CN108510326A - Determination method or initial value and device - Google Patents
Determination method or initial value and device Download PDFInfo
- Publication number
- CN108510326A CN108510326A CN201810274338.4A CN201810274338A CN108510326A CN 108510326 A CN108510326 A CN 108510326A CN 201810274338 A CN201810274338 A CN 201810274338A CN 108510326 A CN108510326 A CN 108510326A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- content characteristic
- clicking rate
- value
- discreet value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The disclosure is directed to Determination method or initial value and devices.Method includes:Identify the content characteristic of the first advertisement;According to the content characteristic of the first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, the clicking rate discreet value of the first advertisement is determined;According to the clicking rate prediction model based on content characteristic, the second advertisement that the difference in advertising aggregator with the clicking rate discreet value of the first advertisement is less than first threshold is searched;According to the history value of the statistical nature of the second advertisement, the initial value of the statistical nature of the first advertisement is determined.The disclosure can improve the chance that new advertisement obtains exposure, to solve the problem of the cold start-up of displaying advertisement avoid in the related technology flow income be difficult to weigh with cold start speed.
Description
Technical field
This disclosure relates to field of computer technology more particularly to Determination method or initial value and device.
Background technology
With the development of Internet technology and universal, businessman is in order to enhance the reputation and promote commodity, often by mutual
Advertisement is launched in networking.
In large-scale advertisement commending system, can all there are a large amount of new advertisements to enter advertisement base daily, due to new advertisement
It is under-exposed so that the statistical information of new advertisement is insufficient, lacks effective feature, is clicked rate predictive algorithm from advertisement base
A large amount of Candidate Sets in the probability very little selected, form the cycle of " under-exposure-shortage validity feature-is difficult to obtain exposure ",
There is serious cold start-up problem.Cold start-up problem is that advertisement commending system needs one of the critical issue solved.
In the related technology, the cold start-up of new advertisement is solved the problems, such as by marking off part dedicated traffic, this partial discharge or
Person is specially allocated to new advertising display, or selects advertising display at random to all advertising aggregators.
Invention content
To overcome the problems in correlation technique, a kind of Determination method or initial value of embodiment of the present disclosure offer and device.
The technical solution is as follows:
According to the first aspect of the embodiments of the present disclosure, a kind of Determination method or initial value is provided, method includes:
Identify the content characteristic of the first advertisement;
According to the content characteristic of first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, really
The clicking rate discreet value of fixed first advertisement;
According to the clicking rate prediction model based on content characteristic, the point with first advertisement in advertising aggregator is searched
The difference for hitting rate discreet value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of second advertisement, the initial of the statistical nature of first advertisement is determined
Value.
In one embodiment, the clicking rate prediction model based on content characteristic described in the basis searches advertising aggregator
In be less than the second advertisement of first threshold with the difference of the clicking rate discreet value of first advertisement, including:
Identify the content characteristic of each advertisement in the advertising aggregator;
According to the content characteristic of each advertisement in the clicking rate prediction model based on content characteristic and the advertising aggregator,
Determine the clicking rate discreet value of each advertisement in the advertising aggregator;
According to the clicking rate discreet value and exposure of each advertisement in the clicking rate discreet value of first advertisement, the advertising aggregator
Light quantity searches the second advertisement, the clicking rate discreet value of second advertisement and first advertisement in the advertising aggregator
The difference of clicking rate discreet value is less than first threshold and the light exposure of second advertisement is more than second threshold.
In one embodiment, the method further includes:
Identify the content characteristic of advertising copy;
According to the content characteristic and clicking rate of the advertising copy, determine that the clicking rate based on content characteristic estimates mould
Type.
In one embodiment, the content characteristic includes text information;
It is described to identify that the content characteristic of the first advertisement includes:The word in first advertisement is identified using character recognition technology
Language.
In one embodiment, the content characteristic includes characteristics of image;
It is described to identify that the content characteristic of the first advertisement includes:The image of first advertisement is extracted using depth convolutional network
Feature.
In one embodiment, the statistical nature includes at least following any or combination:Light exposure, clicking rate or
Download.
According to the second aspect of the embodiment of the present disclosure, a kind of initial value determining device is provided, including:
First identification module, for identification content characteristic of the first advertisement;
First determining module, for according to the content characteristic of first advertisement and obtain in advance based on content characteristic
Clicking rate prediction model determines the clicking rate discreet value of first advertisement;
Searching module, for according to the clicking rate prediction model based on content characteristic, search in advertising aggregator with institute
The difference for stating the clicking rate discreet value of the first advertisement is less than the second advertisement of first threshold;
Second determining module is used for the history value of the statistical nature according to second advertisement, determines first advertisement
Statistical nature initial value.
