CN108510326B - Initial value determination method and device - Google Patents
Initial value determination method and device Download PDFInfo
- Publication number
- CN108510326B CN108510326B CN201810274338.4A CN201810274338A CN108510326B CN 108510326 B CN108510326 B CN 108510326B CN 201810274338 A CN201810274338 A CN 201810274338A CN 108510326 B CN108510326 B CN 108510326B
- Authority
- CN
- China
- Prior art keywords
- advertisement
- click rate
- value
- content characteristics
- click
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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 relates to an initial value determination method and device. The method comprises the following steps: identifying a content characteristic of the first advertisement; determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics; searching a second advertisement in the advertisement set, wherein the difference value of the click rate estimated value of the first advertisement and the click rate estimated value of the second advertisement is smaller than a first threshold value, according to a click rate estimation model based on content characteristics; an initial value of the statistical characteristic of the first advertisement is determined based on the historical value of the statistical characteristic of the second advertisement. The method and the device can improve the exposure opportunity of the new advertisement, solve the cold start problem of the displayed advertisement and avoid the problem that the traffic income and the cold start speed are difficult to balance in the related technology.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to an initial value determination method and apparatus.
Background
With the development and popularization of internet technology, merchants often put advertisements through the internet in order to improve popularity and promote commodities.
In a large-scale advertisement recommendation system, a large number of new advertisements enter an advertisement library every day, due to underexposure of the new advertisements, statistical information of the new advertisements is insufficient, effective characteristics are lacked, the probability that a click rate estimation algorithm selects a large number of candidate sets in the advertisement library is low, a cycle of underexposure, effective characteristics are lacked, and exposure is difficult to obtain is formed, namely a serious cold start problem occurs. The cold start problem is one of the key issues that the advertisement recommendation system needs to solve.
In the related art, the cold start problem of the new advertisement is solved by dividing part of special traffic, and the part of traffic is either specially divided for the new advertisement display or randomly selected for all advertisement sets.
Disclosure of Invention
In order to overcome the problems in the related art, embodiments of the present disclosure provide an initial value determining method and apparatus. The technical scheme is as follows:
according to a first aspect of the embodiments of the present disclosure, there is provided an initial value determining method, including:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimated value of the first advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimated model based on the content characteristics;
and determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement.
In one embodiment, the searching for a second advertisement in the advertisement set, which has a difference value with the estimated click-through rate value of the first advertisement smaller than a first threshold value according to the pre-estimation model of click-through rate based on content features, includes:
identifying content characteristics of each advertisement in the set of advertisements;
determining click rate pre-estimated values of all advertisements in the advertisement set according to the click rate pre-estimated model based on the content characteristics and the content characteristics of all advertisements in the advertisement set;
and searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
In one embodiment, the method further comprises:
identifying content characteristics of an advertising sample;
and determining the click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
In one embodiment, the content features include textual information;
the identifying content characteristics of the first advertisement includes: words in the first advertisement are identified using a word recognition technique.
In one embodiment, the content features include image features;
the identifying content characteristics of the first advertisement includes: extracting image features of the first advertisement using a deep convolutional network.
In one embodiment, the statistical features include at least any one or combination of: exposure, click rate, or download volume.
According to a second aspect of the embodiments of the present disclosure, there is provided an initial value determination apparatus including:
a first identification module for identifying content characteristics of a first advertisement;
the first determining module is used for determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
the searching module is used for searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimation value of the first advertisement and the click rate pre-estimation value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimation model based on the content characteristics;
and the second determining module is used for determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement.
In one embodiment, the lookup module includes:
the identification submodule is used for identifying the content characteristics of each advertisement in the advertisement set;
the determining submodule is used for determining the click rate pre-estimated value of each advertisement in the advertisement set according to the click rate pre-estimation model based on the content characteristics and the content characteristics of each advertisement in the advertisement set;
and the searching submodule is used for searching a second advertisement in the advertisement set according to the click rate estimated value of the first advertisement, the click rate estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate estimated value of the second advertisement and the click rate estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
In one embodiment, the apparatus further comprises:
the second identification module is used for identifying the content characteristics of the advertisement samples;
and the third determining module is used for determining the click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
In one embodiment, the content features include textual information; the first identification module identifies words in the first advertisement using word recognition techniques.
