WO2020134009A1 - Click rate prediction method and apparatus, and display position selection method and apparatus - Google Patents

Click rate prediction method and apparatus, and display position selection method and apparatus Download PDF

Info

Publication number
WO2020134009A1
WO2020134009A1 PCT/CN2019/094738 CN2019094738W WO2020134009A1 WO 2020134009 A1 WO2020134009 A1 WO 2020134009A1 CN 2019094738 W CN2019094738 W CN 2019094738W WO 2020134009 A1 WO2020134009 A1 WO 2020134009A1
Authority
WO
WIPO (PCT)
Prior art keywords
display information
booth
target
prediction model
click rate
Prior art date
Application number
PCT/CN2019/094738
Other languages
French (fr)
Chinese (zh)
Inventor
李亦锬
余林韵
陈嘉闽
黄训蓬
李磊
Original Assignee
北京字节跳动网络技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 北京字节跳动网络技术有限公司 filed Critical 北京字节跳动网络技术有限公司
Publication of WO2020134009A1 publication Critical patent/WO2020134009A1/en

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
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • 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
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • 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
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/01—Social networking

Definitions

  • the present disclosure relates to the field of computer technology, and in particular, to click-through rate prediction and booth selection.
  • Information flow is to present a series of information to the user when the user loads the display interface, which is one of the most important innovations in the social media mobile application.
  • the series of information is a series of display information, and each display information can produce a certain display effect, such as being clicked.
  • the higher the click rate of the display information the more the number of times the display information is clicked, that is, the more popular the display information, so that the display information can obtain greater benefits. Therefore, for a certain display information, it should be possible to get a higher CTR. In different booths, the effect of the same display information may be different. Therefore, it is necessary to predict the effect (for example, click rate).
  • the prediction of the click-through rate often relies on the calculation of the similarity between a large number of booths and booths, and between display information and display information. The amount of data is large and the calculation efficiency is low.
  • a click rate prediction method including:
  • the target display information is input to a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
  • the target prediction model is constructed in the following manner:
  • the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
  • the first training sample further includes the target display information and an actual click rate of the target display information corresponding to a second booth, the second booth is a different booth from the first booth.
  • the target prediction model is constructed in the following manner:
  • the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
  • the second training sample further includes the target display information and an actual click rate of the target display information corresponding to a third booth, the third booth is a booth different from the first booth.
  • a booth selection method comprising:
  • the candidate booth with the highest predicted click rate is determined as the target booth to be selected.
  • a click rate prediction device comprising:
  • the first determining module is used to determine the first booth corresponding to the target display information
  • the prediction module is configured to input the target display information into a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
  • the target prediction model is constructed in the following manner:
  • the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
  • the first training sample further includes the target display information and an actual click rate of the target display information corresponding to a second booth, the second booth is a different booth from the first booth.
  • the target prediction model is constructed in the following manner:
  • the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
  • the second training sample further includes the target display information and an actual click rate of the target display information corresponding to a third booth, the third booth is a booth different from the first booth.
  • a booth selection device comprising:
  • the second determination module is used to determine the candidate booth corresponding to the target display information
  • the click rate prediction device is used to determine the predicted click rate corresponding to each candidate booth
  • the third determining module is used to determine the candidate booth with the highest predicted click rate as the target booth to be selected.
  • a computer-readable storage medium on which a computer program is stored, which when executed by a processor implements the method according to the first aspect of the present disclosure, or the program is executed by a processor To implement the method according to the second aspect of the present disclosure.
  • an electronic device including:
  • the processor is configured to execute the computer program in the memory to implement the method according to the first aspect of the present disclosure, or implement the method according to the second aspect of the present disclosure.
  • a prediction model may be used to predict the possible click rate of the exhibition information at the first booth, so as to know whether the first booth is suitable for the exhibition information.
  • FIG. 1 is a flowchart of a click rate prediction method according to an exemplary embodiment of the present disclosure
  • FIG. 2 is a flowchart of a booth selection method according to an exemplary embodiment of the present disclosure
  • FIG. 3 is a block diagram of a click rate prediction apparatus according to an exemplary embodiment of the present disclosure.
  • FIG. 4 is a block diagram of a booth selection device according to an exemplary embodiment of the present disclosure.
  • FIG. 5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
  • FIG. 6 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
  • FIG. 1 is a flowchart of a click rate prediction method according to an exemplary embodiment of the present disclosure. As shown in FIG. 1, the method may include the following steps 11 to 12.
  • step 11 the first booth corresponding to the target display information is determined.
  • the presentation information may be, for example, text, pictures, videos, or a combination thereof.
  • the booth can be a virtual or actual carrier for displaying the display information, and the display information can be displayed in the booth to obtain the corresponding click-through rate, thereby creating a certain value.
  • click rate prediction can be performed on the existing display information
  • the target display information is the display information of the click rate to be predicted
  • the predicted click rate is the time when the target display information is displayed in the first booth Click through rate to make predictions. Therefore, the first booth corresponding to the target display information can be determined.
  • step 12 the target display information is input to the target prediction model corresponding to the first booth to obtain the predicted click rate corresponding to the target display information.
  • the target prediction model can be constructed in the following ways:
  • the first training sample may include first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information.
  • the first historical display information may be all display information displayed on the first booth, and the actual click rate corresponding to the first historical display information is actual clicks corresponding to all display information displayed on the first booth rate.
  • the first prediction model may be trained according to the first training sample to obtain the target prediction model.
  • the first prediction model can be trained by machine learning algorithms (eg, neural network learning) to obtain the target prediction model.
  • a neural network learning method can be used to obtain a target prediction model.
  • the process of constructing the target prediction model for the supervised neural network training method will be described in detail below, but the method provided by the present disclosure is not limited to this learning method, and is not limited to this training method, the following examples Just as an example.
  • the connection weight of the neural network in the first prediction model may be random.
  • the first historical information is input into the first prediction model.
  • the hidden layer of the first prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently.
  • the above operation is performed on each first historical display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold.
  • the gap between the actual output of the first prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current first prediction model can reach a certain accuracy, so the current first prediction model can be determined It is the target prediction model.
  • the target prediction model may be updated in real time, that is, in the process of using the target prediction model, each time an additional display information is displayed on the first booth, after the display information is displayed,
  • the actual click rate corresponding to the impression will be collected, and the display information corresponding to the actual impression (as the newly added first historical impression information) and the actual click rate will be used as new training data to continuously modify the target prediction model To make the target decision model more and more accurate.
  • the first prediction model can be trained using the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information to obtain the target prediction model. Since the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information are easy to obtain, it is convenient for data collection, and at the same time it is convenient to obtain the target prediction model.
  • the first training sample may also include target display information and target display The actual click rate of the information in the second booth.
  • the second booth is different from the first booth. For example, there may be one second booth, or there may be multiple second booths.
  • the target display information is only displayed in the second booth inside the station, but not in the first booth outside the station, you can collect the target display information and the target display information in the second booth Corresponding actual click-through rate.
  • the first historical display information and/or target display information in the training sample may be input into the first prediction model, and the model may be adjusted according to the gap between the actual output of the first prediction model and the expected output The right to connect internal neural networks.
  • the display information in the training sample can be input into the first prediction model one by one, that is, each time a first historical display information or target display information is input into the first prediction model, according to the actual output of the first prediction model (first prediction The model adjusts the intra-model nerve for the difference between the predicted click rate generated by the input first historical display information or target display information and the expected output (the actual click rate corresponding to the input first historical display information or target display information) The right to connect to the network.
  • the connection weight of the neural network in the first prediction model may be random.
  • the first historical information is input into the first prediction model.
  • the first historical information is a video
  • the video is input to the first prediction model.
  • the hidden layer of the first prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently.
  • the target prediction model may be updated in real time, that is, during the use of the target prediction model, each additional display information is displayed on the first booth, or each additional target display information Show once in the second booth, after the show, the actual click rate corresponding to the impression will be collected, and the display information and actual click rate corresponding to the actual impression will be used as new training data, and the target prediction model will be continuously carried out. Revise to make the target decision model more and more accurate.
  • the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information and the actual click rate of the target display information and target display information corresponding to the second booth can be used to perform the first prediction model Train to get the target prediction model.
  • the target prediction model can be obtained in combination with the relevant information of the first booth and the target display information, so that the obtained target prediction model is more accurate.
  • the target model can be constructed in the following manner:
  • the target display information may exist in the corresponding target display period. Therefore, the prediction of the click rate is to predict the click rate of the target display information in the target display period in the first booth.
  • the display period here refers to a certain period of the day.
  • the target display information may need to be displayed in multiple different display periods. Therefore, the target display period corresponding to the current display of the target display information can be determined.
  • the second training sample may include second historical display information corresponding to the first booth at the target display period and actual click rate corresponding to the second historical display information.
  • the second historical display information may be all display information displayed by the first booth during the target display period
  • the actual click rate corresponding to the second historical display information is all display information displayed by the first booth during the target display period The corresponding click-through rate.
  • the second prediction model may be trained according to the second training sample to obtain the target prediction model.
  • the second prediction model may be trained by machine learning algorithms (eg, neural network learning) to obtain the target prediction model.
  • a neural network learning method can be used to obtain a target prediction model.
  • the process of constructing the target prediction model for the supervised neural network training method will be described in detail below, but the method provided by the present disclosure is not limited to this learning method, and is not limited to this training method, the following examples Just as an example.
  • the connection weight of the neural network in the second prediction model may be random.
  • the second historical information is input into the second prediction model.
  • the hidden layer of the second prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently.
  • the above operation is performed on each second historical display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold.
  • the gap between the actual output of the second prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current second prediction model can reach a certain accuracy, so the current second prediction model can be determined It is the target prediction model.
  • the target prediction model may be updated in real time, that is to say, during the use of the target prediction model, the first booth exhibits one more type of display information during the target display period, and the display information After the impression, the actual click rate corresponding to the impression will be collected, and the impression information corresponding to the actual impression (as the newly added second historical impression information) and the actual click rate will be used as new training data to continuously predict the target
  • the model is revised to make the target decision model more and more accurate.
  • the second prediction model can be trained using the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information to obtain the target prediction model. Since the second historical display information corresponding to the target exhibition period and the actual click rate corresponding to the second historical display information of the first booth are easy to obtain, it is convenient for data collection to obtain the target prediction model, and at the same time, the model is accurate to the time period, so that the model is targeted Sex is stronger.
  • the second training sample may include target display information in addition to the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information And the actual click rate of the target display information in the third booth.
  • the third booth is different from the first booth. For example, there may be one third booth, or there may be multiple third booths.
  • the target display information is only displayed in the third booth in the station, but not in the first booth outside the station, you can collect the target display information and the target display information in the third booth Corresponding actual click-through rate.
  • the actual click rate of the target display information corresponding to the third booth may be the actual click rate of the target display information corresponding to the third booth during the target display period, or it may not be limited to the target display period.
  • the second historical display information and/or target display information in the training sample may be input into the second prediction model, and adjusted according to the gap between the actual output and the expected output of the second prediction model
  • the connection weight of the neural network in the model can be input into the second prediction model one by one, that is, each time a second historical display information or target display information is input into the second prediction model, according to the actual output of the second prediction model (second prediction).
  • the model adjusts the intra-model nerve for the gap between the predicted click rate generated by the input second historical display information or target display information and the expected output (the actual click rate corresponding to the input second historical display information or target display information) The right to connect to the network.
  • the connection weight of the neural network in the second prediction model may be random.
  • the second historical information is input into the second prediction model.
  • the second historical information is a video
  • the video is input to the second prediction model.
  • the hidden layer of the second prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently.
  • the target prediction model may be updated in real time, that is to say, during the use of the target prediction model, the first booth exhibits one more type of display information during the target display period, or, the target Every time the impression information is displayed in the third booth, after the corresponding impression information is displayed, the actual click rate corresponding to the impression will be collected, and the impression information and actual click rate corresponding to the actual impression will be used as new training Data, and constantly modify the target prediction model to make the target decision model more and more accurate.
  • the second historical display information corresponding to the first booth and the actual click rate corresponding to the second historical display information and the actual click rate corresponding to the target display information and target display information at the third booth can be used A prediction model is trained to obtain the target prediction model.
  • the target prediction model can be obtained by combining the relevant information of the first booth and the target display information, and the prediction model is related to the display period, so that the obtained target prediction model is more accurate and targeted.
  • the prediction model may be used to predict the possible click rate of the display information in the first booth, so as to know whether the first booth is suitable for the display information.
  • FIG. 2 is a flowchart of a booth selection method according to an exemplary embodiment of the present disclosure. As shown in FIG. 2, the method may include the following steps 21 to 23.
  • step 21 the candidate booth corresponding to the target display information is determined.
  • each display information may correspond to a corresponding content template.
  • the content template may be used to limit the content types of the display information, such as popular science content and funny content, and the booth can display the display information of a specific content template, so ,
  • the candidate booth corresponding to the target display information can be determined according to the content template.
  • each display information may correspond to the type of display information, such as image, video, text, or a combination thereof.
  • a booth can display a specific type of display information. Therefore, it can be determined according to the type of display information and the target display Candidate booth corresponding to the information.
  • the candidate booth corresponding to the target booth information may be a booth that can display video-type display information.
  • step 22 the predicted click rate corresponding to each candidate booth is determined.
  • click rate prediction can be performed for each candidate booth.
  • the specific embodiment of the click rate prediction has been described above, and will not be repeated here.
  • step 23 the candidate booth with the highest predicted click rate is determined as the target booth.
  • the target booth is the candidate booth that is most suitable for displaying the target display information.
  • the predicted click rate of the target display information displayed to each candidate booth can be obtained, so that an appropriate booth can be selected according to the predicted click rate, so that the target display information displayed to the target booth can obtain the highest click rate.
  • FIG. 3 is a block diagram of a click rate prediction apparatus according to an exemplary embodiment of the present disclosure. As shown in FIG. 3, the device 30 may include:
  • the first determining module 31 is used to determine the first booth corresponding to the target display information
  • the prediction module 32 is configured to input the target display information into the target prediction model corresponding to the first booth to obtain the predicted click rate corresponding to the target display information.
  • the target prediction model is constructed in the following manner:
  • the first training sample at least includes the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information;
  • the first training sample further includes target display information and the actual click rate of the target display information corresponding to the second booth, the second booth is a different booth from the first booth.
  • the target prediction model is constructed in the following manner:
  • the second training sample includes at least the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information;
  • the second training sample further includes target display information and actual click rate of the target display information corresponding to the third booth, the third booth is a different booth from the first booth.
  • the device 40 may include:
  • the second determination module 41 is used to determine the candidate booth corresponding to the target display information
  • the click rate prediction device 42 is used to determine the predicted click rate corresponding to each candidate booth;
  • the third determining module 43 is configured to determine the candidate booth with the highest predicted click rate as the target booth to be selected.
  • FIG. 5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
  • the electronic device 500 may be provided as a server. 5
  • the electronic device 500 may include: one or more processors 522; and a memory 532 for storing computer programs executable by the processor 522.
  • the computer program stored in the memory 532 may include each module corresponding to a set of instructions.
  • the processor 522 may be configured to execute the computer program to perform the above-mentioned click rate prediction method.
  • the electronic device 500 may further include a power component 526 and a communication component 550, which may be configured to perform power management of the electronic device 500, and the communication component 550 may be configured to implement communication of the electronic device 500, for example, wired Or wireless communication.
  • the electronic device 500 may also include an input/output (I/O) interface 558.
  • the electronic device 500 can operate based on an operating system stored in the memory 532, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM, and so on.
  • a computer-readable storage medium including program instructions is also provided.
  • the program instructions are executed by a processor, the above-mentioned click-through rate prediction method is implemented.
  • the computer-readable storage medium may be the above-mentioned memory 532 including program instructions, and the above-mentioned program instructions may be executed by the processor 522 of the electronic device 500 to implement the above-mentioned click rate prediction method.
  • the electronic device 1900 may be provided as a server. 6, the electronic device 1900 may include: one or more processors 1922; and a memory 1932 for storing computer programs executable by the processor 1922.
  • the computer program stored in the memory 1932 may include each module corresponding to a set of instructions.
  • the processor 1922 may be configured to execute the computer program to implement the booth selection method described above.
  • the electronic device 1900 may further include a power supply component 1926 and a communication component 1950, which may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired Or wireless communication.
  • the electronic device 1900 may also include an input/output (I/O) interface 1958.
  • the electronic device 1900 can operate an operating system based on the memory 1932, such as Windows Server TM , Mac OS X TM , Unix TM , Linux TM, and so on.
  • a computer-readable storage medium including program instructions is also provided.
  • the program instructions are executed by a processor, the above booth selection method is implemented.
  • the computer-readable storage medium may be the aforementioned memory 1932 including program instructions, which may be executed by the processor 1922 of the electronic device 1900 to implement the aforementioned booth selection method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Quality & Reliability (AREA)
  • Primary Health Care (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Electrically Operated Instructional Devices (AREA)