In one embodiment, the searching module, including:
Identify submodule, for identification in the advertising aggregator each advertisement content characteristic;
Determination sub-module, for according to each in the clicking rate prediction model based on content characteristic and the advertising aggregator
The content characteristic of advertisement determines the clicking rate discreet value of each advertisement in the advertising aggregator;
Submodule is searched, the clicking rate discreet value according to first advertisement, each advertisement in the advertising aggregator are used for
Clicking rate discreet value and light exposure search the second advertisement, the clicking rate discreet value of second advertisement in the advertising aggregator
The light exposure for being less than first threshold and second advertisement with the difference of the clicking rate discreet value of first advertisement is more than second
Threshold value.
In one embodiment, described device further includes:
Second identification module, for identification content characteristic of advertising copy;
Third determining module determines described based on content for the content characteristic and clicking rate according to the advertising copy
The clicking rate prediction model of feature.
In one embodiment, the content characteristic includes text information;First identification module uses Text region
Technology identifies the word in first advertisement.
In one embodiment, the content characteristic includes characteristics of image;First identification module uses depth convolution
Network extracts the characteristics of image of first advertisement.
According to the third aspect of the embodiment of the present disclosure, a kind of initial value determining device is provided, including:
Processor;
Memory for storing processor-executable instruction;
Wherein, the processor is configured as:
Identify the content characteristic of the first advertisement;
According to the content characteristic of first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, really
The clicking rate discreet value of fixed first advertisement;
According to the clicking rate prediction model based on content characteristic, the point with first advertisement in advertising aggregator is searched
The difference for hitting rate discreet value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of second advertisement, the initial of the statistical nature of first advertisement is determined
Value.
According to the fourth aspect of the embodiment of the present disclosure, a kind of computer readable storage medium is provided, is stored thereon with calculating
The step of machine instructs, which realizes any one of above-mentioned first aspect the method embodiment when being executed by processor.
The technical scheme provided by this disclosed embodiment can include the following benefits:The technical solution passes through study one
A clicking rate prediction model based on content characteristic determines new advertisement and old according to the clicking rate prediction model based on content characteristic
The clicking rate discreet value of advertisement, the lookup old advertisement close enough with the clicking rate discreet value of new advertisement in advertising aggregator, into
And the initial value of the statistical nature of new advertisement is determined according to the history value of the statistical nature of old advertisement, new advertisement can be improved and obtained
The chance of exposure avoids flow income in the related technology and cold start speed difficult to solve the problems, such as the cold start-up of displaying advertisement
The problem of to weigh.
It should be understood that above general description and following detailed description is only exemplary and explanatory, not
The disclosure can be limited.
Description of the drawings
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 a are the flow charts according to the Determination method or initial value shown in an exemplary embodiment.
Fig. 1 b are the principle schematics according to the training clicking rate prediction model shown in an exemplary embodiment.
Fig. 2 is the flow chart according to the Determination method or initial value shown in an exemplary embodiment.
Fig. 3 is the block diagram according to the initial value determining device shown in an exemplary embodiment.
Fig. 4 is the block diagram according to the initial value determining device shown in an exemplary embodiment.
Fig. 5 is the block diagram according to the initial value determining device shown in an exemplary embodiment.
Fig. 6 is the block diagram according to the initial value determining device shown in an exemplary embodiment.
Fig. 7 is the block diagram according to the device shown in an exemplary embodiment.
Specific implementation mode
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
In the related technology, the cold start-up of new advertisement is solved the problems, such as by marking off part dedicated traffic, this partial discharge or
Person is specially allocated to new advertising display, or selects advertising display at random to all advertising aggregators.However, either dividing flow
It is directed to new advertisement and still divides flow and all advertisements are shown at random, all there is following problems:1) receipts of outflow are divided
Enter and has larger reduction;2) the bad determination of flow proportional is marked off, division can cause income to decline to a great extent too much, divide very little cold
Starting can be very slow;And if flow is directed to new advertisement, whether there are flow proportionals to be asked with the variation of the amount of new advertisement
There is new and old advertisement scale effect cold start speed, the relevant technologies flow income if random to all advertisements in topic
It is difficult to weigh with cold start speed.