In one embodiment, the content features include image features; the first identification module extracts image features of the first advertisement using a deep convolutional network.
According to a third aspect of the embodiments of the present disclosure, there is provided an initial value determination apparatus including:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimated value of the first advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimated model based on the content characteristics;
and determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement.
According to a fourth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the steps of the method embodiments of any one of the above-mentioned first aspects.
The technical scheme provided by the embodiment of the disclosure can have the following beneficial effects: according to the technical scheme, a click rate pre-estimation model based on content features is learned, click rate pre-estimation values of new advertisements and old advertisements are determined according to the click rate pre-estimation model based on the content features, the old advertisements close enough to the click rate pre-estimation values of the new advertisements are searched in an advertisement set, then initial values of statistical features of the new advertisements are determined according to historical values of the statistical features of the old advertisements, the exposure opportunities of the new advertisements can be improved, the cold start problem of displaying the advertisements is solved, and the problem that flow income and the cold start speed are difficult to balance in the related technology is solved.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and together with the description, serve to explain the principles of the disclosure.
Fig. 1a is a flow chart illustrating an initial value determination method according to an exemplary embodiment.
FIG. 1b is a schematic diagram illustrating training a click-through rate prediction model according to an exemplary embodiment.
Fig. 2 is a flow chart illustrating an initial value determination method according to an example embodiment.
Fig. 3 is a block diagram illustrating an initial value determination apparatus according to an exemplary embodiment.
Fig. 4 is a block diagram illustrating an initial value determination apparatus according to an exemplary embodiment.
Fig. 5 is a block diagram illustrating an initial value determination apparatus according to an exemplary embodiment.
Fig. 6 is a block diagram illustrating an initial value determination apparatus according to an exemplary embodiment.
Fig. 7 is a block diagram illustrating an apparatus according to an example embodiment.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
In the related art, the cold start problem of the new advertisement is solved by dividing part of special traffic, and the part of traffic is either specially divided for the new advertisement display or randomly selected for all advertisement sets. However, whether the split traffic is targeted to new advertisements or the split traffic is randomly exposed for all advertisements, the following problems exist: 1) the income of the divided flow can be greatly reduced; 2) the flow rate proportion is not well determined, the income is greatly reduced due to too much division, and the cold start is slow when the division is too little; and if the traffic is directed to the new advertisement, the problem that whether the traffic proportion is changed along with the quantity of the new advertisement exists, and if all the advertisements are random, the problem that the cold start speed is influenced by the proportion of the new advertisement and the old advertisement exists, and the traffic income and the cold start speed of the related technology are difficult to balance.
In order to solve the above problem, an embodiment of the present disclosure provides an initial value determining method, where the method includes: identifying a content characteristic of the first advertisement; determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics; searching a second advertisement in the advertisement set, wherein the difference value of the click rate estimated value of the first advertisement and the click rate estimated value of the second advertisement is smaller than a first threshold value, according to a click rate estimation model based on content characteristics; an initial value of the statistical characteristic of the first advertisement is determined based on the historical value of the statistical characteristic of the second advertisement. According to the method and the device, the advertisement is represented by using the content characteristics, the click rate pre-estimation value of the new advertisement and the old advertisement is determined according to the click rate pre-estimation model based on the content characteristics by learning the click rate pre-estimation model based on the content characteristics, the old advertisement close enough to the click rate pre-estimation value of the new advertisement is searched in the advertisement set, the initial value of the statistical characteristics of the new advertisement is determined according to the historical value of the statistical characteristics of the old advertisement, the exposure opportunity of the new advertisement can be improved, the cold start problem of displaying the advertisement is solved, and the problem that the traffic income and the cold start speed are difficult to balance in the related technology is solved.
FIG. 1a is a flow chart illustrating an initial value determination method according to an exemplary embodiment; the execution subject of the method can be a server; as shown in fig. 1a, the method comprises the following steps 101-104:
in step 101, content characteristics of a first advertisement are identified.
