the be up to different allocation plan of n+1 kinds：Any content, distribution are not distributed
Content 1, distribution content 2 ... distribution content n, that is, obtain derivation formula：
F (g1, g2 ... gn)=max (f (g1, g2 ... gn), f (g1-1, g2 ... gn)+w (i, content1), f (g1, g2-
1...gn)+w (i, content2) ... f (g1, g2 ... gn-1)+w (i, contentn)) by above formula be assured that how to
Family group's i distribution content, which just can guarantee, realizes global highest preference score.
In addition, defining s (g1, g2 ... gn) matrix for each user group.For any user group i, s (g1, g2 ...
Gn it) indicates by g1 user group of first content assignment, by g2 user group ... of second content assignment by n-th of content
In the case of distributing gn user group, which kind of allocation strategy user group i should take.That is, the value of s (g1, g2 ... gn)
It indicates to calculate which kind of strategy f (g1, g2 ... gn) uses in above-mentioned derivation：When user group is unassigned, that is, above-mentioned
Max gets first term, note 0 in f derivation formulas；When user group is assigned to content 1, i.e., max gets second in above-mentioned f derivations formula
, note 1；When user group is assigned to content j, i.e., max gets jth+1 in above-mentioned f derivations formula, remembers j.S matrixes should be protected
There are in external cache.
Backward traverse user group-list in this way takes s matrixes to retrodict each user group, that is to say, that is ensuring global highest
Under the premise of total preference score, determine that the allocation strategy of each user group, code can be expressed as：
It is described to determine that the specific algorithm of content assignment scheme includes for each user group：Obtain dividing for each content
With user group quantity；Relative to each user group in each user group, based on it is acquired can distributing user group's quantity, time
It goes through each content assignment scheme for the user group and determines total preference for each content assignment scheme, so as to true
Determine and is stored in the content assignment scheme for the user group in the case of total preference reaches maximum；Each user group is inverse
Sequence arranges；And each user group of the poll through reversing, and obtained for each user group stored it is corresponding interior
Hold allocation plan, to distribute content to each user group according to preference so that total preference reaches maximum.Fig. 4 shows wheel
Ask the flow chart of the algorithm through each user group reversed.In step S401, the assignable user of each content is obtained first
Group's quantity, that is, c (1), c (2) ... c (n).Secondly, in step S403, user's group-list is reversed.Then, in step
S405, determines whether this flow has traversed all user groups.If having stepped through whole user groups, then it represents that completed content
Distribution, therefore, terminates distribution content and method according to the present invention.If not traversing whole user groups, S407 is thened follow the steps.
In step S407, next user group is searched from user's group-list.In step S409, is obtained from external cache and be directed to the user group
S matrixes.In step S411, the value of output s (c (1), c (2) ... c (n)), it is, output is for the interior of user group distribution
Hold number j.In step S413, when the context number j is not equal to 0, that is, will be corresponding interior when distributing content j to the user group
Hold can distributing user group's quantity subtract together return to operation S405.
Algorithm above only seeks a kind of way of example of global optimum, and those skilled in the art can design other schemes
To solve.
It is also possible to be embodied as content distribution device.Specifically, the content distribution device 500 can wrap
It includes：Content tab module 510, is configured to：For each content to be allocated, content tab and respective confidence are determined；User group
Label model 520, is configured to：For each user group, user group label and respective confidence are determined；Preference computing module
530, it is configured to：According to the user group label of each user group determined by user group label model 520 and content tab module
The relevance and respective respective confidence of the content tab of each content to be allocated determined by 510, determine each user group
To the preference of each content to be allocated；And content distribution module 540, it is configured to：According to true by preference computing module 530
Fixed preference distributes content to each user group so that total preference reaches maximum.
In one embodiment, the preference computing module 530 is further configured to：If the content to be allocated
One or more content tabs are associated with one or more user group labels of the user group, then are directed to respectively each
Associated content tab and user group label, calculate the phase of the respective confidence and the user group label of the content tab
The product of confidence level is answered, and each associated content tab will be directed to and summed with the calculated product addition of user group label,
It will be described and be determined as basic preference score of the user group to the content；And if the content to be allocated any
Content tab is not associated with any user group's label of the user group, then the user group is inclined to the basis of the content
Good grades are denoted as 0；And based on the basic preference score, determine preference of each user group to each content to be allocated.