To solve the above-mentioned problems, the embodiment of the present disclosure provides a kind of Determination method or initial value, and method includes:Identification the
The content characteristic of one advertisement;Mould is estimated according to the content characteristic of the first advertisement and the clicking rate based on content characteristic obtained in advance
Type determines the clicking rate discreet value of the first advertisement;According to the clicking rate prediction model based on content characteristic, search in advertising aggregator
It is less than the second advertisement of first threshold with the difference of the clicking rate discreet value of the first advertisement;According to the statistical nature of the second advertisement
History value determines the initial value of the statistical nature of the first advertisement.The embodiment of the present disclosure characterizes advertisement using content characteristic, passes through
A clicking rate prediction model based on content characteristic is practised, new advertisement is determined according to the clicking rate prediction model based on content characteristic
And the clicking rate discreet value of old advertisement, it is searched in advertising aggregator and close enough old wide of the clicking rate discreet value of new advertisement
It accuses, and then determines according to the history value of the statistical nature of old advertisement the initial value of the statistical nature of new advertisement, can improve new wide
The chance for obtaining exposure is accused, to solve the problems, such as the cold start-up of displaying advertisement, avoids flow income and cold start-up in the related technology
Speed is difficult to the problem of weighing.
Fig. 1 a are a kind of flow charts of Determination method or initial value shown according to an exemplary embodiment;The execution of this method
Main body can be server;As shown in Figure 1a, this approach includes the following steps 101-104:
In a step 101, the content characteristic of the first advertisement is identified.
Exemplary, content characteristic refers to and the relevant feature of the content of advertisement.Content characteristic includes at least following any one
Kind feature or combination:Text information or characteristics of image.First advertisement for example can be new wide in advertising aggregator or advertisement base
It accuses.
By taking content characteristic includes text information as an example, the word in the first advertisement is identified using character recognition technology.For example,
Text box is extracted from the image of advertisement using text detection techniques, the word in text box is identified using character recognition technology,
Word is extracted from word using participle technique.
By taking content characteristic includes characteristics of image as an example, the characteristics of image of the first advertisement is extracted using depth convolutional network.Example
Such as, according to advertisement the characteristics of, the depth convolutional network extraction advertisement based on mobile terminal model (mobilenet) this lightweight
Characteristics of image.
In a step 102, pre- according to the content characteristic of the first advertisement and the clicking rate based on content characteristic obtained in advance
Estimate model, determines the clicking rate discreet value of the first advertisement.
Exemplary, the clicking rate prediction model based on content characteristic is for calculating the click that advertisement may be clicked by user
The content characteristic of the advertisement is inputted the clicking rate prediction model based on content characteristic, so that it may to calculate user by rate discreet value
Clicking rate discreet value to the advertisement.The step of learning clicking rate prediction model based on content characteristic may include:It obtains big
The advertising copy of amount;Identify the content characteristic of advertising copy;According to the content characteristic of advertising copy and actual click rate, base is determined
In the clicking rate prediction model of content characteristic.
Referring to Fig. 1 b, based in advertisement text information and characteristics of image both with the relevant feature of ad content, this
It is open to propose to train clicking rate prediction model using the network structure of a kind of interacting depth convolutional network and single layer network;Its
In, the input of single layer network is word W1, word W2, word W3... word WN, it is 0/1 feature of discretization, 0 representative pair
Word is answered not occur in advertising copy, 1 represents corresponding word occurs in advertising copy.The input of depth convolutional network is
The vector T of the characteristics of image for the advertisement extracted using depth convolutional network1、T2、T3……TM.Model output is that clicking rate is pre-
Valuation Y.In training, as advertising copy, the label of every advertising copy is actual point for magnanimity advertisement using in practical application scene
Hit rate.According in advertisement text information and the aspect of characteristics of image two analysis and understanding is carried out to advertisement, this understanding is to pass through end
To the machine learning techniques at end, it is trained the clicking rate prediction model based on content characteristic as target to optimize clicking rate, is imitated
Fruit is preferable.
It is exemplary, by learning the clicking rate prediction model based on content characteristic, according to the clicking rate based on content characteristic
Prediction model can determine the clicking rate discreet value of each old advertisement in new advertisement and advertising aggregator.
In step 103, according to the clicking rate prediction model based on content characteristic, search in advertising aggregator with the first advertisement
Clicking rate discreet value difference be less than first threshold the second advertisement.