By way of example, content characteristics refer to characteristics related to the content of an advertisement. The content features include at least any one or a combination of the following: textual information, or image features. The first advertisement may be, for example, a new advertisement in a collection or library of advertisements.
Using text recognition techniques, for example, where the content features include text information, words in the first advertisement are identified. For example, text boxes are extracted from images of advertisements using text detection techniques, words in text boxes are identified using word recognition techniques, and words are extracted from words using word segmentation techniques.
Taking the example where the content features include image features, the image features of the first advertisement are extracted using a deep convolutional network. For example, according to the characteristics of the advertisement, the image features of the advertisement are extracted based on a light-weight deep convolution network such as a mobile terminal model (mobilenet).
In step 102, a click-through rate pre-estimated value of the first advertisement is determined according to the content characteristics of the first advertisement and a pre-obtained click-through rate pre-estimation model based on the content characteristics.
For example, the click-through rate estimation model based on the content characteristics is used for calculating the click-through rate estimation value of the advertisement which is likely to be clicked by the user, and the click-through rate estimation value of the advertisement for the user can be calculated by inputting the content characteristics of the advertisement into the click-through rate estimation model based on the content characteristics. The step of learning the content feature-based click through rate prediction model may include: obtaining a large number of advertisement samples; identifying content characteristics of an advertising sample; and determining a click rate estimation model based on the content characteristics according to the content characteristics and the actual click rate of the advertisement sample.
Referring to fig. 1b, based on two characteristics related to advertisement content, namely, text information and image characteristics in the advertisement, the present disclosure proposes to train a click rate prediction model using a network structure of a hybrid deep convolutional network and a single-layer network; wherein the input of the single-layer network is a word W1Word W2Word W3… … word WNAnd 0/1 features that are discretized, with 0 representing the absence of the corresponding word in the ad sample and 1 representing the presence of the corresponding word in the ad sample. The input of the deep convolutional network is the extraction by using the deep convolutional networkVector T of image features of incoming advertisements1、T2、T3……TM. The model output is the click rate estimated value Y. In training, mass advertisements in an actual application scene are used as advertisement samples, and the label of each advertisement sample is used as an actual click rate. The advertisement is analyzed and understood according to the character information and the image characteristics in the advertisement, the understanding is that the click rate estimation model based on the content characteristics is trained by using the optimized click rate as a target through an end-to-end machine learning technology, and the effect is good.
For example, by learning the click-through rate prediction model based on the content features, the click-through rate prediction value of each of the new advertisement and the old advertisements in the advertisement set can be determined according to the click-through rate prediction model based on the content features.
In step 103, according to the click-through rate estimation model based on the content features, a second advertisement in the advertisement set, which has a difference value with the click-through rate estimated value of the first advertisement smaller than a first threshold value, is searched.
Illustratively, content characteristics of each advertisement in the set of advertisements are identified; determining click rate pre-evaluation values of the advertisements in the advertisement set according to the click rate pre-evaluation model based on the content characteristics and the content characteristics of the advertisements in the advertisement set; searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the second advertisement needs to meet the following requirements: the difference value between the click rate estimated value of the second advertisement and the click rate estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
In step 104, an initial value of the statistical characteristic of the first advertisement is determined based on the historical value of the statistical characteristic of the second advertisement.
Exemplary, the statistical features include at least any one or combination of: exposure, click rate, or download volume. If the second advertisement meeting the requirement is found in the advertisement set, the historical value of the statistical characteristic of the second advertisement is endowed to the corresponding statistical characteristic of the first advertisement, namely the historical value of the statistical characteristic of the second advertisement is determined as the initial value of the statistical characteristic of the first advertisement, so that the statistical characteristic of the first advertisement in the on-line click rate estimation algorithm has an effective initial value, the exposure probability of the first advertisement from the advertisement set by the on-line click rate estimation algorithm is increased, and after the advertisement is exposed, the first advertisement can be used as an advertisement sample to iteratively fine-tune the on-line click rate estimation algorithm. Therefore, the idea of transfer learning is utilized in the method, the learning of the click rate estimation model based on the content characteristics of the advertisement is transferred to the click rate estimation algorithm on the line, and the cold start problem of the new advertisement is solved. In addition, because the learning process of the click rate estimation model based on the content characteristics and the calculation process of the click rate estimation value of the advertisement are completed on line, the negative influence on an online advertisement recommendation system is small. For new advertisements, near-optimal initialization can be achieved once, and seamless docking can be achieved into an iterative fine-tuning optimization flow of an online click rate estimation algorithm.