Abstract

A click rate prediction method and apparatus, and a display position selection method and apparatus. The click rate prediction method comprises: determining a first display position corresponding to target display information (11); and inputting the target display information into a target prediction model corresponding to the first display position to obtain a predicted click rate corresponding to the target display information (12). In this way, the click rate, which may be obtained in the first display position, of the display information can be predicted by means of the prediction model before the display information is displayed on the first display position, so that whether the first display position is suitable for the display information can be known.

Description

点击率预测方法和装置，以及展位选择方法和装置Click rate prediction method and device, and booth selection method and device
相关申请的交叉引用Cross-reference of related applications
本公开要求于2018年12月29日在中国国家知识产权局提交的申请号为201811641964.9的中国专利申请的权益，该中国专利申请公开的内容通过引用整体并入本文。This disclosure requires the rights and interests of the Chinese patent application with the application number 201811641964.9 filed at the State Intellectual Property Office of China on December 29, 2018. The content of this Chinese patent application is incorporated herein by reference in its entirety.
技术领域Technical field
本公开涉及计算机技术领域，具体地，涉及点击率预测和展位选择。The present disclosure relates to the field of computer technology, and in particular, to click-through rate prediction and booth selection.
背景技术Background technique
信息流(Feeds)是当用户加载显示界面时为用户呈现一系列信息，其是社交媒体移动应用中最重要的创新点之一。该一系列信息就是一系列展示信息，每条展示信息都可以产生一定的展示效果，例如被点击。展示信息的点击率越高，说明该展示信息被点击的次数越多，即该展示信息越受欢迎，从而可以通过该展示信息获得更大的利益。因此，针对某展示信息，应当尽可能使其获得更高的点击率。在不同的展位上，同一展示信息得到的效果可能是不同的，因此，需要对效果(例如，点击率)进行预测。在现有技术中，对于点击率的预测常常依赖于大量的展位与展位之间、展示信息与展示信息之间的相似度计算，数据量大且计算效率低。Information flow (Feeds) is to present a series of information to the user when the user loads the display interface, which is one of the most important innovations in the social media mobile application. The series of information is a series of display information, and each display information can produce a certain display effect, such as being clicked. The higher the click rate of the display information, the more the number of times the display information is clicked, that is, the more popular the display information, so that the display information can obtain greater benefits. Therefore, for a certain display information, it should be possible to get a higher CTR. In different booths, the effect of the same display information may be different. Therefore, it is necessary to predict the effect (for example, click rate). In the prior art, the prediction of the click-through rate often relies on the calculation of the similarity between a large number of booths and booths, and between display information and display information. The amount of data is large and the calculation efficiency is low.
发明内容Summary of the invention
根据本公开的第一方面，提供一种点击率预测方法，所述方法包括：According to a first aspect of the present disclosure, a click rate prediction method is provided, the method including:
确定与目标展示信息对应的第一展位；Determine the first booth corresponding to the target display information;
将所述目标展示信息输入至与所述第一展位对应的目标预测模型，以获得所述目标展示信息对应的预测点击率。The target display information is input to a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
可选地，所述目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
获取第一训练样本，所述第一训练样本至少包括所述第一展位对应的第一历史展示信息以及与所述第一历史展示信息对应的实际点击率；Obtaining a first training sample, the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
根据所述第一训练样本对第一预测模型进行训练，以获得所述目标预测模型。Training the first prediction model according to the first training sample to obtain the target prediction model.
可选地，所述第一训练样本还包括所述目标展示信息以及所述目标展示信息在第二展位对应的实际点击率，所述第二展位为与所述第一展位不同的展位。Optionally, the first training sample further includes the target display information and an actual click rate of the target display information corresponding to a second booth, the second booth is a different booth from the first booth.
可选地，所述目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
确定与所述目标展示信息对应的目标展示时段；Determine a target display period corresponding to the target display information;
获取第二训练样本，所述第二训练样本至少包括所述第一展位在所述目标展示时段对应的第二历史展示信息以及与所述第二历史展示信息对应的实际点击率；Acquiring a second training sample, the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
根据所述第二训练样本对第二预测模型进行训练，以获得所述目标预测模型。Training the second prediction model according to the second training sample to obtain the target prediction model.
可选地，所述第二训练样本还包括所述目标展示信息以及所述目标展示信息在第三展位对应的实际点击率，所述第三展位为与所述第一展位不同的展位。Optionally, the second training sample further includes the target display information and an actual click rate of the target display information corresponding to a third booth, the third booth is a booth different from the first booth.
根据本公开的第二方面，提供一种展位选择方法，所述方法包括：According to a second aspect of the present disclosure, there is provided a booth selection method, the method comprising:
确定与目标展示信息对应的候选展位；Identify the candidate booth corresponding to the target display information;
根据本公开第一方面所述的点击率预测方法确定每个所述候选展位对应的预测点击率；Determining the predicted click rate corresponding to each candidate booth according to the click rate prediction method described in the first aspect of the present disclosure;
将预测点击率最高的候选展位确定为待选择的目标展位。The candidate booth with the highest predicted click rate is determined as the target booth to be selected.
根据本公开的第三方面，提供一种点击率预测装置，所述装置包括：According to a third aspect of the present disclosure, there is provided a click rate prediction device, the device comprising:
第一确定模块，用于确定与目标展示信息对应的第一展位；The first determining module is used to determine the first booth corresponding to the target display information;
预测模块，用于将所述目标展示信息输入至与所述第一展位对应的目标预测模型，以获得所述目标展示信息对应的预测点击率。The prediction module is configured to input the target display information into a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
可选地，所述目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
获取第一训练样本，所述第一训练样本至少包括所述第一展位对应的第一历史展示信息以及与所述第一历史展示信息对应的实际点击率；Obtaining a first training sample, the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
根据所述第一训练样本对第一预测模型进行训练，以获得所述目标预测模型。Training the first prediction model according to the first training sample to obtain the target prediction model.
可选地，所述第一训练样本还包括所述目标展示信息以及所述目标展示信息在第二展位对应的实际点击率，所述第二展位为与所述第一展位不同的展位。Optionally, the first training sample further includes the target display information and an actual click rate of the target display information corresponding to a second booth, the second booth is a different booth from the first booth.
可选地，所述目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
确定与所述目标展示信息对应的目标展示时段；Determine a target display period corresponding to the target display information;
获取第二训练样本，所述第二训练样本至少包括所述第一展位在所述目标展示时段对应的第二历史展示信息以及与所述第二历史展示信息对应的实际点击率；Acquiring a second training sample, the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
根据所述第二训练样本对第二预测模型进行训练，以获得所述目标预测模型。Training the second prediction model according to the second training sample to obtain the target prediction model.
可选地，所述第二训练样本还包括所述目标展示信息以及所述目标展示信息在第三展位对应的实际点击率，所述第三展位为与所述第一展位不同的展位。Optionally, the second training sample further includes the target display information and an actual click rate of the target display information corresponding to a third booth, the third booth is a booth different from the first booth.
根据本公开的第四方面，提供一种展位选择装置，所述装置包括：According to a fourth aspect of the present disclosure, there is provided a booth selection device, the device comprising:
第二确定模块，用于确定与目标展示信息对应的候选展位；The second determination module is used to determine the candidate booth corresponding to the target display information;
本公开第三方面所述的点击率预测装置，用于确定每个所述候选展位对应的预测点击率；The click rate prediction device according to the third aspect of the present disclosure is used to determine the predicted click rate corresponding to each candidate booth;
第三确定模块，用于将预测点击率最高的候选展位确定为待选择的目标展位。The third determining module is used to determine the candidate booth with the highest predicted click rate as the target booth to be selected.
根据本公开的第五方面，提供一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现根据本公开的第一方面的方法，或者，该程序被处理器执行时实现根据本公开的第二方面的方法。According to a fifth aspect of the present disclosure, there is provided a computer-readable storage medium on which a computer program is stored, which when executed by a processor implements the method according to the first aspect of the present disclosure, or the program is executed by a processor To implement the method according to the second aspect of the present disclosure.
根据本公开的第六方面，提供一种电子设备，包括：According to a sixth aspect of the present disclosure, an electronic device is provided, including:
存储器，其上存储有计算机程序；Memory, on which computer programs are stored;
处理器，用于执行所述存储器中的所述计算机程序，以实现根据本公开的第一方面的方法，或者，实现根据本公开的第二方面的方法。The processor is configured to execute the computer program in the memory to implement the method according to the first aspect of the present disclosure, or implement the method according to the second aspect of the present disclosure.
根据本公开的实施例，可以在将展示信息展示至第一展位之前，通过 预测模型预测该展示信息在第一展位可能得到的点击率，从而可以获知该第一展位是否适合该展示信息。According to an embodiment of the present disclosure, before displaying the exhibition information to the first booth, a prediction model may be used to predict the possible click rate of the exhibition information at the first booth, so as to know whether the first booth is suitable for the exhibition information.