It is exemplary, identify the content characteristic of each advertisement in advertising aggregator;Mould is estimated according to the clicking rate based on content characteristic
The content characteristic of each advertisement in type and advertising aggregator determines the clicking rate discreet value of each advertisement in advertising aggregator;It is wide according to first
The clicking rate discreet value of each advertisement and light exposure, search second in advertising aggregator in the clicking rate discreet value of announcement, advertising aggregator
Advertisement, the requirement that the second advertisement needs meet are:The clicking rate discreet value of second advertisement and the clicking rate discreet value of the first advertisement
Difference be less than the light exposure of first threshold and the second advertisement and be more than second threshold.
At step 104, according to the history value of the statistical nature of the second advertisement, the first of the statistical nature of the first advertisement is determined
Initial value.
Exemplary, statistical nature includes at least following any or combination:Light exposure, clicking rate or download.If wide
It accuses in set and finds the second advertisement met the requirements, then assign the history value of the statistical nature of the second advertisement to the first advertisement phase
The history value of the statistical nature of second advertisement is determined as the initial value of the statistical nature of the first advertisement by the statistical nature answered,
Make the statistical nature of the first advertisement in clicking rate predictive algorithm on line that there is effective initial value, to increase the first advertisement quilt
Clicking rate predictive algorithm on line chooses the probability being exposed from advertising aggregator, also, after exposure advertisement, the first advertisement
It again can be as clicking rate predictive algorithm on advertising copy iteratively trimming-wire.As it can be seen that the disclosure utilized is transfer learning
Thought, by the clicking rate predictive algorithm in the transfer of learning to line of the clicking rate prediction model of the content characteristic based on advertisement, with
Solve the problems, such as the cold start-up of new advertisement.Also, the learning process due to the clicking rate prediction model based on content characteristic and advertisement
Clicking rate discreet value calculating process be all it is online lower complete, to the negative effect very little of the advertisement commending system on line.It is right
In new advertisement, it can disposably reach the initialization of near-optimization, and can seamlessly dock and be estimated into clicking rate on line
In the Iterative minor adjustment Optimizing Flow of algorithm.
If the number more than one of the advertisement met the requirements is found in advertising aggregator, from the advertisement met the requirements
Choose clicking rate discreet value and the difference minimum of the clicking rate discreet value of the first advertisement one.
For example, for displaying advertisement (display ads), image is the carrier that user perceives advertisement, disclosure profit
With the means of transfer learning, the content characteristic of user's clicking rate is influenced in analysis and understanding image, if new advertisement and old advertisement have
Same or analogous content characteristic then can assign new advertisement the history value of the statistical nature of the sufficient old advertisement of exposure, make new
Advertisement has the initial value of effective statistical nature, so that new advertisement is easy to get chance for exposure, is opened to solve the cold of new advertisement
Dynamic problem.
The technical solution that the embodiment of the present disclosure provides characterizes advertisement using content characteristic, special based on content by learning one
The clicking rate prediction model of sign, according to the clicking rate for determining new advertisement and old advertisement based on the clicking rate prediction model of content characteristic
Discreet value, the lookup old advertisement close enough with the clicking rate discreet value of new advertisement in advertising aggregator, and then according to old advertisement
Statistical nature history value determine new advertisement statistical nature initial value, the chance that new advertisement obtains exposure can be improved,
To solve the problems, such as displaying advertisement cold start-up, avoid in the related technology flow income asked with what cold start speed was difficult to weigh
Topic.
Fig. 2 is a kind of flow chart of Determination method or initial value shown according to an exemplary embodiment.As shown in Fig. 2,
On the basis of Fig. 1 a illustrated embodiments, this disclosure relates to Determination method or initial value may comprise steps of 201-208:
In step 201, the content characteristic of advertising copy is identified.
In step 202, according to the content characteristic of advertising copy and clicking rate, determine that the clicking rate based on content characteristic is pre-
Estimate model.
In step 203, the content characteristic of the first advertisement is identified.
In step 204, pre- according to the content characteristic of the first advertisement and the clicking rate based on content characteristic obtained in advance
Estimate model, determines the clicking rate discreet value of the first advertisement.
In step 205, the content characteristic of each advertisement in advertising aggregator is identified.
In step 206, according to the content of each advertisement in clicking rate prediction model and advertising aggregator based on content characteristic
Feature determines the clicking rate discreet value of each advertisement in advertising aggregator.
In step 207, according to the clicking rate discreet value of each advertisement in the clicking rate discreet value of the first advertisement, advertising aggregator
And light exposure, the second advertisement is searched in advertising aggregator, the clicking rate discreet value of the second advertisement and the clicking rate of the first advertisement are pre-
The difference of valuation is less than first threshold and the light exposure of the second advertisement is more than second threshold.