And if more than one advertisement meeting the requirements is found in the advertisement set, selecting one advertisement meeting the requirements with the smallest difference value between the click rate estimated value and the click rate estimated value of the first advertisement.
For example, for display ads (display ads), an image is a carrier for a user to perceive an advertisement, the present disclosure analyzes and understands content features affecting a user click rate in the image by means of migration learning, and if a new advertisement and an old advertisement have the same or similar content features, a history value of statistical features of the old advertisement with sufficient exposure can be given to the new advertisement, so that the new advertisement has an initial value of effective statistical features, so that the new advertisement is easy to obtain an exposure opportunity, thereby solving a cold start problem of the new advertisement.
The technical scheme provided by the embodiment of the disclosure uses the content characteristics to represent the advertisement, and determines the click rate pre-estimation values of the new advertisement and the old advertisement according to the click rate pre-estimation model based on the content characteristics by learning the click rate pre-estimation model based on the content characteristics, searches the old advertisement which is close enough to the click rate pre-estimation value of the new advertisement in the advertisement set, and further determines the initial value of the statistical characteristics of the new advertisement according to the historical value of the statistical characteristics of the old advertisement, so that the exposure opportunity of the new advertisement can be improved, the cold start problem of displaying the advertisement is solved, and the problem that the balance between the flow income and the cold start speed in the related technology is difficult is avoided.
Fig. 2 is a flow chart illustrating an initial value determination method according to an example embodiment. As shown in fig. 2, on the basis of the embodiment shown in fig. 1a, the initial value determination method according to the present disclosure may include the following steps 201 and 208:
in step 201, content characteristics of a sample of advertisements are identified.
In step 202, according to the content characteristics and the click-through rate of the advertisement sample, a click-through rate estimation model based on the content characteristics is determined.
In step 203, content characteristics of the first advertisement are identified.
In step 204, a click-through rate prediction value of the first advertisement is determined according to the content characteristics of the first advertisement and a pre-obtained click-through rate prediction model based on the content characteristics.
In step 205, content characteristics of each advertisement in the set of advertisements are identified.
In step 206, a click-through rate estimate value of each advertisement in the advertisement set is determined according to the click-through rate estimation model based on the content characteristics and the content characteristics of each advertisement in the advertisement set.
In step 207, a second advertisement is searched in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set, and the exposure amount of the second advertisement, wherein the difference between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure amount of the second advertisement is larger than a second threshold value.
In step 208, an initial value of the statistical characteristic of the first advertisement is determined based on the historical values of the statistical characteristic of the second advertisement.
According to the technical scheme provided by the embodiment of the disclosure, the click rate estimation model based on the content characteristics is determined according to the content characteristics and the click rate of the advertisement sample, the click rate estimation model based on the content characteristics searches for the old advertisement which is close enough to the click rate estimation value of the new advertisement in the advertisement set, and the initial value of the statistical characteristics of the new advertisement is determined according to the historical value of the statistical characteristics of the old advertisement, so that the exposure opportunity of the new advertisement can be improved, the cold start problem of displaying the advertisement is solved, and the problem that the flow income and the cold start speed are difficult to balance in the related technology is avoided.
The following are embodiments of the disclosed apparatus that may be used to perform embodiments of the disclosed methods.