In addition, in another embodiment, the content distribution module 540 is further configured to：It obtains and is directed to each content
Can distributing user group's quantity；It, can distributing user group based on acquired relative to each user group in each user group
Quantity, each content assignment scheme and determining total preference for each content assignment scheme of traversal for the user group
Degree, to be determined and stored in the content assignment scheme for the user group in the case of total preference reaches maximum；It will
Each user group reverses；And each user group of the poll through reversing, and for each user group obtain stored with
Corresponding content assignment scheme, to distribute content to each user group according to preference so that total preference reaches maximum.
To sum up, a kind of content identification method and system based on Dynamic Programming are described, the method can be based on big number
Carry out a large amount of labels of process content and user group according to technological frame, to more accurately evaluate user group to the inclined of each content
Good degree, and under the constraint that each content need to distribute different number user group, the allocation strategy of global optimum can be calculated,
Promote whole personalised effects.
For the technology that the disclosure proposes, such as assessment can be executed to its effect by ABTest on execution line.Example
Such as, the user in each user group can be randomly divided into two groups of AB.To each user group, A groups press the side that this programme provides
Method distributes content, and heuristically method distributes content to B groups, and continuance test one week observes feedback data, compares personalised effects.
According to test, the preference that the technology of the disclosure can more accurately evaluate user group to each content is found, promoted complete
Total preference score of office, to realize optimum allocation.
It should be noted that above scheme is only to show a specific implementation of present inventive concept, the present invention is not limited to above-mentioned
Implementation.The part processing in above-mentioned implementation is can be omitted or skips, without departing from the spirit and scope of the present invention.
The method of front can be realized in the form of the program command that can be held by a variety of computer installations and be recorded in calculating
In machine readable medium recording program performing.In this case, computer readable recording medium storing program for performing may include individual program command, data text
Part, data structure or combinations thereof.Meanwhile the program command recorded in the recording medium specially can not count or be configured to this hair
Bright or computer software fields technical staff's known applications.Computer readable recording medium storing program for performing includes such as hard disk, floppy disk
Or the magnetic mediums such as tape, the optical medium such as compact disk read-only memory (CD-ROM) or digital versatile disc (DVD), such as
The magnet-optical medium of floptical disk and the hardware device such as storing and executing ROM, RAM of program command, flash memory.In addition, journey
Sequence order includes the high-level language that the machine language code that compiler is formed and computer can perform by using interpretive program.Before
The hardware device in face can be configured to be operated as at least one software module to execute the operation of the present invention, and contrary operation
It is also the same.
Although the operation of context of methods has shown and described with particular order, the operation of each method can be changed
Sequentially so that specific operation can be executed with reverse order or allow to execute spy simultaneously with other operations at least partly
Fixed operation.Additionally, this invention is not limited to the above example embodiments, it can be in the premise for not departing from spirit and scope of the present disclosure
Under, including one or more other components or operation, or omit one or more other components or operation.
The preferred embodiment of the present invention is had been combined above and shows the present invention, but those skilled in the art will manage
Solution, without departing from the spirit and scope of the present invention, can carry out various modifications the present invention, replaces and change.Cause
This, the present invention should not be limited by above-described embodiment, and should be limited by appended claims and its equivalent.
Claims (11)
1. a kind of content distribution method, including：
For each content to be allocated, content tab and respective confidence are determined；
For each user group, user group label and respective confidence are determined；
According to the user group label of each user group and the relevance of the content tab of each content to be allocated and respective corresponding
Confidence level determines preference of each user group to each content to be allocated；And
According to the preference, content is distributed to each user group so that total preference reaches maximum.
2. according to the method described in claim 1, the wherein described determining content tab and respective confidence include：By contents attribute
It is determined as content tab.
3. according to the method described in claim 1, further including：Determine the maximum allocated user group quantity that the content allows.
4. according to the method described in claim 1, the wherein described determining user group label and respective confidence include：According to user
The historical behavior record of group, determines user group label.
5. according to the method described in claim 1, the user group label of the wherein described each user group of basis with it is each to be allocated
The relevance of the content tab of content and respective respective confidence determine each user group to the inclined of each content to be allocated
Degree includes well：
If one or more user groups of one or more content tabs of the content to be allocated and the user group
Label is associated, then is directed to each associated content tab and user group label respectively, calculates the corresponding of the content tab
The product of confidence level and the respective confidence of the user group label, and each associated content tab and user group will be directed to
Label calculated product addition summation, will be described and be determined as basic preference score of the user group to the content；
And
It, will if any content tab of the content to be allocated is not associated with any user group's label of the user group
The user group is denoted as 0 to the basic preference score of the content；And
Based on the basic preference score, preference of each user group to each content to be allocated is determined.