附图说明BRIEF DESCRIPTION
附图是用来帮助对本公开的进一步理解，并且构成说明书的一部分，与下面的详细描述一起用于解释本公开，但并不构成对本公开的限制。在附图中：The drawings are used to help further understanding of the present disclosure, and constitute a part of the specification, together with the following detailed description to explain the present disclosure, but do not constitute a limitation of the present disclosure. In the drawings:
图1是根据本公开的示例性实施例的点击率预测方法的流程图；FIG. 1 is a flowchart of a click rate prediction method according to an exemplary embodiment of the present disclosure;
图2是根据本公开的示例性实施例的展位选择方法的流程图；2 is a flowchart of a booth selection method according to an exemplary embodiment of the present disclosure;
图3是根据本公开的示例性实施例的点击率预测装置的框图；3 is a block diagram of a click rate prediction apparatus according to an exemplary embodiment of the present disclosure;
图4是根据本公开的示例性实施例的展位选择装置的框图；4 is a block diagram of a booth selection device according to an exemplary embodiment of the present disclosure;
图5是根据本公开的示例性实施例的一种电子设备的框图；5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure;
图6是根据本公开的示例性实施例的一种电子设备的框图。6 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
具体实施方式detailed description
以下结合附图对本公开的具体实施例进行详细说明。应当理解的是，此处所描述的具体实施例仅用于说明和解释本公开，并不用于限制本公开。The specific embodiments of the present disclosure will be described in detail below with reference to the drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure, and are not intended to limit the present disclosure.
需要说明的是，本公开的说明书和权利要求书以及上述附图中的术语“第一”、“第二”、“第三”等是用于区别类似的对象，而不必理解为描述特定的顺序或先后次序。It should be noted that the terms "first", "second", "third", etc. in the specification and claims of the present disclosure and the above drawings are used to distinguish similar objects, and do not have to be understood as describing specific Order or sequence.
图1是根据本公开的示例性实施例的点击率预测方法的流程图。如图1所示，该方法可以包括以下步骤11至12。FIG. 1 is a flowchart of a click rate prediction method according to an exemplary embodiment of the present disclosure. As shown in FIG. 1, the method may include the following steps 11 to 12.
在步骤11中，确定与目标展示信息对应的第一展位。In step 11, the first booth corresponding to the target display information is determined.
示例地，展示信息可以例如是文字、图片、视频或其组合等。展位可以为用于对展示信息进行展示的虚拟或实际的载体，展示信息可以在展位进行展示以获得相应的点击率，从而创造一定的价值。利用本公开提供的方法，可以针对已有的展示信息进行点击率预测，目标展示信息就是待预测点击率的展示信息，而这里预测的点击率则为对目标展示信息在第一展 位展示时的点击率进行预测。因此，可以确定与目标展示信息对应的第一展位。Illustratively, the presentation information may be, for example, text, pictures, videos, or a combination thereof. The booth can be a virtual or actual carrier for displaying the display information, and the display information can be displayed in the booth to obtain the corresponding click-through rate, thereby creating a certain value. By using the method provided by the present disclosure, click rate prediction can be performed on the existing display information, the target display information is the display information of the click rate to be predicted, and the predicted click rate here is the time when the target display information is displayed in the first booth Click through rate to make predictions. Therefore, the first booth corresponding to the target display information can be determined.
在步骤12中，将目标展示信息输入至与第一展位对应的目标预测模型，以获得目标展示信息对应的预测点击率。In step 12, the target display information is input to the target prediction model corresponding to the first booth to obtain the predicted click rate corresponding to the target display information.
在一种可能的实施例中，目标预测模型可以通过以下方式构建：In a possible embodiment, the target prediction model can be constructed in the following ways:
获取第一训练样本；Obtain the first training sample;
根据第一训练样本对第一预测模型进行训练，以获得目标预测模型。Train the first prediction model according to the first training sample to obtain the target prediction model.
在一种可能的实施例中，第一训练样本可以包括第一展位对应的第一历史展示信息以及与第一历史展示信息对应的实际点击率。其中，第一历史展示信息可以为在第一展位上展示过的所有展示信息，与第一历史展示信息对应的实际点击率则为在第一展位上展示过的所有展示信息各自对应的实际点击率。In a possible embodiment, the first training sample may include first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information. The first historical display information may be all display information displayed on the first booth, and the actual click rate corresponding to the first historical display information is actual clicks corresponding to all display information displayed on the first booth rate.
在获取到第一训练样本后，可以根据第一训练样本对第一预测模型进行训练，以获得目标预测模型。示例地，可以通过机器学习算法(例如，神经网络学习)对第一预测模型进行训练，以获得目标预测模型。After obtaining the first training sample, the first prediction model may be trained according to the first training sample to obtain the target prediction model. Illustratively, the first prediction model can be trained by machine learning algorithms (eg, neural network learning) to obtain the target prediction model.
示例地，可以利用神经网络学习方法获得目标预测模型。下面将针对有监督的神经网络训练方法构建目标预测模型的过程进行详细说明，但是本公开所提供的方法并不局限于此学习方法，并且，也不局限于这一种训练方式，下述实例仅作为示例性说明。Illustratively, a neural network learning method can be used to obtain a target prediction model. The process of constructing the target prediction model for the supervised neural network training method will be described in detail below, but the method provided by the present disclosure is not limited to this learning method, and is not limited to this training method, the following examples Just as an example.
将训练样本中的一个第一历史展示信息输入第一预测模型，根据第一预测模型的实际输出(第一预测模型针对输入的第一历史展示信息所生成的预测点击率)与期望输出(与输入的第一历史展示信息对应的实际点击率)间的差距来调整模型内神经网络的连接权。初始情况下，第一预测模型内神经网络的连接权可以是随机的。其中，将第一历史信息输入第一预测模型。举例来说，若第一历史信息为图片，则将该图片输入第一预测模型。示例地，第一预测模型的隐层中可以包含能够对输入的展示信息进行特征提取的层，以获得输入信息的特征，从而可以更加便利地训练预测模型。之后，对训练样本中的每一个第一历史展示信息执行上述操作，直至实际输出与期望输出之间的差距小于预设的差距阈值。当第一预测模型的 实际输出与期望输出之间的差距小于预设的差距阈值时，说明当前的第一预测模型的预测可以达到一定的准确率，因此，可以将当前的第一预测模型确定为目标预测模型。在一种可能的实施例中，目标预测模型可以是实时更新的，也就是说，在目标预测模型使用的过程中，第一展位上每多展示一种展示信息，在该展示信息展示后则会收集该次展示所对应的实际点击率，并将该次实际展示对应的展示信息(作为新增的第一历史展示信息)和实际点击率作为新的训练数据，不断对目标预测模型进行修正，以使目标决策模型越来越精确。Input the first historical display information in the training sample into the first prediction model, according to the actual output of the first prediction model (the predicted click rate generated by the first prediction model for the input first historical display information) and the expected output (and The gap between the actual click rate corresponding to the input first historical display information is used to adjust the connection weight of the neural network in the model. Initially, the connection weight of the neural network in the first prediction model may be random. Among them, the first historical information is input into the first prediction model. For example, if the first historical information is a picture, the picture is input to the first prediction model. Exemplarily, the hidden layer of the first prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently. After that, the above operation is performed on each first historical display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold. When the gap between the actual output of the first prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current first prediction model can reach a certain accuracy, so the current first prediction model can be determined It is the target prediction model. In a possible embodiment, the target prediction model may be updated in real time, that is, in the process of using the target prediction model, each time an additional display information is displayed on the first booth, after the display information is displayed, The actual click rate corresponding to the impression will be collected, and the display information corresponding to the actual impression (as the newly added first historical impression information) and the actual click rate will be used as new training data to continuously modify the target prediction model To make the target decision model more and more accurate.
通过上述方式，可以利用第一展位对应的第一历史展示信息和第一历史展示信息对应的实际点击率对第一预测模型进行训练，以得到目标预测模型。由于第一展位对应的第一历史展示信息和第一历史展示信息对应的实际点击率易于获得，便于进行数据采集，同时便于获得目标预测模型。In the above manner, the first prediction model can be trained using the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information to obtain the target prediction model. Since the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information are easy to obtain, it is convenient for data collection, and at the same time it is convenient to obtain the target prediction model.
在另一种可能的实施例中，第一训练样本除了可以包括第一展位对应的第一历史展示信息以及与第一历史展示信息对应的实际点击率外，还可以包括目标展示信息以及目标展示信息在第二展位对应的实际点击率。其中，第二展位为与第一展位不同的展位。示例地，可以存在一个第二展位，或者，可以存在多个第二展位。示例地，若对于某一网站，目标展示信息仅在站内的第二展位上展示过，而并未在站外的第一展位展示过，则可以收集目标展示信息以及目标展示信息在第二展位对应的实际点击率。In another possible embodiment, in addition to the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information, the first training sample may also include target display information and target display The actual click rate of the information in the second booth. Among them, the second booth is different from the first booth. For example, there may be one second booth, or there may be multiple second booths. For example, if for a website, the target display information is only displayed in the second booth inside the station, but not in the first booth outside the station, you can collect the target display information and the target display information in the second booth Corresponding actual click-through rate.