FIG. 3 is a block diagram illustrating an initial value determination apparatus according to an exemplary embodiment; the apparatus may be implemented in various ways, for example with all components of the apparatus being implemented in a server or with components of the apparatus being implemented in a coupled manner on the server side; the apparatus may implement the method related to the present disclosure through software, hardware, or a combination of the two, as shown in fig. 3, the initial value determining apparatus includes: a first identification module 301, a first determination module 302, a lookup module 303, and a second determination module 304, wherein:
the first identification module 301 is configured to identify a content characteristic of the first advertisement;
the first determining module 302 is configured to determine a click-through rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-obtained click-through rate pre-evaluation model based on the content characteristics;
the searching module 303 is configured to search for a second advertisement in the advertisement set, where a difference value between the click rate pre-evaluation value of the first advertisement and the click rate pre-evaluation value of the first advertisement is smaller than a first threshold value, according to the click rate pre-evaluation model based on the content features;
the second determination module 304 is configured to determine an initial value of the statistical characteristic of the first advertisement based on the historical values of the statistical characteristic of the second advertisement.
The apparatus provided in the embodiment of the present disclosure can be used to implement the technical solution of the embodiment shown in fig. 1a, and the implementation manner and the beneficial effects are similar, and are not described herein again.
In a possible implementation, as shown in fig. 4, the initial value determining apparatus shown in fig. 3 may further include a lookup module 303 configured to include: an identification submodule 401, a determination submodule 402 and a search submodule 403, wherein:
the identifying sub-module 401 is configured to identify content characteristics of each advertisement in the set of advertisements;
the determining submodule 402 is configured to determine a click rate pre-evaluation value of each advertisement in the advertisement set according to the click rate pre-evaluation model based on the content characteristics and the content characteristics of each advertisement in the advertisement set;
the searching sub-module 403 is configured to search for a second advertisement in the advertisement set according to the estimated click-through rate value of the first advertisement, the estimated click-through rate value of each advertisement in the advertisement set, and the exposure amount, wherein the difference between the estimated click-through rate value of the second advertisement and the estimated click-through rate value of the first advertisement is smaller than a first threshold, and the exposure amount of the second advertisement is larger than a second threshold.
In a possible implementation, as shown in fig. 5, the initial value determining apparatus shown in fig. 3 may further include: a second identification module 501 and a third determination module 502, wherein:
the second identification module 501 is configured to identify content characteristics of the ad sample;
the third determining module 502 is configured to determine a click-through rate estimation model based on the content characteristics according to the content characteristics and the click-through rate of the advertisement sample.
In one possible implementation, the content features include textual information; the first identification module 301 identifies words in the first advertisement using word recognition techniques.
In one possible implementation, the content features include image features; the first identification module 301 extracts image features of the first advertisement using a deep convolutional network.
Fig. 6 is a block diagram illustrating an initial value determination apparatus 600 according to an exemplary embodiment, the initial value determination apparatus 600 is applied to a server, and the initial value determination apparatus 600 includes:
a processor 601;
a memory 602 for storing processor-executable instructions;
wherein the processor 601 is configured to:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value of the click rate estimated value of the first advertisement and the click rate estimated value of the second advertisement is smaller than a first threshold value, according to a click rate estimation model based on content characteristics;
an initial value of the statistical characteristic of the first advertisement is determined based on the historical value of the statistical characteristic of the second advertisement.
In one embodiment, the processor 601 may be further configured to:
identifying content characteristics of each advertisement in the advertisement set;
determining click rate pre-evaluation values of the advertisements in the advertisement set according to the click rate pre-evaluation model based on the content characteristics and the content characteristics of the advertisements in the advertisement set;
and searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
In one embodiment, the processor 601 may be further configured to:
identifying content characteristics of an advertising sample;
and determining a click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
In one embodiment, the content features include textual information; the processor 601 may be further configured to: words in the first advertisement are identified using a word recognition technique.
In one embodiment, the content features include image features; the processor 601 may be further configured to: an image feature of the first advertisement is extracted using a deep convolutional network.
In one embodiment, the statistical features include at least any one or combination of: exposure, click rate, or download volume.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
FIG. 7 is a block diagram illustrating an apparatus in accordance with an example embodiment. For example, the apparatus 700 may be provided as a server. The apparatus 700 includes a processing component 702 that further includes one or more processors, and memory resources, represented by memory 703, for storing instructions, such as application programs, that are executable by the processing component 702. The application programs stored in memory 703 may include one or more modules that each correspond to a set of instructions. Further, the processing component 702 is configured to execute instructions to perform the above-described methods.
The apparatus 700 may also include a power component 706 configured to perform power management of the apparatus 700, a wired or wireless network interface 705 configured to connect the apparatus 700 to a network, and an input output (I/O) interface 708. The apparatus 700 may operate based on an operating system stored in memory 703, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, or the like.
A non-transitory computer readable storage medium, instructions in the storage medium, when executed by a processor of an apparatus 700, enable the apparatus 700 to perform a method of:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value of the click rate estimated value of the first advertisement and the click rate estimated value of the second advertisement is smaller than a first threshold value, according to a click rate estimation model based on content characteristics;
an initial value of the statistical characteristic of the first advertisement is determined based on the historical value of the statistical characteristic of the second advertisement.
In one embodiment, according to the click-through rate prediction model based on the content features, searching for a second advertisement in the advertisement set, wherein the difference between the click-through rate prediction value of the first advertisement and the click-through rate prediction value of the second advertisement is smaller than a first threshold value, comprises:
identifying content characteristics of each advertisement in the advertisement set;
determining click rate pre-evaluation values of the advertisements in the advertisement set according to the click rate pre-evaluation model based on the content characteristics and the content characteristics of the advertisements in the advertisement set;
and searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
In one embodiment, the method further comprises:
identifying content characteristics of an advertising sample;
and determining a click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
In one embodiment, the content features include textual information;
identifying content characteristics of the first advertisement includes: words in the first advertisement are identified using a word recognition technique.
In one embodiment, the content features include image features;
identifying content characteristics of the first advertisement includes: an image feature of the first advertisement is extracted using a deep convolutional network.
In one embodiment, the statistical features include at least any one or combination of: exposure, click rate, or download volume.
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (11)
1. An initial value determination method, comprising:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimated value of the first advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimated model based on the content characteristics;
determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement;
the searching for the second advertisement in the advertisement set, in which the difference value between the click-through rate pre-evaluation value of the first advertisement and the click-through rate pre-evaluation value of the first advertisement is smaller than a first threshold value according to the click-through rate pre-evaluation model based on the content characteristics includes:
identifying content characteristics of each advertisement in the set of advertisements;
determining click rate pre-estimated values of all advertisements in the advertisement set according to the click rate pre-estimated model based on the content characteristics and the content characteristics of all advertisements in the advertisement set;
and searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
2. The method of claim 1, further comprising:
identifying content characteristics of an advertising sample;
and determining the click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
3. The method of claim 1, wherein the content features comprise textual information;
the identifying content characteristics of the first advertisement includes: words in the first advertisement are identified using a word recognition technique.
4. The method of claim 1, wherein the content features comprise image features;
the identifying content characteristics of the first advertisement includes: extracting image features of the first advertisement using a deep convolutional network.
5. The method according to any one of claims 1 to 4, wherein the statistical features comprise at least any one or a combination of: exposure, click rate, or download volume.
6. An initial value determination device, comprising:
a first identification module for identifying content characteristics of a first advertisement;
the first determining module is used for determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
the searching module is used for searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimation value of the first advertisement and the click rate pre-estimation value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimation model based on the content characteristics;
the second determining module is used for determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement;
the search module comprises:
the identification submodule is used for identifying the content characteristics of each advertisement in the advertisement set;
the determining submodule is used for determining the click rate pre-estimated value of each advertisement in the advertisement set according to the click rate pre-estimation model based on the content characteristics and the content characteristics of each advertisement in the advertisement set;
and the searching submodule is used for searching a second advertisement in the advertisement set according to the click rate estimated value of the first advertisement, the click rate estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate estimated value of the second advertisement and the click rate estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
7. The apparatus of claim 6, further comprising:
the second identification module is used for identifying the content characteristics of the advertisement samples;
and the third determining module is used for determining the click rate estimation model based on the content characteristics according to the content characteristics and the click rate of the advertisement sample.
8. The apparatus of claim 6, wherein the content features comprise textual information; the first identification module identifies words in the first advertisement using word recognition techniques.
9. The apparatus of claim 6, wherein the content features comprise image features; the first identification module extracts image features of the first advertisement using a deep convolutional network.
10. An initial value determination device, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to:
identifying a content characteristic of the first advertisement;
determining a click rate pre-evaluation value of the first advertisement according to the content characteristics of the first advertisement and a pre-acquired click rate pre-evaluation model based on the content characteristics;
searching a second advertisement in the advertisement set, wherein the difference value between the click rate pre-estimated value of the first advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, according to the click rate pre-estimated model based on the content characteristics;
determining an initial value of the statistical characteristic of the first advertisement according to the historical value of the statistical characteristic of the second advertisement;
the searching for the second advertisement in the advertisement set, in which the difference value between the click-through rate pre-evaluation value of the first advertisement and the click-through rate pre-evaluation value of the first advertisement is smaller than a first threshold value according to the click-through rate pre-evaluation model based on the content characteristics includes:
identifying content characteristics of each advertisement in the set of advertisements;
determining click rate pre-estimated values of all advertisements in the advertisement set according to the click rate pre-estimated model based on the content characteristics and the content characteristics of all advertisements in the advertisement set;
and searching a second advertisement in the advertisement set according to the click rate pre-estimated value of the first advertisement, the click rate pre-estimated value of each advertisement in the advertisement set and the exposure, wherein the difference value between the click rate pre-estimated value of the second advertisement and the click rate pre-estimated value of the first advertisement is smaller than a first threshold value, and the exposure of the second advertisement is larger than a second threshold value.
11. A computer-readable storage medium having stored thereon computer instructions, which, when executed by a processor, carry out the steps of the method according to any one of claims 1 to 5.
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 CN108510326A (en) | 2018-09-07 |
| CN108510326B true 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) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN110489644B (en) * | 2019-08-13 | 2024-06-25 | 腾讯科技（北京）有限公司 | Information pushing method, information pushing 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108510326A (en) | 2018-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107679039B (en) | Method and device for determining statement intention | |
| US20190392487A1 (en) | System, Device, and Method of Automatic Construction of Digital Advertisements | |
| CN108510326B (en) | Initial value determination method and device | |
| CN113239275B (en) | Information pushing method, device, electronic equipment and storage medium | |
| CN114036398B (en) | Content recommendation and ranking model training method, device, equipment and storage medium | |
| CN110765973B (en) | Account type identification method and device | |
| CN111274485B (en) | Personalized recommendation method based on community discovery | |
| CN113469298B (en) | Model training method and resource recommendation method | |
| CN111400586A (en) | Group display method, terminal, server, system and storage medium | |
| CN107291774B (en) | Error sample identification method and device | |
| US10853417B2 (en) | Generating a platform-based representative image for a digital video | |
| CN111144936B (en) | Similar crowd expansion method and device based on user labels | |
| CN111897950A (en) | Method and apparatus for generating information | |
| CN114881685A (en) | Advertisement delivery method, device, electronic device and storage medium | |
| CN110737820B (en) | Method and apparatus for generating event information | |
| CN112906800B (en) | Image group self-adaptive collaborative saliency detection method | |
| CN112148958A (en) | Method, apparatus, and computer storage medium for information recommendation | |
| CN107368923B (en) | Scenic spot heat prediction method and device | |
| CN108734498B (en) | Advertisement pushing method and device | |
| CN113705683B (en) | Training method and device of recommendation model, electronic equipment and storage medium | |
| US20220358347A1 (en) | Computerized system and method for distilled deep prediction for personalized stream ranking | |
| CN115098729A (en) | Video processing method, sample generation method, model training method and device | |
| CN112905892A (en) | Big data processing method and big data server applied to user portrait mining | |
| CN112446738A (en) | Advertisement data processing method, device, medium and electronic equipment | |
| CN112200602A (en) | Neural network model training method and device for advertisement recommendation |
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 |