6. according to the method described in claim 5, further including：The quality point of the content is assessed,
The user group label of the wherein described each user group of basis is with each relevance of the content tab of content to be allocated and respectively
From respective confidence include to the preference of each content to be allocated to determine each user group：
The product for calculating basic preference score and the quality point of the content, is determined as the user group to described by the product
The final preference score of content；
Based on the final preference score, preference of each user group to each content to be allocated is determined.
7. according to the method described in claim 1, described distribute content according to the preference to each user group so that total preference
Degree reaches maximum and includes：
Acquisition can distributing user group's quantity for each content；
Relative to each user group in each user group, based on it is acquired can distributing user group's quantity, traversal is directed to institute
It states each content assignment scheme of user group and determines total preference for each content assignment scheme, to determine and to store
Reach the maximum content assignment scheme for the user group in total preference；
Each user group is reversed；And
Each user group of the poll through reversing, and the corresponding content assignment side stored is obtained for each user group
Case, to distribute content to each user group according to preference so that total preference reaches maximum.
8. a kind of content distribution device, including：
Content tab module, is configured to：For each content to be allocated, content tab and respective confidence are determined；
User group label model, is configured to：For each user group, user group label and respective confidence are determined；
Preference computing module, is configured to：According to the user group label of each user group determined by user group label model with
The relevance and respective respective confidence of the content tab of each content to be allocated determined by content tab module determine every
Preference of a user group to each content to be allocated；And
Content distribution module is configured to：According to the preference determined by preference computing module, content is distributed to each user group,
So that total preference reaches maximum.
9. device according to claim 8, wherein the preference computing module is further configured to：
If one or more user groups of one or more content tabs of the content to be allocated and the user group
Label is associated, then is directed to each associated content tab and user group label respectively, calculates the corresponding of the content tab
The product of confidence level and the respective confidence of the user group label, and each associated content tab and user group will be directed to
Label calculated product addition summation, will be described and be determined as basic preference score of the user group to the content；
And
It, will if any content tab of the content to be allocated is not associated with any user group's label of the user group
The user group is denoted as 0 to the basic preference score of the content；And
Based on the basic preference score, preference of each user group to each content to be allocated is determined.
10. device according to claim 8, the content distribution module is further configured to：
Acquisition can distributing user group's quantity for each content；
Relative to each user group in each user group, based on it is acquired can distributing user group's quantity, traversal is directed to institute
It states each content assignment scheme of user group and determines total preference for each content assignment scheme, to determine and to store
Reach the maximum content assignment scheme for the user group in total preference；
Each user group is reversed；And
Each user group of the poll through reversing, and the corresponding content assignment side stored is obtained for each user group
Case, to distribute content to each user group according to preference so that total preference reaches maximum.
11. a kind of content allocation system, including：
Memory is configured to storage for the content tab and respective confidence of content to be allocated and for each user group
User group label and respective confidence；And
Processor is connected via wired or wireless way with memory, and is configured to execute any power as in claim 1-7
Profit requires the method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710063920.1A CN108389066B (en) | 2017-02-03 | 2017-02-03 | Content distribution method, device, system and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710063920.1A CN108389066B (en) | 2017-02-03 | 2017-02-03 | Content distribution method, device, system and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108389066A true CN108389066A (en) | 2018-08-10 |
| CN108389066B CN108389066B (en) | 2022-02-01 |
Family
ID=63076040
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710063920.1A Active CN108389066B (en) | 2017-02-03 | 2017-02-03 | Content distribution method, device, system and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108389066B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109961308A (en) * | 2017-12-25 | 2019-07-02 | 北京京东尚科信息技术有限公司 | The method and apparatus of assessment tag data |