In a step 208, according to the history value of the statistical nature of the second advertisement, the first of the statistical nature of the first advertisement is determined
Initial value.
In the technical solution that the embodiment of the present disclosure provides, by determining base according to the content characteristic and clicking rate of advertising copy
In the clicking rate prediction model of content characteristic, the clicking rate prediction model based on content characteristic searched in advertising aggregator with it is new wide
The close enough old advertisement of the clicking rate discreet value of announcement, the statistics of new advertisement is determined according to the history value of the statistical nature of old advertisement
The initial value of feature can improve the chance that new advertisement obtains exposure and avoid phase to solve the problems, such as the cold start-up of displaying advertisement
The problem of flow income is difficult to weigh with cold start speed in the technology of pass.
Following is embodiment of the present disclosure, can be used for executing embodiments of the present disclosure.
Fig. 3 is a kind of block diagram of initial value determining device shown according to an exemplary embodiment;The device may be used
Various modes are implemented, such as all components of implementation in the server, alternatively, real in a coupled manner in server side
Apply the component in device;The device can by software, hardware or both be implemented in combination with it is above-mentioned this disclosure relates to method,
As shown in figure 3, the initial value determining device includes:First identification module 301, the first determining module 302, searching module 303 and
Second determining module 304, wherein:
First identification module 301 is configured as the content characteristic of the first advertisement of identification;
First determining module 302 be configured as the content characteristic according to the first advertisement and obtain in advance based on content characteristic
Clicking rate prediction model, determine the clicking rate discreet value of the first advertisement;
Searching module 303 is configured as according to the clicking rate prediction model based on content characteristic, search in advertising aggregator with
The difference of the clicking rate discreet value of first advertisement is less than the second advertisement of first threshold;
Second determining module 304 is configured as the history value of the statistical nature according to the second advertisement, determines the first advertisement
The initial value of statistical nature.
The device that the embodiment of the present disclosure provides can be used in executing the technical solution of Fig. 1 a illustrated embodiments, executive mode
Similar with advantageous effect, details are not described herein again.
In a kind of possible embodiment, as shown in figure 4, the initial value determining device shown in Fig. 3 can also include handle
Searching module 303 is configured to:It identifies submodule 401, determination sub-module 402 and searches submodule 403, wherein:
Identification submodule 401 is configured as the content characteristic of each advertisement in identification advertising aggregator;
Determination sub-module 402 is configured as according to each wide in clicking rate prediction model and advertising aggregator based on content characteristic
The content characteristic of announcement determines the clicking rate discreet value of each advertisement in advertising aggregator;
Search submodule 403 be configured as the clicking rate discreet value according to the first advertisement, in advertising aggregator each advertisement point
Rate discreet value and light exposure are hit, the second advertisement, clicking rate discreet value and the first advertisement of the second advertisement are searched in advertising aggregator
The difference of clicking rate discreet value be less than the light exposure of first threshold and the second advertisement and be more than second threshold.
In a kind of possible embodiment, as shown in figure 5, the initial value determining device shown in Fig. 3 can also include:The
Two identification modules 501 and third determining module 502, wherein:
Second identification module 501 is configured as the content characteristic of identification advertising copy;
Third determining module 502 is configured as content characteristic and clicking rate according to advertising copy, determines special based on content
The clicking rate prediction model of sign.
In a kind of possible embodiment, content characteristic includes text information;First identification module 301 is known using word
Other technology identifies the word in the first advertisement.
In a kind of possible embodiment, content characteristic includes characteristics of image;First identification module 301 is rolled up using depth
Product network extracts the characteristics of image of the first advertisement.
Fig. 6 is a kind of block diagram of initial value determining device 600 shown according to an exemplary embodiment, and initial value determines dress
It sets 600 and is suitable for server, initial value determining device 600 includes:
Processor 601;
Memory 602 for storing processor-executable instruction;
Wherein, processor 601 is configured as:
Identify the content characteristic of the first advertisement;
According to the content characteristic of the first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, is determined
The clicking rate discreet value of one advertisement;
According to the clicking rate prediction model based on content characteristic, searches the clicking rate in advertising aggregator with the first advertisement and estimate
The difference of value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of the second advertisement, the initial value of the statistical nature of the first advertisement is determined.
In one embodiment, above-mentioned processor 601 is also configured to:
Identify the content characteristic of each advertisement in advertising aggregator;
According to the content characteristic of each advertisement in clicking rate prediction model and advertising aggregator based on content characteristic, advertisement is determined
The clicking rate discreet value of each advertisement in set;
According to the clicking rate discreet value of each advertisement and light exposure in the clicking rate discreet value of the first advertisement, advertising aggregator,
The second advertisement is searched in advertising aggregator, the clicking rate discreet value of the second advertisement and the difference of the clicking rate discreet value of the first advertisement are small
In first threshold and the light exposure of the second advertisement is more than second threshold.
In one embodiment, above-mentioned processor 601 is also configured to:
Identify the content characteristic of advertising copy;
According to the content characteristic and clicking rate of advertising copy, the clicking rate prediction model based on content characteristic is determined.
In one embodiment, content characteristic includes text information;Above-mentioned processor 601 is also configured to:Use text
Word identification technology identifies the word in the first advertisement.
In one embodiment, content characteristic includes characteristics of image;Above-mentioned processor 601 is also configured to:Use depth
Spend the characteristics of image that convolutional network extracts the first advertisement.
In one embodiment, statistical nature includes at least following any or combination:Light exposure, clicking rate are downloaded
Amount.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
Fig. 7 is a kind of block diagram of device shown according to an exemplary embodiment.For example, device 700 may be provided as
One server.Device 700 includes processing component 702, further comprises one or more processors, and by memory 703
Representative memory resource, can be by the instruction of the execution of processing component 702, such as application program for storing.Memory 703
The application program of middle storage may include it is one or more each correspond to one group of instruction module.In addition, processing
Component 702 is configured as executing instruction, to execute the above method.
Device 700 can also include the power management that a power supply module 706 is configured as executive device 700, and one has
Line or radio network interface 705 are configured as device 700 being connected to network and input and output (I/O) interface 708.Dress
Setting 700 can operate based on the operating system for being stored in memory 703, such as Windows ServerTM, Mac OS XTM,
UnixTM, LinuxTM, FreeBSDTM or similar.
A kind of non-transitorycomputer readable storage medium, when the instruction in storage medium is held by the processor of device 700
When row so that device 700 is able to carry out following method:
Identify the content characteristic of the first advertisement;
According to the content characteristic of the first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, is determined
The clicking rate discreet value of one advertisement;
According to the clicking rate prediction model based on content characteristic, searches the clicking rate in advertising aggregator with the first advertisement and estimate
The difference of value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of the second advertisement, the initial value of the statistical nature of the first advertisement is determined.
In one embodiment, it according to the clicking rate prediction model based on content characteristic, searches in advertising aggregator with first
The difference of the clicking rate discreet value of advertisement is less than the second advertisement of first threshold, including:
Identify the content characteristic of each advertisement in advertising aggregator;
According to the content characteristic of each advertisement in clicking rate prediction model and advertising aggregator based on content characteristic, advertisement is determined
The clicking rate discreet value of each advertisement in set;
According to the clicking rate discreet value of each advertisement and light exposure in the clicking rate discreet value of the first advertisement, advertising aggregator,
The second advertisement is searched in advertising aggregator, the clicking rate discreet value of the second advertisement and the difference of the clicking rate discreet value of the first advertisement are small
In first threshold and the light exposure of the second advertisement is more than second threshold.
In one embodiment, method further includes:
Identify the content characteristic of advertising copy;
According to the content characteristic and clicking rate of advertising copy, the clicking rate prediction model based on content characteristic is determined.
In one embodiment, content characteristic includes text information;
Identify that the content characteristic of the first advertisement includes:The word in the first advertisement is identified using character recognition technology.
In one embodiment, content characteristic includes characteristics of image;
Identify that the content characteristic of the first advertisement includes:The characteristics of image of the first advertisement is extracted using depth convolutional network.
In one embodiment, statistical nature includes at least following any or combination:Light exposure, clicking rate are downloaded
Amount.
Those skilled in the art will readily occur to its of the disclosure after considering specification and putting into practice disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and includes the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following
Claim is pointed out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (13)
1. a kind of Determination method or initial value, which is characterized in that including:
Identify the content characteristic of the first advertisement;
According to the content characteristic of first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, institute is determined
State the clicking rate discreet value of the first advertisement;
According to the clicking rate prediction model based on content characteristic, the clicking rate with first advertisement in advertising aggregator is searched
The difference of discreet value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of second advertisement, the initial value of the statistical nature of first advertisement is determined.
2. according to the method described in claim 1, it is characterized in that, the clicking rate based on content characteristic described in the basis is estimated
Model searches the second advertisement that the difference in advertising aggregator with the clicking rate discreet value of first advertisement is less than first threshold,
Including:
Identify the content characteristic of each advertisement in the advertising aggregator;
According to the content characteristic of each advertisement in the clicking rate prediction model based on content characteristic and the advertising aggregator, determine
The clicking rate discreet value of each advertisement in the advertising aggregator;
According to the clicking rate discreet value and exposure of each advertisement in the clicking rate discreet value of first advertisement, the advertising aggregator
Amount, searches the second advertisement, the point of the clicking rate discreet value of second advertisement and first advertisement in the advertising aggregator
The difference for hitting rate discreet value is less than first threshold and the light exposure of second advertisement is more than second threshold.
3. according to the method described in claim 1, it is characterized in that, the method further includes:
Identify the content characteristic of advertising copy;
According to the content characteristic and clicking rate of the advertising copy, the clicking rate prediction model based on content characteristic is determined.
4. according to the method described in claim 1, it is characterized in that, the content characteristic includes text information;
It is described to identify that the content characteristic of the first advertisement includes:The word in first advertisement is identified using character recognition technology.
5. according to the method described in claim 1, it is characterized in that, the content characteristic includes characteristics of image;
It is described to identify that the content characteristic of the first advertisement includes:The image that first advertisement is extracted using depth convolutional network is special
Sign.
6. the method according to any one of claims 1 to 5, it is characterized in that, the statistical nature is including at least following
Any or combination:Light exposure, clicking rate or download.
7. a kind of initial value determining device, which is characterized in that including:
First identification module, for identification content characteristic of the first advertisement;
First determining module, for according to the content characteristic of first advertisement and the click based on content characteristic obtained in advance
Rate prediction model determines the clicking rate discreet value of first advertisement;
Searching module, for according to the clicking rate prediction model based on content characteristic, searching in advertising aggregator with described the
The difference of the clicking rate discreet value of one advertisement is less than the second advertisement of first threshold;
Second determining module is used for the history value of the statistical nature according to second advertisement, determines the system of first advertisement
Count the initial value of feature.
8. device according to claim 7, which is characterized in that the searching module, including:
Identify submodule, for identification in the advertising aggregator each advertisement content characteristic;
Determination sub-module, for according to each advertisement in the clicking rate prediction model based on content characteristic and the advertising aggregator
Content characteristic, determine the clicking rate discreet value of each advertisement in the advertising aggregator;
Submodule is searched, for the click according to each advertisement in the clicking rate discreet value of first advertisement, the advertising aggregator
Rate discreet value and light exposure search the second advertisement, the clicking rate discreet value of second advertisement and institute in the advertising aggregator
The difference for stating the clicking rate discreet value of the first advertisement is less than first threshold and the light exposure of second advertisement is more than the second threshold
Value.
9. device according to claim 7, which is characterized in that described device further includes:
Second identification module, for identification content characteristic of advertising copy;
Third determining module determines described based on content characteristic for the content characteristic and clicking rate according to the advertising copy
Clicking rate prediction model.
10. device according to claim 7, which is characterized in that the content characteristic includes text information;Described first knows
Other module identifies the word in first advertisement using character recognition technology.
11. device according to claim 7, which is characterized in that the content characteristic includes characteristics of image;Described first knows
Other module extracts the characteristics of image of first advertisement using depth convolutional network.
12. a kind of initial value determining device, which is characterized in that including:
Processor;
Memory for storing processor-executable instruction;
Wherein, the processor is configured as:
Identify the content characteristic of the first advertisement;
According to the content characteristic of first advertisement and the clicking rate prediction model based on content characteristic obtained in advance, institute is determined
State the clicking rate discreet value of the first advertisement;
According to the clicking rate prediction model based on content characteristic, the clicking rate with first advertisement in advertising aggregator is searched
The difference of discreet value is less than the second advertisement of first threshold;
According to the history value of the statistical nature of second advertisement, the initial value of the statistical nature of first advertisement is determined.
13. a kind of computer readable storage medium, is stored thereon with computer instruction, which is characterized in that the instruction is by processor
The step of any one of claim 1-6 the methods are realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810274338.4A CN108510326B (en) | 2018-03-29 | 2018-03-29 | Initial value determination method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810274338.4A CN108510326B (en) | 2018-03-29 | 2018-03-29 | Initial value determination method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108510326A true CN108510326A (en) | 2018-09-07 |
CN108510326B CN108510326B (en) | 2022-02-11 |
Family
ID=63379529
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810274338.4A Active CN108510326B (en) | 2018-03-29 | 2018-03-29 | Initial value determination method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108510326B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109670632A (en) * | 2018-11-26 | 2019-04-23 | 北京达佳互联信息技术有限公司 | The predictor method of ad click rate, the estimating device of ad click rate, electronic equipment and storage medium |
CN110489644A (en) * | 2019-08-13 | 2019-11-22 | 腾讯科技(北京)有限公司 | Information-pushing method, device, computer readable storage medium and computer equipment |
WO2021081962A1 (en) * | 2019-10-31 | 2021-05-06 | 华为技术有限公司 | Recommendation model training method, recommendation method, device, and computer-readable medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104574124A (en) * | 2013-10-10 | 2015-04-29 | 深圳市腾讯计算机系统有限公司 | Method and device for determining exhibition effect of advertisement data |
CN104636940A (en) * | 2013-11-07 | 2015-05-20 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method, advertisement delivery system, advertisement screening method, advertisement screening device, and server |
CN107480124A (en) * | 2017-07-05 | 2017-12-15 | 小草数语(北京)科技有限公司 | Advertisement placement method and device |
-
2018
- 2018-03-29 CN CN201810274338.4A patent/CN108510326B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104574124A (en) * | 2013-10-10 | 2015-04-29 | 深圳市腾讯计算机系统有限公司 | Method and device for determining exhibition effect of advertisement data |
CN104636940A (en) * | 2013-11-07 | 2015-05-20 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method, advertisement delivery system, advertisement screening method, advertisement screening device, and server |
CN107480124A (en) * | 2017-07-05 | 2017-12-15 | 小草数语(北京)科技有限公司 | Advertisement placement method and device |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109670632A (en) * | 2018-11-26 | 2019-04-23 | 北京达佳互联信息技术有限公司 | The predictor method of ad click rate, the estimating device of ad click rate, electronic equipment and storage medium |
CN109670632B (en) * | 2018-11-26 | 2021-01-29 | 北京达佳互联信息技术有限公司 | Advertisement click rate estimation method, advertisement click rate estimation device, electronic device and storage medium |
CN110489644A (en) * | 2019-08-13 | 2019-11-22 | 腾讯科技(北京)有限公司 | Information-pushing method, device, computer readable storage medium and computer equipment |
WO2021081962A1 (en) * | 2019-10-31 | 2021-05-06 | 华为技术有限公司 | Recommendation model training method, recommendation method, device, and computer-readable medium |
Also Published As
Publication number | Publication date |
---|---|
CN108510326B (en) | 2022-02-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20210049622A1 (en) | Marketing method and apparatus based on deep reinforcement learning | |
CN106547908B (en) | Information pushing method and system | |
CN105160545B (en) | Method and device for determining release information style | |
CN109300059B (en) | Dish recommending method and device | |
CN107688605B (en) | Cross-platform data matching process, device, computer equipment and storage medium | |
CN109325179B (en) | Content promotion method and device | |
US20130346182A1 (en) | Multimedia features for click prediction of new advertisements | |
US20090271228A1 (en) | Construction of predictive user profiles for advertising | |
US20180039675A1 (en) | Extended search method and apparatus | |
CN108510326A (en) | Determination method or initial value and device | |
CN105183772A (en) | Release information click rate estimation method and apparatus | |
US11138249B1 (en) | Systems and methods for the creation, update and use of concept networks to select destinations in artificial intelligence systems | |
CN108596695B (en) | Entity pushing method and system | |
CN111210258A (en) | Advertisement putting method and device, electronic equipment and readable storage medium | |
CN103295145A (en) | Mobile phone advertising method based on user consumption feature vector | |
CN105740268A (en) | Information pushing method and apparatus | |
CN110019703B (en) | Data marking method and device and intelligent question-answering method and system | |
CN109063104A (en) | Method for refreshing, device, storage medium and the terminal device of recommendation information | |
WO2017143773A1 (en) | Crowdsourcing learning method and device | |
CN111008335B (en) | Information processing method, device, equipment and storage medium | |
CN111144933A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
CN104915354A (en) | Multimedia file pushing method and device | |
Yankov et al. | Evaluation of explore-exploit policies in multi-result ranking systems | |
CN111144936B (en) | Similar crowd expansion method and device based on user labels | |
CN112559777A (en) | Content item delivery method and device, computer equipment and storage medium |
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 |