示例地，在构建目标预测模型时，可以将训练样本中的第一历史展示信息和/或目标展示信息输入第一预测模型，根据第一预测模型的实际输出与期望输出间的差距来调整模型内神经网络的连接权。例如，可以逐个将训练样本中的展示信息输入第一预测模型，也就是每一次将一个第一历史展示信息或目标展示信息输入第一预测模型，根据第一预测模型的实际输出(第一预测模型针对输入的第一历史展示信息或目标展示信息所生成的预测点击率)与期望输出(与输入的第一历史展示信息或目标展示信息对应的实际点击率)间的差距来调整模型内神经网络的连接权。初始情况下，第一预测模型内神经网络的连接权可以是随机的。其中，将第一历史信息输入第一预测模型。举例来说，若第一历史信息为视频，则将该视频 输入第一预测模型。示例地，第一预测模型的隐层中可以包含能够对输入的展示信息进行特征提取的层，以获得输入信息的特征，从而可以更加便利地训练预测模型。之后，对训练样本中的每一个第一历史展示信息或目标展示信息执行上述操作，直至实际输出与期望输出之间的差距小于预设的差距阈值。当第一预测模型的实际输出与期望输出之间的差距小于预设的差距阈值时，说明当前的第一预测模型的预测可以达到一定的准确率，因此，可以将当前的第一预测模型确定为目标预测模型。在一种可能的实施例中，目标预测模型可以是实时更新的，也就是说，在目标预测模型使用的过程中，第一展位上每多展示一种展示信息，或者，目标展示信息每多在第二展位上展示一次，在展示后则会收集该次展示所对应的实际点击率，并将该次实际展示对应的展示信息和实际点击率作为新的训练数据，不断对目标预测模型进行修正，以使目标决策模型越来越精确。For example, when constructing the target prediction model, the first historical display information and/or target display information in the training sample may be input into the first prediction model, and the model may be adjusted according to the gap between the actual output of the first prediction model and the expected output The right to connect internal neural networks. For example, the display information in the training sample can be input into the first prediction model one by one, that is, each time a first historical display information or target display information is input into the first prediction model, according to the actual output of the first prediction model (first prediction The model adjusts the intra-model nerve for the difference between the predicted click rate generated by the input first historical display information or target display information and the expected output (the actual click rate corresponding to the input first historical display information or target display information) The right to connect to the network. Initially, the connection weight of the neural network in the first prediction model may be random. Among them, the first historical information is input into the first prediction model. For example, if the first historical information is a video, the video is input to the first prediction model. Exemplarily, the hidden layer of the first prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently. After that, the above operation is performed on each first historical display information or target display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold. When the gap between the actual output of the first prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current first prediction model can reach a certain accuracy, so the current first prediction model can be determined It is the target prediction model. In a possible embodiment, the target prediction model may be updated in real time, that is, during the use of the target prediction model, each additional display information is displayed on the first booth, or each additional target display information Show once in the second booth, after the show, the actual click rate corresponding to the impression will be collected, and the display information and actual click rate corresponding to the actual impression will be used as new training data, and the target prediction model will be continuously carried out. Revise to make the target decision model more and more accurate.
通过上述方式，可以利用第一展位对应的第一历史展示信息和第一历史展示信息对应的实际点击率以及目标展示信息和目标展示信息在第二展位对应的实际点击率对第一预测模型进行训练，以得到目标预测模型。这样，可以结合第一展位的相关信息以及目标展示信息的相关信息获得目标预测模型，使得到的目标预测模型更加精确。In the above manner, the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information and the actual click rate of the target display information and target display information corresponding to the second booth can be used to perform the first prediction model Train to get the target prediction model. In this way, the target prediction model can be obtained in combination with the relevant information of the first booth and the target display information, so that the obtained target prediction model is more accurate.
在另一种可能的实施例中，目标模型可以通过以下方式构建：In another possible embodiment, the target model can be constructed in the following manner:
确定与目标展示信息对应的目标展示时段；Determine the target display period corresponding to the target display information;
获取第二训练样本；Obtain the second training sample;
根据第二训练样本对第二预测模型进行训练，以获得目标预测模型。Train the second prediction model according to the second training sample to obtain the target prediction model.
目标展示信息可以存在对应的目标展示时段，因此，对点击率进行预测，就是对目标展示信息在第一展位处于目标展示时段的点击率进行预测。这里的展示时段是指一天中的某个时段，目标展示信息可能需要在多个不同的展示时段进行展示，因此，可以确定目标展示信息当前展示所对应的目标展示时段。The target display information may exist in the corresponding target display period. Therefore, the prediction of the click rate is to predict the click rate of the target display information in the target display period in the first booth. The display period here refers to a certain period of the day. The target display information may need to be displayed in multiple different display periods. Therefore, the target display period corresponding to the current display of the target display information can be determined.
在一种可能的实施例中，第二训练样本可以包括第一展位在目标展示时段对应的第二历史展示信息以及与第二历史展示信息对应的实际点击率。其中，第二历史展示信息可以为第一展位在目标展示时段展示过的所 有展示信息，与第二历史展示信息对应的实际点击率则为第一展位在目标展示时段内展示过的所有展示信息各自对应的点击率。In a possible embodiment, the second training sample may include second historical display information corresponding to the first booth at the target display period and actual click rate corresponding to the second historical display information. Wherein, the second historical display information may be all display information displayed by the first booth during the target display period, and the actual click rate corresponding to the second historical display information is all display information displayed by the first booth during the target display period The corresponding click-through rate.
在获取到第二训练样本后，可以根据第二训练样本对第二预测模型进行训练，以获得目标预测模型。示例地，可以通过机器学习算法(例如，神经网络学习)对第二预测模型进行训练，以获得目标预测模型。After obtaining the second training sample, the second prediction model may be trained according to the second training sample to obtain the target prediction model. Illustratively, the second prediction model may be trained by machine learning algorithms (eg, neural network learning) to obtain the target prediction model.
示例地，可以利用神经网络学习方法获得目标预测模型。下面将针对有监督的神经网络训练方法构建目标预测模型的过程进行详细说明，但是本公开所提供的方法并不局限于此学习方法，并且，也不局限于这一种训练方式，下述实例仅作为示例性说明。Illustratively, a neural network learning method can be used to obtain a target prediction model. The process of constructing the target prediction model for the supervised neural network training method will be described in detail below, but the method provided by the present disclosure is not limited to this learning method, and is not limited to this training method, the following examples Just as an example.
将训练样本中的一个第二历史展示信息输入第二预测模型，根据第二预测模型的实际输出(第二预测模型针对输入的第二历史展示信息所生成的预测点击率)与期望输出(与输入的第二历史展示信息对应的实际点击率)间的差距来调整模型内神经网络的连接权。初始情况下，第二预测模型内神经网络的连接权可以是随机的。其中，将第二历史信息输入第二预测模型。举例来说，若第二历史信息为文字，则将该文字输入第二预测模型。示例地，第二预测模型的隐层中可以包含能够对输入的展示信息进行特征提取的层，以获得输入信息的特征，从而可以更加便利地训练预测模型。之后，对训练样本中的每一个第二历史展示信息执行上述操作，直至实际输出与期望输出之间的差距小于预设的差距阈值。当第二预测模型的实际输出与期望输出之间的差距小于预设的差距阈值时，说明当前的第二预测模型的预测可以达到一定的准确率，因此，可以将当前的第二预测模型确定为目标预测模型。在一种可能的实施例中，目标预测模型可以是实时更新的，也就是说，在目标预测模型使用的过程中，第一展位在目标展示时段每多展示一种展示信息，在该展示信息展示后则会收集该次展示所对应的实际点击率，并将该次实际展示对应的展示信息(作为新增的第二历史展示信息)和实际点击率作为新的训练数据，不断对目标预测模型进行修正，以使目标决策模型越来越精确。Input the second historical display information in the training sample into the second prediction model, according to the actual output of the second prediction model (the predicted click rate generated by the second prediction model for the input second historical display information) and the expected output (and The actual click rate corresponding to the input second historical display information is used to adjust the connection weight of the neural network in the model. Initially, the connection weight of the neural network in the second prediction model may be random. Among them, the second historical information is input into the second prediction model. For example, if the second historical information is text, the text is input into the second prediction model. Exemplarily, the hidden layer of the second prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently. After that, the above operation is performed on each second historical display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold. When the gap between the actual output of the second prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current second prediction model can reach a certain accuracy, so the current second prediction model can be determined It is the target prediction model. In a possible embodiment, the target prediction model may be updated in real time, that is to say, during the use of the target prediction model, the first booth exhibits one more type of display information during the target display period, and the display information After the impression, the actual click rate corresponding to the impression will be collected, and the impression information corresponding to the actual impression (as the newly added second historical impression information) and the actual click rate will be used as new training data to continuously predict the target The model is revised to make the target decision model more and more accurate.
通过上述方式，可以利用第一展位在目标展示时段对应的第二历史展示信息和第二历史展示信息对应的实际点击率对第二预测模型进行训练， 以得到目标预测模型。由于第一展位在目标展示时段对应的第二历史展示信息和第二历史展示信息对应的实际点击率易于获得，便于进行数据采集，以获得目标预测模型，同时将模型精确到时段，使得模型针对性更强。In the above manner, the second prediction model can be trained using the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information to obtain the target prediction model. Since the second historical display information corresponding to the target exhibition period and the actual click rate corresponding to the second historical display information of the first booth are easy to obtain, it is convenient for data collection to obtain the target prediction model, and at the same time, the model is accurate to the time period, so that the model is targeted Sex is stronger.
在另一种可能的实施例中，第二训练样本除了包括第一展位在目标展示时段对应的第二历史展示信息以及与第二历史展示信息对应的实际点击率以外，还可以包括目标展示信息以及目标展示信息在第三展位对应的实际点击率。其中，第三展位为与第一展位不同的展位。示例地，可以存在一个第三展位，或者，可以存在多个第三展位。示例地，若对于某一网站，目标展示信息仅在站内的第三展位上展示过，而并未在站外的第一展位展示过，则可以收集目标展示信息以及目标展示信息在第三展位对应的实际点击率。这里目标展示信息在第三展位对应的实际点击率可以是目标展示时段内目标展示信息在第三展位对应的实际点击率，或者，也可以不限于目标展示时段。In another possible embodiment, the second training sample may include target display information in addition to the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information And the actual click rate of the target display information in the third booth. Among them, the third booth is different from the first booth. For example, there may be one third booth, or there may be multiple third booths. For example, if for a website, the target display information is only displayed in the third booth in the station, but not in the first booth outside the station, you can collect the target display information and the target display information in the third booth Corresponding actual click-through rate. Here, the actual click rate of the target display information corresponding to the third booth may be the actual click rate of the target display information corresponding to the third booth during the target display period, or it may not be limited to the target display period.
示例地，在构建目标预测模型时，可以将训练样本中的第二历史展示信息和/或目标展示信息输入第二预测模型，根据第二预测模型的实际输出与期望输出之间的差距来调整模型内神经网络的连接权。例如，可以逐个将训练样本中的展示信息输入第二预测模型，也就是每一次将一个第二历史展示信息或目标展示信息输入第二预测模型，根据第二预测模型的实际输出(第二预测模型针对输入的第二历史展示信息或目标展示信息所生成的预测点击率)与期望输出(与输入的第二历史展示信息或目标展示信息对应的实际点击率)间的差距来调整模型内神经网络的连接权。初始情况下，第二预测模型内神经网络的连接权可以是随机的。其中，将第二历史信息输入第二预测模型。举例来说，若第二历史信息为视频，则将该视频输入第二预测模型。示例地，第二预测模型的隐层中可以包含能够对输入的展示信息进行特征提取的层，以获得输入信息的特征，从而可以更加便利地训练预测模型。之后，对训练样本中的每一个第二历史展示信息或目标展示信息执行上述操作，直至实际输出与期望输出之间的差距小于预设的差距阈值。当第二预测模型的实际输出与期望输出之间的差距小于预设的差距阈值时，说明当前的第二预测模型的预测可以达到一定的准确 率，因此，可以将当前的第二预测模型确定为目标预测模型。在一种可能的实施例中，目标预测模型可以是实时更新的，也就是说，在目标预测模型使用的过程中，第一展位在目标展示时段内每多展示一种展示信息，或者，目标展示信息每多在第三展位上展示一次，在相应的展示信息展示后则会收集该次展示所对应的实际点击率，并将该次实际展示对应的展示信息和实际点击率作为新的训练数据，不断对目标预测模型进行修正，以使目标决策模型越来越精确。For example, when constructing the target prediction model, the second historical display information and/or target display information in the training sample may be input into the second prediction model, and adjusted according to the gap between the actual output and the expected output of the second prediction model The connection weight of the neural network in the model. For example, the display information in the training sample can be input into the second prediction model one by one, that is, each time a second historical display information or target display information is input into the second prediction model, according to the actual output of the second prediction model (second prediction The model adjusts the intra-model nerve for the gap between the predicted click rate generated by the input second historical display information or target display information and the expected output (the actual click rate corresponding to the input second historical display information or target display information) The right to connect to the network. Initially, the connection weight of the neural network in the second prediction model may be random. Among them, the second historical information is input into the second prediction model. For example, if the second historical information is a video, the video is input to the second prediction model. Exemplarily, the hidden layer of the second prediction model may include a layer capable of extracting the input display information to obtain the characteristics of the input information, so that the prediction model can be trained more conveniently. After that, the above operation is performed on each second historical display information or target display information in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold. When the gap between the actual output of the second prediction model and the expected output is less than the preset gap threshold, it means that the prediction of the current second prediction model can reach a certain accuracy, so the current second prediction model can be determined It is the target prediction model. In a possible embodiment, the target prediction model may be updated in real time, that is to say, during the use of the target prediction model, the first booth exhibits one more type of display information during the target display period, or, the target Every time the impression information is displayed in the third booth, after the corresponding impression information is displayed, the actual click rate corresponding to the impression will be collected, and the impression information and actual click rate corresponding to the actual impression will be used as new training Data, and constantly modify the target prediction model to make the target decision model more and more accurate.
通过上述方式，可以利用第一展位在目标展示时段对应的第二历史展示信息和第二历史展示信息对应的实际点击率以及目标展示信息和目标展示信息在第三展位对应的实际点击率对第一预测模型进行训练，以得到目标预测模型。这样，可以结合第一展位的相关信息以及目标展示信息的相关信息获得目标预测模型，且预测模型与展示时段相关，使得到的目标预测模型更加精确而具有针对性。In the above manner, the second historical display information corresponding to the first booth and the actual click rate corresponding to the second historical display information and the actual click rate corresponding to the target display information and target display information at the third booth can be used A prediction model is trained to obtain the target prediction model. In this way, the target prediction model can be obtained by combining the relevant information of the first booth and the target display information, and the prediction model is related to the display period, so that the obtained target prediction model is more accurate and targeted.
需要说明的是，利用机器学习算法构建模型的方法与流程均为本领域技术人员公知，为理解方便在上文中对于其中的几种可能的情况进行了简要说明，但是本公开中构建模型的方式并不局限于此，对于其他的实现方式此处不赘述。It should be noted that the methods and processes for constructing models using machine learning algorithms are well known to those skilled in the art. For the sake of understanding, several possible situations are briefly described above, but the method of constructing models in the present disclosure It is not limited to this, and the other implementation manners will not be repeated here.
将目标展示信息输入至第一展位对应的目标预测模型，即可获得目标展示信息对应的预测点击率。Enter the target display information into the target prediction model corresponding to the first booth to obtain the predicted click rate corresponding to the target display information.
这样，可以在将展示信息展示至第一展位之前，通过预测模型预测该展示信息在第一展位可能得到的点击率，从而可以获知该第一展位是否适合该展示信息。In this way, before displaying the display information to the first booth, the prediction model may be used to predict the possible click rate of the display information in the first booth, so as to know whether the first booth is suitable for the display information.
图2是根据本公开的示例性实施例的展位选择方法的流程图。如图2所示，该方法可以包括以下步骤21至23。2 is a flowchart of a booth selection method according to an exemplary embodiment of the present disclosure. As shown in FIG. 2, the method may include the following steps 21 to 23.
在步骤21中，确定与目标展示信息对应的候选展位。In step 21, the candidate booth corresponding to the target display information is determined.
对于目标展示信息，可能存在多个可以用于展示该目标展示信息的候选展位。示例地，每个展示信息可以对应于相应的内容模板，该内容模板可以用于限制展示信息的内容类型，例如科普类内容、搞笑类内容，而展位可以展示特定的内容模板的展示信息，因此，可以根据内容模板确定与 目标展示信息对应的候选展位。再例如，每个展示信息可以对应于展示信息类型，例如图片类、视频类、文字类或其组合等，一个展位可以展示特定的类型的展示信息，因此，可以根据展示信息类型确定与目标展示信息对应的候选展位。例如，目标展示信息为视频类展示信息，那么与目标展位信息对应的候选展位可以为可展示视频类展示信息的展位。For the target display information, there may be multiple candidate booths that can be used to display the target display information. For example, each display information may correspond to a corresponding content template. The content template may be used to limit the content types of the display information, such as popular science content and funny content, and the booth can display the display information of a specific content template, so , The candidate booth corresponding to the target display information can be determined according to the content template. As another example, each display information may correspond to the type of display information, such as image, video, text, or a combination thereof. A booth can display a specific type of display information. Therefore, it can be determined according to the type of display information and the target display Candidate booth corresponding to the information. For example, if the target display information is video-type display information, the candidate booth corresponding to the target booth information may be a booth that can display video-type display information.
在步骤22中，确定每个候选展位对应的预测点击率。In step 22, the predicted click rate corresponding to each candidate booth is determined.
根据本公开任意实施例提供的点击率预测方法，可以针对每一个候选展位进行点击率预测。其中，点击率预测的具体实施例在上文中已有说明，此处不赘述。According to the click rate prediction method provided by any embodiment of the present disclosure, click rate prediction can be performed for each candidate booth. Among them, the specific embodiment of the click rate prediction has been described above, and will not be repeated here.
在步骤23中，将预测点击率最高的候选展位确定为目标展位。In step 23, the candidate booth with the highest predicted click rate is determined as the target booth.
目标展位即为候选展位中最适合展示目标展示信息的展位。The target booth is the candidate booth that is most suitable for displaying the target display information.
通过上述方式，可以得到目标展示信息展示至各个候选展位的预测点击率，从而可以根据该预测点击率选择合适的展位，以使目标展示信息展示至目标展位能够获得最高的点击率。In the above manner, the predicted click rate of the target display information displayed to each candidate booth can be obtained, so that an appropriate booth can be selected according to the predicted click rate, so that the target display information displayed to the target booth can obtain the highest click rate.
图3是根据本公开的示例性实施例的点击率预测装置的框图。如图3所示，该装置30可以包括：FIG. 3 is a block diagram of a click rate prediction apparatus according to an exemplary embodiment of the present disclosure. As shown in FIG. 3, the device 30 may include:
第一确定模块31，用于确定与目标展示信息对应的第一展位；The first determining module 31 is used to determine the first booth corresponding to the target display information;
预测模块32，用于将目标展示信息输入至与第一展位对应的目标预测模型，以获得目标展示信息对应的预测点击率。The prediction module 32 is configured to input the target display information into the target prediction model corresponding to the first booth to obtain the predicted click rate corresponding to the target display information.
可选地，目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
获取第一训练样本，第一训练样本至少包括第一展位对应的第一历史展示信息以及与第一历史展示信息对应的实际点击率；Obtaining a first training sample, the first training sample at least includes the first historical display information corresponding to the first booth and the actual click rate corresponding to the first historical display information;
根据第一训练样本对第一预测模型进行训练，以获得目标预测模型。Train the first prediction model according to the first training sample to obtain the target prediction model.
可选地，第一训练样本还包括目标展示信息以及目标展示信息在第二展位对应的实际点击率，第二展位为与第一展位不同的展位。Optionally, the first training sample further includes target display information and the actual click rate of the target display information corresponding to the second booth, the second booth is a different booth from the first booth.
可选地，目标预测模型通过以下方式构建：Optionally, the target prediction model is constructed in the following manner:
确定与目标展示信息对应的目标展示时段；Determine the target display period corresponding to the target display information;
获取第二训练样本，第二训练样本至少包括第一展位在目标展示时段对应的第二历史展示信息以及与第二历史展示信息对应的实际点击率；Obtaining a second training sample, the second training sample includes at least the second historical display information corresponding to the first booth at the target display period and the actual click rate corresponding to the second historical display information;
根据第二训练样本对第二预测模型进行训练，以获得目标预测模型。Train the second prediction model according to the second training sample to obtain the target prediction model.
可选地，第二训练样本还包括目标展示信息以及目标展示信息在第三展位对应的实际点击率，第三展位为与第一展位不同的展位。Optionally, the second training sample further includes target display information and actual click rate of the target display information corresponding to the third booth, the third booth is a different booth from the first booth.
图4是根据本公开的示例性实施例的展位选择装置的框图。如图4所示，该装置40可以包括：4 is a block diagram of a booth selection apparatus according to an exemplary embodiment of the present disclosure. As shown in FIG. 4, the device 40 may include:
第二确定模块41，用于确定与目标展示信息对应的候选展位；The second determination module 41 is used to determine the candidate booth corresponding to the target display information;
根据本公开的上述实施例的点击率预测装置42，用于确定每个候选展位对应的预测点击率；The click rate prediction device 42 according to the above embodiment of the present disclosure is used to determine the predicted click rate corresponding to each candidate booth;
第三确定模块43，用于将预测点击率最高的候选展位确定为待选择的目标展位。The third determining module 43 is configured to determine the candidate booth with the highest predicted click rate as the target booth to be selected.
关于上述实施例中的装置，其中各个模块执行操作的具体方式已经在有关该方法的实施例中进行了详细描述，此处将不做详细阐述说明。Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment related to the method, and will not be elaborated here.
图5是根据本公开的示例性实施例的一种电子设备的框图。例如，电子设备500可以被提供为服务器。参照图5，电子设备500可以包括：一个或多个处理器522；以及存储器532，用于存储可由处理器522执行的计算机程序。存储器532中存储的计算机程序可以包括每一个对应于一组指令的模块。此外，处理器522可以被配置为执行该计算机程序，以执行上述的点击率预测方法。FIG. 5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. For example, the electronic device 500 may be provided as a server. 5, the electronic device 500 may include: one or more processors 522; and a memory 532 for storing computer programs executable by the processor 522. The computer program stored in the memory 532 may include each module corresponding to a set of instructions. In addition, the processor 522 may be configured to execute the computer program to perform the above-mentioned click rate prediction method.
另外，电子设备500还可以包括电源组件526和通信组件550，该电源组件526可以被配置为执行电子设备500的电源管理，该通信组件550可以被配置为实现电子设备500的通信，例如，有线或无线通信。此外，该电子设备500还可以包括输入/输出(I/O)接口558。电子设备500可以操作基于存储在存储器532的操作***，例如Windows Server TM、Mac OS X TM、Unix TM、Linux TM等等。 In addition, the electronic device 500 may further include a power component 526 and a communication component 550, which may be configured to perform power management of the electronic device 500, and the communication component 550 may be configured to implement communication of the electronic device 500, for example, wired Or wireless communication. In addition, the electronic device 500 may also include an input/output (I/O) interface 558. The electronic device 500 can operate based on an operating system stored in the memory 532, such as Windows Server ™ , Mac OS X ™ , Unix ™ , Linux ™, and so on.
在另一示例性实施例中，还提供了一种包括程序指令的计算机可读存储介质，该程序指令被处理器执行时实现上述的点击率预测方法。例如，该计算机可读存储介质可以为上述包括程序指令的存储器532，上述程序指令可由电子设备500的处理器522执行以实现上述的点击率预测方法。In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When the program instructions are executed by a processor, the above-mentioned click-through rate prediction method is implemented. For example, the computer-readable storage medium may be the above-mentioned memory 532 including program instructions, and the above-mentioned program instructions may be executed by the processor 522 of the electronic device 500 to implement the above-mentioned click rate prediction method.
图6是根据本公开的示例性实施例的一种电子设备的框图。例如，电 子设备1900可以被提供为服务器。参照图6，电子设备1900可以包括：一个或多个处理器1922；以及存储器1932，用于存储可由处理器1922执行的计算机程序。存储器1932中存储的计算机程序可以包括每一个对应于一组指令的模块。此外，处理器1922可以被配置为执行该计算机程序，以实现上述的展位选择方法。6 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. For example, the electronic device 1900 may be provided as a server. 6, the electronic device 1900 may include: one or more processors 1922; and a memory 1932 for storing computer programs executable by the processor 1922. The computer program stored in the memory 1932 may include each module corresponding to a set of instructions. In addition, the processor 1922 may be configured to execute the computer program to implement the booth selection method described above.
另外，电子设备1900还可以包括电源组件1926和通信组件1950，该电源组件1926可以被配置为执行电子设备1900的电源管理，该通信组件1950可以被配置为实现电子设备1900的通信，例如，有线或无线通信。此外，该电子设备1900还可以包括输入/输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作***，例如Windows Server TM、Mac OS X TM、Unix TM、Linux TM等等。 In addition, the electronic device 1900 may further include a power supply component 1926 and a communication component 1950, which may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired Or wireless communication. In addition, the electronic device 1900 may also include an input/output (I/O) interface 1958. The electronic device 1900 can operate an operating system based on the memory 1932, such as Windows Server ™ , Mac OS X ™ , Unix ™ , Linux ™, and so on.
在另一示例性实施例中，还提供了一种包括程序指令的计算机可读存储介质，该程序指令被处理器执行时实现上述的展位选择方法。例如，该计算机可读存储介质可以为上述包括程序指令的存储器1932，上述程序指令可由电子设备1900的处理器1922执行以实现上述的展位选择方法。In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When the program instructions are executed by a processor, the above booth selection method is implemented. For example, the computer-readable storage medium may be the aforementioned memory 1932 including program instructions, which may be executed by the processor 1922 of the electronic device 1900 to implement the aforementioned booth selection method.
以上结合附图详细描述了本公开的示例性实施例，但是，本公开并不限于上述实施例中的具体细节，在本公开的技术构思范围内，可以对本公开的实施例进行多种简单变型，这些简单变型均属于本公开的保护范围。The exemplary embodiments of the present disclosure have been described in detail above with reference to the drawings. However, the present disclosure is not limited to the specific details in the above embodiments, and within the scope of the technical idea of the present disclosure, various simple modifications can be made to the embodiments of the present disclosure These simple modifications all fall within the protection scope of the present disclosure.
另外需要说明的是，在上述具体实施例中所描述的各个具体技术特征，在不矛盾的情况下，可以通过任何合适的方式进行组合。为了避免不必要的重复，本公开对各种可能的组合方式不再另行说明。In addition, it should be noted that the specific technical features described in the foregoing specific embodiments can be combined in any suitable manner without contradictions. In order to avoid unnecessary repetition, the present disclosure does not describe various possible combinations.
此外，本公开的各种不同的实施例之间也可以进行任意组合，只要其不违背本公开的构思，就应当被视为在本公开的范围内。In addition, any combination of various embodiments of the present disclosure can also be arbitrarily combined, as long as it does not violate the concept of the present disclosure, it should be regarded as within the scope of the present disclosure.

Claims (14)

  1. 一种点击率预测方法，包括：A click-through rate prediction method, including:
    确定与目标展示信息对应的第一展位；Determine the first booth corresponding to the target display information;
    将所述目标展示信息输入至与所述第一展位对应的目标预测模型，以获得所述目标展示信息对应的预测点击率。The target display information is input to a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
  2. 根据权利要求1所述的方法，其中，所述目标预测模型通过以下方式构建：The method according to claim 1, wherein the target prediction model is constructed in the following manner:
    获取第一训练样本，所述第一训练样本至少包括所述第一展位对应的第一历史展示信息以及与所述第一历史展示信息对应的实际点击率；Obtaining a first training sample, the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
    根据所述第一训练样本对第一预测模型进行训练，以获得所述目标预测模型。Training the first prediction model according to the first training sample to obtain the target prediction model.
  3. 根据权利要求2所述的方法，其中，所述第一训练样本还包括所述目标展示信息以及所述目标展示信息在第二展位对应的实际点击率，所述第二展位为与所述第一展位不同的展位。The method according to claim 2, wherein the first training sample further includes the target display information and the actual click rate of the target display information corresponding to the second booth, the second booth is One booth with different booths.
  4. 根据权利要求1所述的方法，其中，所述目标预测模型通过以下方式构建：The method according to claim 1, wherein the target prediction model is constructed in the following manner:
    确定与所述目标展示信息对应的目标展示时段；Determine a target display period corresponding to the target display information;
    获取第二训练样本，所述第二训练样本至少包括所述第一展位在所述目标展示时段对应的第二历史展示信息以及与所述第二历史展示信息对应的实际点击率；Acquiring a second training sample, the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
    根据所述第二训练样本对第二预测模型进行训练，以获得所述目标预测模型。Training the second prediction model according to the second training sample to obtain the target prediction model.
  5. 根据权利要求4所述的方法，其中，所述第二训练样本还包括所述目标展示信息以及所述目标展示信息在第三展位对应的实际点击率，所述第三 展位为与所述第一展位不同的展位。The method according to claim 4, wherein the second training sample further includes the target display information and the actual click rate of the target display information corresponding to the third booth, the third booth is One booth with different booths.
  6. 一种展位选择方法，包括：A booth selection method, including:
    确定与目标展示信息对应的候选展位；Identify the candidate booth corresponding to the target display information;
    根据权利要求1-5中任一项所述的点击率预测方法，确定每个所述候选展位对应的预测点击率；According to the click rate prediction method according to any one of claims 1 to 5, determining the predicted click rate corresponding to each of the candidate booths;
    将预测点击率最高的候选展位确定为待选择的目标展位。The candidate booth with the highest predicted click rate is determined as the target booth to be selected.
  7. 一种点击率预测装置，包括：A click rate prediction device, including:
    第一确定模块，用于确定与目标展示信息对应的第一展位；The first determining module is used to determine the first booth corresponding to the target display information;
    预测模块，用于将所述目标展示信息输入至与所述第一展位对应的目标预测模型，以获得所述目标展示信息对应的预测点击率。The prediction module is configured to input the target display information into a target prediction model corresponding to the first booth to obtain a predicted click rate corresponding to the target display information.
  8. 根据权利要求7所述的装置，其中，所述目标预测模型通过以下方式构建：The apparatus according to claim 7, wherein the target prediction model is constructed in the following manner:
    获取第一训练样本，所述第一训练样本至少包括所述第一展位对应的第一历史展示信息以及与所述第一历史展示信息对应的实际点击率；Obtaining a first training sample, the first training sample includes at least first historical display information corresponding to the first booth and actual click rate corresponding to the first historical display information;
    根据所述第一训练样本对第一预测模型进行训练，以获得所述目标预测模型。Training the first prediction model according to the first training sample to obtain the target prediction model.
  9. 根据权利要求8所述的装置，其中，所述第一训练样本还包括所述目标展示信息以及所述目标展示信息在第二展位对应的实际点击率，所述第二展位为与所述第一展位不同的展位。The apparatus according to claim 8, wherein the first training sample further includes the target display information and an actual click rate of the target display information corresponding to the second booth, the second booth is One booth with different booths.
  10. 根据权利要求7所述的装置，其中，所述目标预测模型通过以下方式构建：The apparatus according to claim 7, wherein the target prediction model is constructed in the following manner:
    确定与所述目标展示信息对应的目标展示时段；Determine a target display period corresponding to the target display information;
    获取第二训练样本，所述第二训练样本至少包括所述第一展位在所述目标展示时段对应的第二历史展示信息以及与所述第二历史展示信息对应的实 际点击率；Obtaining a second training sample, the second training sample at least includes second historical display information corresponding to the first booth during the target display period and an actual click rate corresponding to the second historical display information;
    根据所述第二训练样本对第二预测模型进行训练，以获得所述目标预测模型。Training the second prediction model according to the second training sample to obtain the target prediction model.
  11. 根据权利要求10所述的装置，其中，所述第二训练样本还包括所述目标展示信息以及所述目标展示信息在第三展位对应的实际点击率，所述第三展位为与所述第一展位不同的展位。The apparatus according to claim 10, wherein the second training sample further includes the target display information and an actual click rate of the target display information corresponding to the third booth, the third booth is One booth with different booths.
  12. 一种展位选择装置，包括：A booth selection device, including:
    第二确定模块，用于确定与目标展示信息对应的候选展位；The second determination module is used to determine the candidate booth corresponding to the target display information;
    权利要求7-11中任一项所述的点击率预测装置，用于确定每个所述候选展位对应的预测点击率；The click-through rate prediction device according to any one of claims 7-11, used to determine the predicted click-through rate corresponding to each of the candidate booths;
    第三确定模块，用于将预测点击率最高的候选展位确定为待选择的目标展位。The third determining module is used to determine the candidate booth with the highest predicted click rate as the target booth to be selected.
  13. 一种计算机可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现权利要求1-5中任一项所述的方法，或者，该程序被处理器执行时实现权利要求6所述的方法。A computer-readable storage medium on which a computer program is stored, which implements the method of any one of claims 1-5 when executed by a processor, or which implements claim 6 when the program is executed by a processor The method.
  14. 一种电子设备，包括：An electronic device, including:
    存储器，其上存储有计算机程序；Memory, on which computer programs are stored;
    处理器，用于执行所述存储器中的所述计算机程序，以实现权利要求1-5中任一项所述的方法，或者实现权利要求6所述的方法。A processor, configured to execute the computer program in the memory, to implement the method according to any one of claims 1-5, or implement the method according to claim 6.
PCT/CN2019/094738 2018-12-29 2019-07-04 Click rate prediction method and apparatus, and display position selection method and apparatus WO2020134009A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811641964.9 2018-12-29
CN201811641964.9A CN111401592A (en) 2018-12-29 2018-12-29 Click rate prediction method, booth selection method, apparatus, medium, and electronic device

Publications (1)

Publication Number Publication Date
WO2020134009A1 true WO2020134009A1 (en) 2020-07-02

Family

ID=71125941

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/094738 WO2020134009A1 (en) 2018-12-29 2019-07-04 Click rate prediction method and apparatus, and display position selection method and apparatus

Country Status (2)

Country Link
CN (1) CN111401592A (en)
WO (1) WO2020134009A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103268344A (en) * 2013-05-24 2013-08-28 华东师范大学 Method for matching advertisements and pages without position bias influence
US20150278376A1 (en) * 2014-04-01 2015-10-01 Baidu (China) Co., Ltd. Method and apparatus for presenting search result
CN105183772A (en) * 2015-08-07 2015-12-23 百度在线网络技术（北京）有限公司 Release information click rate estimation method and apparatus
CN106445997A (en) * 2016-07-20 2017-02-22 腾讯科技（北京）有限公司 Information processing method and server
CN106453447A (en) * 2015-08-06 2017-02-22 阿里巴巴集团控股有限公司 Information display method and apparatus
CN107767174A (en) * 2017-10-19 2018-03-06 厦门美柚信息科技有限公司 The Forecasting Methodology and device of a kind of ad click rate

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107105031A (en) * 2017-04-20 2017-08-29 北京京东尚科信息技术有限公司 Information-pushing method and device

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103268344A (en) * 2013-05-24 2013-08-28 华东师范大学 Method for matching advertisements and pages without position bias influence
US20150278376A1 (en) * 2014-04-01 2015-10-01 Baidu (China) Co., Ltd. Method and apparatus for presenting search result
CN106453447A (en) * 2015-08-06 2017-02-22 阿里巴巴集团控股有限公司 Information display method and apparatus
CN105183772A (en) * 2015-08-07 2015-12-23 百度在线网络技术（北京）有限公司 Release information click rate estimation method and apparatus
CN106445997A (en) * 2016-07-20 2017-02-22 腾讯科技（北京）有限公司 Information processing method and server
CN107767174A (en) * 2017-10-19 2018-03-06 厦门美柚信息科技有限公司 The Forecasting Methodology and device of a kind of ad click rate

Also Published As

Publication number Publication date
CN111401592A (en) 2020-07-10

Similar Documents

Publication Publication Date Title
US20210042666A1 (en) Localized Learning From A Global Model
US11188950B2 (en) Audience expansion for online social network content
WO2021151336A1 (en) Road image target detection method based on attentional mechanism and related device
WO2017080170A1 (en) Group user profiling method and system
WO2015183370A1 (en) Optimizing non-convex function for learning to rank
US11593860B2 (en) Method, medium, and system for utilizing item-level importance sampling models for digital content selection policies
US11483408B2 (en) Feature-based network embedding
WO2021174877A1 (en) Processing method for smart decision-based target detection model, and related device
US20220156956A1 (en) Active image depth prediction
EP2972950B1 (en) Segmentation of content delivery
US20180203857A1 (en) Automated data exploration and validation
US20220334837A1 (en) Method, apparatus, and system for outputting a development unit performance insight interface component comprising a visual emphasis element in response to an insight interface component request
US10936601B2 (en) Combined predictions methodology
US9478042B1 (en) Determining visibility of rendered content
CN110084317B (en) Method and device for recognizing images
WO2019117970A1 (en) Adaptive object tracking policy
CN112381236A (en) Data processing method, device, equipment and storage medium for federal transfer learning
US11899823B2 (en) Privacy safe anonymized identity matching
WO2020134009A1 (en) Click rate prediction method and apparatus, and display position selection method and apparatus
WO2023115974A1 (en) Multimedia resource recommendation method and apparatus and object representation network generation method and apparatus
CN109034085B (en) Method and apparatus for generating information
US11200284B1 (en) Optimization of feature embeddings for deep learning models
WO2015158224A1 (en) Message recommendation method and apparatus
US20160042277A1 (en) Social action and social tie prediction
US20210166263A1 (en) Identification of software robot activity

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19903587

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 05.10.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19903587

Country of ref document: EP

Kind code of ref document: A1