| CN112989175A (en) * | 2019-12-12 | 2021-06-18 | 北京沃东天骏信息技术有限公司 | Article pushing method, device, equipment and medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5487168A (en) * | 1992-06-15 | 1996-01-23 | International Business Machines Corporation | Method and system for global optimization of device allocation |
| US20070220103A1 (en) * | 2006-03-14 | 2007-09-20 | Michael Rogers | Targeting personalized content to mobile users |
| US7283971B1 (en) * | 2000-09-06 | 2007-10-16 | Masterlink Corporation | System and method for managing mobile workers |
| US20140074269A1 (en) * | 2012-09-11 | 2014-03-13 | Google Inc. | Method for Recommending Musical Entities to a User |
| CN104504098A (en) * | 2014-12-29 | 2015-04-08 | 北京奇虎科技有限公司 | Information recommending method and device |
| CN104915861A (en) * | 2015-06-15 | 2015-09-16 | 浙江经贸职业技术学院 | An electronic commerce recommendation method for a user group model constructed based on scores and labels |
| CN105740468A (en) * | 2016-03-07 | 2016-07-06 | 达而观信息科技（上海）有限公司 | Individuation recommendation method and system combined with content publisher information |
| CN106327090A (en) * | 2016-08-29 | 2017-01-11 | 安徽慧达通信网络科技股份有限公司 | Real task allocation method applied to preference crowd-sourcing system |
-
2017
- 2017-02-03 CN CN201710063920.1A patent/CN108389066B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5487168A (en) * | 1992-06-15 | 1996-01-23 | International Business Machines Corporation | Method and system for global optimization of device allocation |
| US7283971B1 (en) * | 2000-09-06 | 2007-10-16 | Masterlink Corporation | System and method for managing mobile workers |
| US20070220103A1 (en) * | 2006-03-14 | 2007-09-20 | Michael Rogers | Targeting personalized content to mobile users |
| US20140074269A1 (en) * | 2012-09-11 | 2014-03-13 | Google Inc. | Method for Recommending Musical Entities to a User |
| CN104504098A (en) * | 2014-12-29 | 2015-04-08 | 北京奇虎科技有限公司 | Information recommending method and device |
| CN104915861A (en) * | 2015-06-15 | 2015-09-16 | 浙江经贸职业技术学院 | An electronic commerce recommendation method for a user group model constructed based on scores and labels |
| CN105740468A (en) * | 2016-03-07 | 2016-07-06 | 达而观信息科技（上海）有限公司 | Individuation recommendation method and system combined with content publisher information |
| CN106327090A (en) * | 2016-08-29 | 2017-01-11 | 安徽慧达通信网络科技股份有限公司 | Real task allocation method applied to preference crowd-sourcing system |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109961308A (en) * | 2017-12-25 | 2019-07-02 | 北京京东尚科信息技术有限公司 | The method and apparatus of assessment tag data |
| CN112989175A (en) * | 2019-12-12 | 2021-06-18 | 北京沃东天骏信息技术有限公司 | Article pushing method, device, equipment and medium |
| CN112989175B (en) * | 2019-12-12 | 2025-02-28 | 北京沃东天骏信息技术有限公司 | Method, device, equipment and medium for pushing items |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108389066B (en) | 2022-02-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aggarwal et al. | Examining the factors influencing fintech adoption behaviour of gen Y in India | |
| Mahajan et al. | An approach to normative segmentation | |
| Montgomery et al. | Marketing decision-information systems: an emerging view | |
| Bettman | Perceived risk and its components: A model and empirical test | |
| Skenderi et al. | Well googled is half done: Multimodal forecasting of new fashion product sales with image‐based google trends | |
| Murari | Financial development–economic growth nexus: Evidence from South Asian middle-income countries | |
| Bukhari et al. | The journey of Pakistan’s banking industry towards green banking adoption | |
| Fonseca et al. | Granger causality between tourism and income: A meta-regression analysis | |
| Qin et al. | Uncertain random portfolio optimization models based on value-at-risk | |
| Doré et al. | Empirical literature on economic growth, 1991–2020: Uncovering extant gaps and avenues for future research | |
| Fforde | Yes, but what about services: is development doctrine changing? | |
| Baum et al. | The BDS test of independence | |
| Raheem | More finance or better finance in Feldstein–Horioka puzzle: Evidence from SSA countries | |
| Ahmed | Performance evaluation of regional rural banks: Evidence from Indian rural banks | |
| Yang et al. | Data envelopment analysis may obfuscate corporate financial data: using support vector machine and data envelopment analysis to predict corporate failure for nonmanufacturing firms | |
| CN108389066A (en) | Content distribution method based on Dynamic Programming and system | |
| JP2020047229A (en) | Article analysis device and article analysis method | |
| Ramesh | Bad loans of public sector banks in India: A panel data study | |
| Lemos et al. | Free apps and paid apps: monetization strategies for health apps in the Portuguese market | |
| Majeed et al. | Under-Informed Policy Interventions and Long-Run Damage to the MSME Sector in India: An Analysis of the Aatmanirbhar Bharat Abhiyan | |
| CN113919869A (en) | Equity distribution method and device based on sales increment model and electronic equipment | |
| Singh | Cointegration and Causality Between Foreign Direct Investment, Trade and Economic Growth: Empirical Evidence from India’s Post Economic Reform Period | |
| Giesecke et al. | Two-step analysis of hierarchical data | |
| CN116881246A (en) | Data processing methods, devices, storage media and electronic equipment | |
| Singh et al. | Efficiency assessment parameters of public sector banks in India |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |