WO2020134012A1 - 价值组合确定方法、装置、存储介质及电子设备 - Google Patents

价值组合确定方法、装置、存储介质及电子设备 Download PDF

Info

Publication number
WO2020134012A1
WO2020134012A1 PCT/CN2019/094747 CN2019094747W WO2020134012A1 WO 2020134012 A1 WO2020134012 A1 WO 2020134012A1 CN 2019094747 W CN2019094747 W CN 2019094747W WO 2020134012 A1 WO2020134012 A1 WO 2020134012A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
target
display information
combination
expected
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/094747
Other languages
English (en)
French (fr)
Inventor
李亦锬
余林韵
陈嘉闽
黄训蓬
李磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing ByteDance Network Technology Co Ltd
Original Assignee
Beijing ByteDance Network Technology Co Ltd
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 Beijing ByteDance Network Technology Co Ltd filed Critical Beijing ByteDance Network Technology Co Ltd
Publication of WO2020134012A1 publication Critical patent/WO2020134012A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/40Business processes related to social networking or social networking services
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06QINFORMATION 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
    • G06Q50/00Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism

Definitions

  • the present disclosure relates to the field of computer technology, and in particular, to a value combination determination method, device, storage medium, and electronic equipment.
  • 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 has a predetermined value, and the corresponding value can be obtained by displaying such display information. Therefore, for this type of display information, it is usually displayed according to the order of its planned value, that is, the higher the planned value, the higher the display priority. In the actual situation, different display information providers will give the planned value and expected value corresponding to the display information.
  • the expected value of the display information can be considered as the total cost of the display information in a display cycle under the planned value value.
  • the return value After the display information is actually displayed, the return value will be obtained, that is, the return obtained during the display period, and the ultimate purpose of the display information display is to obtain a high return value.
  • the planned value and the expected value need to be determined after manual data collection and human analysis. In this way, on the one hand, a large amount of manpower, material resources, and time are required, and the efficiency is very low; on the other hand, because the method of manual analysis is limited by factors such as experience and environment, the determined proposed value and expected value may not meet the requirements in real time And accuracy may be insufficient.
  • a method for determining a value combination including:
  • the value combination corresponding to the display information includes the proposed value of the display information and the expected value of the display information
  • the proposed value in the first value combination with the largest predicted return value is determined as the target intended value of the target display information, and the expected value in the first value combination with the largest predicted return value is determined as the target of the target display information Expected value.
  • the target prediction model is obtained as follows:
  • Obtaining a training sample corresponding to the target display information the training sample at least including a second value combination corresponding to the target display information and an actual return value corresponding to the second value combination, wherein the second The value combination is the value combination corresponding to the target display information in the historical actual display;
  • the training sample further includes a third value combination corresponding to historical display information and an actual return value corresponding to the third value combination
  • the historical display information is a display that has a corresponding relationship with the target display information information.
  • the multiple first value combinations corresponding to the generated target display information include:
  • the plurality of first value combinations are generated.
  • the first value combination with the lowest expected value is determined as the target first value combination
  • a value combination determination device comprising:
  • a generating module configured to generate a plurality of first value combinations corresponding to the target display information, wherein the value combination corresponding to the display information includes the planned value of the display information and the expected value of the display information;
  • a processing module configured to input the plurality of first value combinations into a target prediction model to obtain a corresponding predicted return value
  • a determining module configured to determine the planned value in the first value combination with the largest predicted return value as the target planned value of the target display information, and determine the expected value in the first value combination with the largest predicted return value as the The goal shows the expected value of the information.
  • the target prediction model is obtained as follows:
  • Obtaining a training sample corresponding to the target display information the training sample at least including a second value combination corresponding to the target display information and an actual return value corresponding to the second value combination, wherein the second The value combination is the value combination corresponding to the target display information in the historical actual display;
  • the training sample further includes a third value combination corresponding to historical display information and an actual return value corresponding to the third value combination
  • the historical display information is a display that has a corresponding relationship with the target display information information.
  • the generating module includes:
  • a generation submodule configured to generate the plurality of first value combinations according to the expected value range and the planned value range.
  • the determination module includes:
  • a first determining submodule configured to determine the first value combination with the lowest expected value as the target first value combination when there are multiple first value combinations with the largest predicted return value
  • a second determining submodule configured to determine the proposed value in the target first value combination as the target intended value of the target display information, and determine the expected value in the target first value combination as the target Demonstrate the stated expected value of the information.
  • 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.
  • an electronic device including:
  • a processor for executing the computer program in the memory to implement the method according to the first aspect of the present disclosure.
  • the target proposed value and target expected value be automatically generated for the target display information, but there is no need to manually determine the planned value and expected value, thereby saving manpower and improving efficiency, but also making the generated target proposed value and target Expect value to get the best return value possible, so as to maximize user benefits.
  • FIG. 1 is a flowchart of a value combination determination method according to an exemplary embodiment of the present disclosure
  • FIG. 2 is a flowchart of an exemplary implementation of the steps of generating a plurality of first value combinations corresponding to target display information in a value combination determination method according to an exemplary embodiment of the present disclosure
  • FIG. 3 is a block diagram of a value combination determination apparatus according to an exemplary embodiment of the present disclosure.
  • FIG. 4 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure.
  • FIG. 1 is a flowchart of a value combination determination method according to an exemplary embodiment of the present disclosure. As shown in FIG. 1, the method may include the following steps 11 to 13.
  • step 11 multiple first value combinations corresponding to the target display information are generated.
  • the presentation information may be, for example, text, pictures, videos, or a combination thereof.
  • the value combination corresponding to the display information may include the planned value of the display information and the expected value of the display information.
  • the proposed value of the display information can be regarded as the value of the display information that is required to be clicked or viewed once after being displayed on the booth.
  • the booth can be understood as a virtual or actual carrier for displaying the display information, and the expectations of the display information
  • the value can be regarded as the total value of the impression information that is required to be clicked or viewed a desired number of times (ie, one impression cycle) at the proposed value.
  • users can give the planned value and budget value.
  • the user corresponding to the currently highest proposed value will currently have the highest priority displayed in the booth.
  • each first value combination includes a planned value and an expected value.
  • the first value combination can be considered when generating Factors, that is to say, when generating multiple first value combinations corresponding to the target display information, for example, the condition that the expected value in the same first value combination is greater than or equal to the planned value can be set so that the generated first value combinations It is usable and improves the generation efficiency of the first value combination.
  • multiple first value combinations may be directly generated, for example, randomly.
  • the proposed value and the expected value may be randomly selected from positive real numbers, for example.
  • step 11 may include the following steps 21-22.
  • step 21 the expected value range and the planned value range corresponding to the target display information are obtained.
  • the value of various display information may be estimated in advance, that is, for different display information, respective expected value ranges and planned value ranges may be set for them in advance.
  • the expected value range and the planned value range can be set according to the content of the displayed information, the historical planned value and the historical expected value of the displayed information. Therefore, for the target display information, the expected value range and the planned value range corresponding to the target display information can be obtained.
  • step 22 multiple first value combinations are generated based on the expected value range and the planned value range.
  • multiple first value combinations can be generated.
  • the optional expected value and the optional planned value within the expected value range and the planned value range may be traversed to obtain multiple first value combinations.
  • the multiple first value combinations generated may be mutually different.
  • step 12 the multiple first value combinations are input into the target prediction model to obtain the corresponding predicted return value.
  • the target prediction model may generate and output a predicted return value corresponding to the first value combination according to the input first value combination.
  • the following is an example of how to obtain the target prediction model.
  • the target prediction model can be obtained as follows:
  • the training sample may include the second value combination corresponding to the target display information and the actual return value corresponding to the second value combination.
  • the second value combination may be the value combination corresponding to the target display information in the historical actual display.
  • the training sample corresponding to the target display information may include a second value combination corresponding to the target display information and an actual return value corresponding to the second value combination, where there may be multiple second value combinations, and each second value combination corresponds to one Actual return value.
  • the second value combination corresponding to the target display information and the actual return value corresponding to the second value combination may be the value combination of the target display information in the historical actual display and the actual return value of the value combination in the historical actual display, respectively.
  • the second value combination corresponding to the target display information may be the historical value combination of the target display information in the most recent period, that is, all the value combinations used by the target display information in the actual display in the past period of time, and
  • the actual return value corresponding to the second value combination is the actual return value of the target display information corresponding to all the used value combinations in the actual display of the period of time.
  • the target display information may exist in a corresponding booth, and the target display information may have been displayed in multiple different booths in the past.
  • the target booth currently corresponding to the target display information can be determined, so that the corresponding The two-value combination can be the value combination of the target display information corresponding to the target booth, that is, all the value combinations used by the target display information in the actual display of the target booth in the past, and the actual return value corresponding to the second value combination is
  • the information displayed for the target corresponds to the actual return value of all value combinations used in the actual display of the target booth.
  • the target display information may exist in a corresponding display period.
  • the display period here refers to a certain period of the day.
  • the target display information may have been displayed in multiple different display periods in the past.
  • the target display information is currently Corresponding target display period, so that the second value combination corresponding to the target display information can be the value combination of the target display information corresponding to the target display period, that is, all the target display information used in the actual display of the target display period in the past
  • the actual return value corresponding to the second value combination is the actual return value of the target display information corresponding to all the used value combinations in the actual display of the target display period.
  • the target display period and target booth currently corresponding to the target display information can be determined, so that the second value combination corresponding to the target display information can be the value combination of the target display information corresponding to the target booth and target display period, that is, the target display All value combinations used by the information in the target display period and the actual display of the target booth in the past, the actual return value corresponding to the second value combination is the time of the target display information in the target display period and in the target booth The actual return value corresponding to all value combinations used in the display.
  • the above training samples can be directly obtained from the data storage space corresponding to the display information.
  • the first prediction model can be trained according to the training samples 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 the above learning method, and is not limited to this training method, the following implementation The examples are for illustrative purposes only.
  • Input a set of proposed value and expected value (that is, a first value combination) in the training sample into the first prediction model, based on the actual output of the first prediction model (the first prediction model generates a prediction for the input first value combination).
  • the difference between the return value) and the expected output (the actual return value corresponding to the first value combination) adjusts the connection weight of the neural network in the model.
  • the connection weight of the neural network in the first prediction model may be randomly determined. After that, the above operation is performed on each first value combination in the training sample until the gap between the actual output and the expected output is less than the preset gap threshold.
  • the target prediction model may be updated in real time, that is, during the use of the target prediction model, the target impression information is actually displayed once more, and the actual impression will be collected after the impression Corresponding actual return value, and using the value combination and actual return value corresponding to the actual display as new training data, constantly revise the target prediction model to make the target decision model more and more accurate.
  • the first prediction model can be trained using the second value combination corresponding to the target display information and the actual return value corresponding to the second value combination to obtain the target prediction model. Because the second value combination corresponding to the target display information and the actual return value corresponding to the second value combination 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 training sample may also include a third value combination corresponding to the historical display information and The actual return value corresponding to the third value combination.
  • the historical display information may include display information corresponding to the target display information, for example, display information of the same type as the target display information or having a competitive relationship.
  • the same type may refer to the same type of information, for example, the same type of display information, the same type of video display information, and the same type of image display information.
  • the existence of a competitive relationship may refer to the same target audience for displaying information, for example, both of which are technology videos and funny images.
  • the third value combination and the third value combination of the historical display information corresponding to the target display information are also considered, so that The collected data is more comprehensive, and the historical display information can also reflect the display information display situation of the global market to a certain extent, compared to using only the second value combination corresponding to the target display information and the actual return value of the second value combination , The target prediction model obtained will be more accurate.
  • step 13 the proposed value in the first value combination with the largest predicted return value is determined as the target intended value of the target display information, and the expected value in the first value combination with the largest predicted return value is determined as the target display information.
  • Target expected value is determined as the target display information.
  • the proposed value in the largest first value combination can be determined as the target intended value of the target display information, and the expected value in the first value combination with the largest predicted return value can be determined as the target Demonstrate the expected value of the information.
  • the proposed value corresponding to the maximum predicted return value may be directly determined as the target proposed value and the corresponding expected value as the target Expected value. After that, you can use the set of target value and target expected value to compete for the display rights of the booth.
  • step 13 if there are multiple first value combinations with the largest predicted return value, the first value combination with the lowest expected value is determined as the target first value combination; the target The proposed value in the first value combination is determined as the target intended value of the target display information, and the expected value in the target first value combination is determined as the target expected value of the target display information.
  • the lower the expected value can guarantee the less value of the current cost, and the current expected value can be reduced as much as possible while the return value is guaranteed to reduce the current need Expected value of cost, that is to reduce the cost of the current cycle. Therefore, if there are multiple first value combinations with the largest predicted return value, the first value combination with the lowest expected value can be determined as the target first value combination. Then, the proposed value in the target first value combination is determined as the target proposed value of the target display information, and the expected value in the target first value combination is determined as the target expected value of the target display information, in order to use the set of target prepared value Compete with the target's expected value for the display rights of the booth.
  • a group can be randomly selected from these first value combinations, and the proposed value of the group of value combinations can be used as the target Value, and use the expected value of this group of value combinations as the target expected value, and then use the group's target to formulate the value and the target expected value to compete for the exhibition authority of the booth.
  • the target proposed value and target expected value be automatically generated for the target display information, but there is no need to manually determine the planned value and expected value, thereby saving manpower and improving efficiency, but also making the generated target proposed value and target Expect value to get the best return value possible, so as to maximize user benefits.
  • FIG. 3 is a block diagram of a value combination determination apparatus according to an exemplary embodiment of the present disclosure. As shown in FIG. 3, the device 30 may include:
  • the generating module 31 is configured to generate a plurality of first value combinations corresponding to the target display information, wherein the value combinations corresponding to the display information include the planned value of the display information and the expected value of the display information;
  • the processing module 32 is configured to input the plurality of first value combinations into the target prediction model to obtain the corresponding predicted return value;
  • the determining module 33 is configured to determine the planned value in the first value combination with the largest predicted return value as the target planned value of the target display information, and determine the expected value in the first value combination with the largest predicted return value as the target display information. The expected value of the goal.
  • the target prediction model is obtained as follows:
  • the training sample includes at least a second value combination corresponding to the target display information and an actual return value corresponding to the second value combination.
  • the second value combination is the historical value of the target display information.
  • the training sample further includes a third value combination corresponding to the historical display information and an actual return value corresponding to the third value combination.
  • the historical display information is display information that has a corresponding relationship with the target display information.
  • the generation module 31 includes:
  • a generation submodule is used to generate multiple first value combinations according to the expected value range and the planned value range.
  • the determination module 33 includes:
  • the first determining submodule is used to determine the first value combination with the lowest expected value as the target first value combination when there are multiple first value combinations with the largest predicted return value;
  • the second determination submodule is used to determine the planned value in the target first value combination as the target planned value of the target display information, and determine the expected value in the target first value combination as the target expected value of the target display information.
  • the electronic device 1900 may be provided as a server. 4, 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 one or more modules each corresponding to a set of instructions.
  • the processor 1922 may be configured to execute the computer program to execute the above-mentioned value combination determination method.
  • 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 value combination determination 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 complete the aforementioned value combination determination method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • User Interface Of Digital Computer (AREA)

Abstract

本公开涉及一种价值组合确定方法、装置、存储介质及电子设备。所述方法包括:生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;将所述多个第一价值组合输入目标预测模型,以获得相应的预测回报值;将预测回报值最大的第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为所述目标展示信息的目标期望价值。这样,不仅可以节省人力、提升效率,同时还可以使用户利益最大化。

Description

价值组合确定方法、装置、存储介质及电子设备
相关申请的交叉引用
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201811644344.0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。
技术领域
本公开涉及计算机技术领域,具体地,涉及一种价值组合确定方法、装置、存储介质及电子设备。
背景技术
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,每条展示信息都有拟定价值,通过展示此类展示信息可获得相应的价值。因而,针对这类展示信息,通常根据其拟定价值的高低顺序对其进行展示,即拟定价值越高,其展示优先级越高。在实际情况中,不同的展示信息提供者会给出与展示信息相对应的拟定价值和期望价值,展示信息的期望价值可以认为是该展示信息在拟定价值下的一个展示周期所需要花费的总价值。展示信息在实际展示后,会得到回报值,也就是在展示期间所得到的回报,而展示信息展示的最终目的就是得到高回报值。现有技术中,对于拟定价值和期望价值,需要人工收集数据后进行人为分析后进行确定。这样,一方面需要投入大量的人力、物力、时间等,效率很低;另一方面,由于人工分析的方式受限于经验、环境等因素,确定出的拟定价值和期望价值可能无法实时满足要求,而且准确性可能不足。
发明内容
根据本公开的第一方面,提供一种价值组合确定方法,所述方法包括:
生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;
将所述多个第一价值组合输入目标预测模型,以获得相应的预测回报值;
将预测回报值最大的第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为所述目标展示信息的目标期望价值。
可选地,所述目标预测模型通过如下方式获得:
获取与所述目标展示信息对应的训练样本,所述训练样本中至少包括所述目标展示信息对应的第二价值组合和与所述第二价值组合对应的实际回报值,其中,所述第二价值组合为所述目标展示信息在历史的实际展示中对应的价值组合;
根据所述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。
可选地,所述训练样本还包括历史展示信息对应的第三价值组合和与所述第三价值组合对应的实际回报值,所述历史展示信息为与所述目标展示信息存在对应关系的展示信息。
可选地,所述生成目标展示信息对应的多个第一价值组合包括:
获取所述目标展示信息对应的期望价值范围和拟定价值范围;
根据所述期望价值范围和所述拟定价值范围,生成所述多个第一价值组合。
可选地,若预测回报值最大的第一价值组合为多个,则将其中期望价值最低的第一价值组合确定为目标第一价值组合;
将所述目标第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将所述目标第一价值组合中的期望价值确定为所述目标展示信息的所述目标期望价值。
根据本公开的第二方面,提供一种价值组合确定装置,所述装置包括:
生成模块,用于生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;
处理模块,用于将所述多个第一价值组合输入目标预测模型,以获得相应 的预测回报值;
确定模块,用于将预测回报值最大的第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为所述目标展示信息的目标期望价值。
可选地,所述目标预测模型通过如下方式获得:
获取与所述目标展示信息对应的训练样本,所述训练样本中至少包括所述目标展示信息对应的第二价值组合和与所述第二价值组合对应的实际回报值,其中,所述第二价值组合为所述目标展示信息在历史的实际展示中对应的价值组合;
根据所述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。
可选地,所述训练样本还包括历史展示信息对应的第三价值组合和与所述第三价值组合对应的实际回报值,所述历史展示信息为与所述目标展示信息存在对应关系的展示信息。
可选地,所述生成模块包括:
获取子模块,用于获取所述目标展示信息对应的期望价值范围和拟定价值范围;
生成子模块,用于根据所述期望价值范围和所述拟定价值范围,生成所述多个第一价值组合。
可选地,所述确定模块包括:
第一确定子模块,用于在预测回报值最大的第一价值组合为多个时,将其中期望价值最低的第一价值组合确定为目标第一价值组合;
第二确定子模块,用于将所述目标第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将所述目标第一价值组合中的期望价值确定为所述目标展示信息的所述目标期望价值。
根据本公开的第三方面,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。
根据本公开的第四方面,提供一种电子设备,包括:
存储器,其上存储有计算机程序;
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。
根据本公开的实施例,不仅可以针对目标展示信息自动生成目标拟定价值和目标期望价值,而无需人为确定拟定价值和期望价值,从而节省人力且提升效率,还可以使得生成的目标拟定价值和目标期望价值尽可能获得最优的回报值,从而使用户利益最大化。
附图说明
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:
图1是根据本公开的示例性实施例的价值组合确定方法的流程图;
图2是在根据本公开的示例性实施例的价值组合确定方法中,生成目标展示信息对应的多个第一价值组合的步骤的一种示例性实现方式的流程图;
图3是根据本公开的示例性实施例的价值组合确定装置的框图;
图4是根据本公开的示例性实施例的一种电子设备的框图。
具体实施方式
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。
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。
图1是根据本公开的示例性实施例的价值组合确定方法的流程图。如图1所示,该方法可以包括以下步骤11至13。
在步骤11中,生成目标展示信息对应的多个第一价值组合。
示例地,展示信息可以例如是文字、图片、视频或其组合等。展示信息对应的价值组合可以包括该展示信息的拟定价值和该展示信息的期望价值。展示信息的拟定价值可以认为是该展示信息在展位上展示后被点击或观看一次所需要花费的价值,展位可以理解为用于对展示信息进行展示的虚拟或实际的载体,而展示信息的期望价值则可以认为是该展示信息在拟定价值下被点 击或观看期望次数(即,一个展示周期)所需要花费的总价值。当存在一个空闲的展位,用户可以给出拟定价值和预算价值。在多个用户给出的多个拟定价值和多个预算价值中,当前最高的拟定价值所对应的用户将在当前拥有在该展位展示的最高优先级。
针对目标展示信息,可以生成与目标展示信息对应的多个第一价值组合,每个第一价值组合包括拟定价值和期望价值。可选地,由上文中所提及的拟定价值和期望价值的概念可知,一般情况下,同一价值组合中的期望价值是大于或者等于拟定价值的,因此,生成第一价值组合时可以考虑该因素,也就是说,在生成与目标展示信息对应的多个第一价值组合时,可以设置例如同一第一价值组合中期望价值大于或者等于拟定价值的条件,以使生成的第一价值组合均是可使用的,提升第一价值组合的生成效率。
在一种可能的实施例中,针对目标展示信息,可以例如随机地直接生成多个第一价值组合。示例地,可以例如从正实数中随机选取拟定价值和期望价值。
在另一种可能的实施例中,如图2所示,步骤11可以包括以下步骤21至22。
在步骤21中,获取目标展示信息对应的期望价值范围和拟定价值范围。
示例地,针对不同的展示信息,可以为各种展示信息的价值进行预先估计,也就是说,针对不同的展示信息,可以预先为其设置各自的期望价值范围和拟定价值范围。其中,该期望价值范围和拟定价值范围可以根据展示信息的内容、展示信息的历史拟定价值和历史期望价值进行设定。因此,针对目标展示信息,可以获取该目标展示信息对应的期望价值范围和拟定价值范围。
在步骤22中,根据期望价值范围和拟定价值范围,生成多个第一价值组合。
根据目标展示信息对应的期望价值范围和拟定价值范围,可以生成多个第一价值组合。示例地,可以在期望价值范围内随机选取一个期望价值,并在拟定价值范围内随机选取一个拟定价值,从而生成以该期望价值和该拟定价值构成的第一价值组合,而后按照上述方式生成多个第一价值组合,直至第一价值组合的个数满足预设的要求。再例如,可以针对期望价值范围和拟定价值范围,对期望价值范围和拟定价值范围内的可选期望价值和可选拟定价值进 行遍历,以获得多个第一价值组合。可选地,生成的多个第一价值组合可以是互异的。
通过上述方式,根据目标展示信息对应的期望价值范围和拟定价值范围,生成多个第一价值组合,可以使得生成的第一价值组合更具有针对性,更加适合目标展示信息,实用性强。
现在回到图1,在步骤12中,将该多个第一价值组合输入目标预测模型,以获得相应的预测回报值。
目标预测模型可以根据输入的第一价值组合生成并输出与第一价值组合对应的预测回报值。下面将首先对目标预测模型的获得方式进行举例说明。
示例地,目标预测模型可以通过如下方式获得:
获取与目标展示信息对应的训练样本;
根据训练样本对第一预测模型进行训练,以获得目标预测模型。
在一种可能的实施例中,训练样本中可以包括目标展示信息对应的第二价值组合和第二价值组合对应的实际回报值。其中,第二价值组合可以为目标展示信息在历史的实际展示中对应的价值组合。
与目标展示信息对应的训练样本可以包括目标展示信息对应的第二价值组合和与第二价值组合对应的实际回报值,其中,第二价值组合可以为多个,每个第二价值组合对应一个实际回报值。目标展示信息对应的第二价值组合和与第二价值组合对应的实际回报值可以分别为目标展示信息在历史的实际展示中的价值组合和该价值组合在历史的实际展示中的实际回报值。示例地,目标展示信息对应的第二价值组合可以为目标展示信息在最近一段时间内的历史价值组合,也就是目标展示信息在过去一段时间的实际展示中所使用过的所有的价值组合,而与该第二价值组合对应的实际回报值则为目标展示信息在该段时间的实际展示中对应于使用过的所有价值组合的实际回报值。再例如,目标展示信息可以存在对应的展位,而目标展示信息过去可能在多个不同的展位上进行过展示,因此,可以确定目标展示信息当前对应的目标展位,从而,目标展示信息对应的第二价值组合可以为目标展示信息对应于目标展位的价值组合,也就是目标展示信息过去在目标展位的实际展示中所使用过的所有价值组合,而与该第二价值组合对应的实际回报值则为目标展示信息 在该目标展位的实际展示中对应于使用过的所有价值组合的实际回报值。再例如,目标展示信息可以存在对应的展示时段,这里的展示时段是指一天中的某个时段,目标展示信息过去可能在多个不同的展示时段进行过展示,因此,可以确定目标展示信息当前对应的目标展示时段,从而,目标展示信息对应的第二价值组合可以为目标展示信息对应于目标展示时段的价值组合,也就是目标展示信息过去在目标展示时段的实际展示中所使用过的所有价值组合,与该第二价值组合对应的实际回报值则为目标展示信息在该目标展示时段的实际展示中对应于使用过的所有价值组合的实际回报值。再例如,可以确定目标展示信息当前对应的目标展示时段和目标展位,从而,目标展示信息对应的第二价值组合可以为目标展示信息对应于目标展位和目标展示时段的价值组合,也就是目标展示信息过去在目标展示时段且在目标展位的实际展示中所使用过的所有价值组合,与该第二价值组合对应的实际回报值则为目标展示信息在该目标展示时段且在该目标展位的时间展示中对应于使用过的所有价值组合的实际回报值。
示例地,上述训练样本可以直接从展示信息对应的数据存储空间中获取到。在获取到与目标展示信息对应的训练样本后,可以根据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。
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上述学习方法,并且,也不局限于这一种训练方式,下述实施例仅作为示例性说明。
将训练样本中的一组拟定价值和期望价值(即一个第一价值组合)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价值组合所生成的预测回报值)与期望输出(与第一价值组合对应的实际回报值)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确定的。之后,对训练样本中的每一个第一价值组合执行上述操作,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,目标展示信息每多实际展示一次,在展示后则会收集该次实际展示所对应的实际回报值,并将该次实际展示对应的价值组合和实际回报值作为新的训练数据,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。
通过上述方式,可以利用目标展示信息对应的第二价值组合和第二价值组合对应的实际回报值对第一预测模型进行训练,以得到目标预测模型。由于目标展示信息对应的第二价值组合和第二价值组合对应的实际回报值易于获得,便于进行数据采集,同时便于获得目标预测模型。
在另一种可能的实施例中,训练样本中除了包括目标展示信息对应的第二价值组合和第二价值组合对应的实际回报值之外,还可以包括历史展示信息对应的第三价值组合和与第三价值组合对应的实际回报值。其中,历史展示信息可以包括与目标展示信息存在对应关系的展示信息,例如与目标展示信息类型相同或存在竞争关系的展示信息。示例地,类型相同可以指信息类型相同,例如同为文字类展示信息、同为视频类展示信息、同为图片类展示信息等。再例如,存在竞争关系可以指展示信息的目标受众相同,例如同为科技类视频、同为搞笑类图片等。
对于训练样本的限定及获取、以及利用训练样本对第一预测模型训练以得到目标预测模型等的相关说明均与上文中相似且原理相同,此处不赘述。
通过上述方式,在获取目标展示信息相关的第二价值组合和实际回报值之外,还考虑了与目标展示信息对应的历史展示信息的第三价值组合和第三价值组合的实际回报值,使得收集的数据更加全面,并且,历史展示信息也能够一定程度上反应全局市场的展示信息展示情况,相比于仅利用目标展示信息对应的第二价值组合及第二价值组合的实际回报值来说,获得的目标预测模型将会更加精确。
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一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。
在步骤13中,将预测回报值最大的第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为目标展示信息的目标期望价值。
针对得到的多个预测回报值,可以将其中最大的第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为目标展示信息的目标期望价值。
在一种可能的实施例中,可能存在一个预测回报值最大的第一价值组合,那么可以直接将该最大预测回报值对应的拟定价值确定为目标拟定价值并将其对应的期望价值确定为目标期望价值。此后,可以利用该组目标拟定价值和目标期望价值竞争展位的展示权限。
在另一种可能的实施例中,在步骤13中,若预测回报值最大的第一价值组合为多个,则将其中期望价值最低的第一价值组合确定为目标第一价值组合;将目标第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将目标第一价值组合中的期望价值确定为目标展示信息的目标期望价值。
实际应用中,在能够得到相同的回报值的情况下,期望价值越低可以保证当前花费的价值越少,可以在保证回报值的情况下尽可能降低当前的期望价值,以减小当前所需要花费的期望价值,也就是减小当前周期的花费。因此,若预测回报值最大的第一价值组合为多个,可以将其中期望价值最低的第一价值组合确定为目标第一价值组合。而后,将目标第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将目标第一价值组合中的期望价值确定为目标展示信息的目标期望价值,以便利用该组目标拟定价值和目标期望价值竞争展位的展示权限。
采用上述方式,可以在保证高回报值的情况下,在给出期望价值时保证当前所花费的期望价值是最低的。
在另一种可能的实施例中,若预测回报值最大的第一价值组合为多个,那么可以从这些第一价值组合中随机选择一组,并将该组价值组合的拟定价值作为目标拟定价值,并将该组价值组合的期望价值作为目标期望价值,之后利用该组目标拟定价值和目标期望价值竞争展位的展示权限。
根据本公开的实施例,不仅可以针对目标展示信息自动生成目标拟定价值和目标期望价值,而无需人为确定拟定价值和期望价值,从而节省人力且提升效率,还可以使得生成的目标拟定价值和目标期望价值尽可能获得最优的回报值,从而使用户利益最大化。
图3是根据本公开的示例性实施例的价值组合确定装置的框图。如图3所示,该装置30可以包括:
生成模块31,用于生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;
处理模块32,用于将该多个第一价值组合输入目标预测模型,以获得相应的预测回报值;
确定模块33,用于将预测回报值最大的第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为目标展示信息的目标期望价值。
可选地,目标预测模型通过如下方式获得:
获取与目标展示信息对应的训练样本,该训练样本中至少包括目标展示信息对应的第二价值组合和与第二价值组合对应的实际回报值,其中,第二价值组合为目标展示信息在历史的实际展示中对应的价值组合;
根据训练样本对第一预测模型进行训练,以获得目标预测模型。
可选地,训练样本还包括历史展示信息对应的第三价值组合和与第三价值组合对应的实际回报值,历史展示信息为与目标展示信息存在对应关系的展示信息。
可选地,生成模块31包括:
获取子模块,用于获取目标展示信息对应的期望价值范围和拟定价值范围;
生成子模块,用于根据期望价值范围和拟定价值范围,生成多个第一价值组合。
可选地,确定模块33包括:
第一确定子模块,用于在预测回报值最大的第一价值组合为多个时,将其 中期望价值最低的第一价值组合确定为目标第一价值组合;
第二确定子模块,用于将目标第一价值组合中的拟定价值确定为目标展示信息的目标拟定价值,并将目标第一价值组合中的期望价值确定为目标展示信息的目标期望价值。
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。
图4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4,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一个或一个以上的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,以执行上述的价值组合确定方法。
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的操作系统,例如Windows Server TM、Mac OS X TM、Unix TM、Linux TM等等。
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价值组合确定方法。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执行以完成上述的价值组合确定方法。
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。
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。
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。

Claims (12)

  1. 一种价值组合确定方法,包括:
    生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;
    将所述多个第一价值组合输入目标预测模型,以获得相应的预测回报值;
    将预测回报值最大的第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为所述目标展示信息的目标期望价值。
  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如下方式获得:
    获取与所述目标展示信息对应的训练样本,所述训练样本中至少包括所述目标展示信息对应的第二价值组合和与所述第二价值组合对应的实际回报值,其中,所述第二价值组合为所述目标展示信息在历史的实际展示中对应的价值组合;
    根据所述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。
  3. 根据权利要求2所述的方法,其中,所述训练样本还包括历史展示信息对应的第三价值组合和与所述第三价值组合对应的实际回报值,所述历史展示信息为与所述目标展示信息存在对应关系的展示信息。
  4. 根据权利要求1-3中任一项所述的方法,其中,所述生成目标展示信息对应的多个第一价值组合包括:
    获取所述目标展示信息对应的期望价值范围和拟定价值范围;
    根据所述期望价值范围和所述拟定价值范围,生成所述多个第一价值组合。
  5. 根据权利要求1-3中任一项所述的方法,其中,
    若预测回报值最大的第一价值组合为多个,则将其中期望价值最低的第一价值组合确定为目标第一价值组合;
    将所述目标第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将所述目标第一价值组合中的期望价值确定为所述目标展示信息的所述目标期望价值。
  6. 一种价值组合确定装置,包括:
    生成模块,用于生成目标展示信息对应的多个第一价值组合,其中,展示信息对应的价值组合包括该展示信息的拟定价值和该展示信息的期望价值;
    处理模块,用于将所述多个第一价值组合输入目标预测模型,以获得相应的预测回报值;
    确定模块,用于将预测回报值最大的第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将预测回报值最大的第一价值组合中的期望价值确定为所述目标展示信息的目标期望价值。
  7. 根据权利要求6所述的装置,其中,所述目标预测模型通过如下方式获得:
    获取与所述目标展示信息对应的训练样本,所述训练样本中至少包括所述目标展示信息对应的第二价值组合和与所述第二价值组合对应的实际回报值,其中,所述第二价值组合为所述目标展示信息在历史的实际展示中对应的价值组合;
    根据所述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。
  8. 根据权利要求7所述的装置,其中,所述训练样本还包括历史展示信息对应的第三价值组合和与所述第三价值组合对应的实际回报值,所述历史展示信息为与所述目标展示信息存在对应关系的展示信息。
  9. 根据权利要求6-8中任一项所述的装置,其中,所述生成模块包括:
    获取子模块,用于获取所述目标展示信息对应的期望价值范围和拟定价 值范围;
    生成子模块,用于根据所述期望价值范围和所述拟定价值范围,生成所述多个第一价值组合。
  10. 根据权利要求6-8中任一项所述的装置,其中,所述确定模块包括:
    第一确定子模块,用于在预测回报值最大的第一价值组合为多个时,将其中期望价值最低的第一价值组合确定为目标第一价值组合;
    第二确定子模块,用于将所述目标第一价值组合中的拟定价值确定为所述目标展示信息的目标拟定价值,并将所述目标第一价值组合中的期望价值确定为所述目标展示信息的所述目标期望价值。
  11. 一种计算机可读存储介质,其上存储有计算机程序,所述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。
  12. 一种电子设备,包括:
    存储器,其上存储有计算机程序;
    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。
PCT/CN2019/094747 2018-12-29 2019-07-04 价值组合确定方法、装置、存储介质及电子设备 Ceased WO2020134012A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811644344.0A CN111401593A (zh) 2018-12-29 2018-12-29 价值组合确定方法、装置、可读存储介质及电子设备
CN201811644344.0 2018-12-29

Publications (1)

Publication Number Publication Date
WO2020134012A1 true WO2020134012A1 (zh) 2020-07-02

Family

ID=71128255

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/094747 Ceased WO2020134012A1 (zh) 2018-12-29 2019-07-04 价值组合确定方法、装置、存储介质及电子设备

Country Status (2)

Country Link
CN (1) CN111401593A (zh)
WO (1) WO2020134012A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006138400A1 (en) * 2005-06-15 2006-12-28 Google, Inc. Time-multiplexing documents based on preferences or relatedness
CN102592235A (zh) * 2011-12-28 2012-07-18 北京品友互动信息技术有限公司 一种互联网广告投放系统
CN103136311A (zh) * 2011-12-05 2013-06-05 微软公司 付费搜索中的投标和预算分配的联合优化
CN103164805A (zh) * 2011-12-19 2013-06-19 阿里巴巴集团控股有限公司 关键词投放价格优化处理方法及系统
CN105431875A (zh) * 2013-08-05 2016-03-23 雅虎公司 搜索引擎营销优化器
CN107194729A (zh) * 2017-05-22 2017-09-22 北京京东尚科信息技术有限公司 广告竞价方法、装置、电子装置及计算机可读介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006138400A1 (en) * 2005-06-15 2006-12-28 Google, Inc. Time-multiplexing documents based on preferences or relatedness
CN103136311A (zh) * 2011-12-05 2013-06-05 微软公司 付费搜索中的投标和预算分配的联合优化
CN103164805A (zh) * 2011-12-19 2013-06-19 阿里巴巴集团控股有限公司 关键词投放价格优化处理方法及系统
CN102592235A (zh) * 2011-12-28 2012-07-18 北京品友互动信息技术有限公司 一种互联网广告投放系统
CN105431875A (zh) * 2013-08-05 2016-03-23 雅虎公司 搜索引擎营销优化器
CN107194729A (zh) * 2017-05-22 2017-09-22 北京京东尚科信息技术有限公司 广告竞价方法、装置、电子装置及计算机可读介质

Also Published As

Publication number Publication date
CN111401593A (zh) 2020-07-10

Similar Documents

Publication Publication Date Title
CN113516255B (zh) 联邦学习建模优化方法、设备、可读存储介质及程序产品
Koch et al. Optimising resource costs of cloud computing for education
US10423631B2 (en) Automated data exploration and validation
CN110599393B (zh) 图片风格转换方法、装置、设备及计算机可读存储介质
Hassani et al. Context-aware recruitment scheme for opportunistic mobile crowdsensing
US10313746B2 (en) Server, client and video processing method
US12189629B2 (en) Optimizing job runtimes via prediction-based token allocation
US20150178134A1 (en) Hybrid Crowdsourcing Platform
WO2019117970A1 (en) Adaptive object tracking policy
US20170228349A1 (en) Combined predictions methodology
US20180012378A1 (en) Determining visibility of rendered content
CN109697637A (zh) 对象类别确定方法、装置、电子设备及计算机存储介质
CN115185606B (zh) 业务配置参数的获得方法、装置、设备及存储介质
WO2023116138A1 (zh) 多任务模型的建模方法、推广内容处理方法及相关装置
Jarrett et al. Combining human and machine computing elements for analysis via crowdsourcing
Li et al. PTASIM: Incentivizing crowdsensing with POI-tagging cooperation over edge clouds
Stone et al. Methods for determining the optimal number of simultaneous contributors for multi-user CAD parts
JP2019040417A (ja) 情報処理装置及びプログラム
Singh et al. Asynchronous federated learning with grey wolf optimization for the heterogeneity iot devices
WO2019085356A1 (zh) 一种净推荐值的显示方法及其终端
CN109034085B (zh) 用于生成信息的方法和装置
US10482403B2 (en) Methods and systems for designing of tasks for crowdsourcing
WO2020134012A1 (zh) 价值组合确定方法、装置、存储介质及电子设备
US20140019394A1 (en) Providing expert elicitation
CN119168086A (zh) 数据处理方法、装置和电子设备

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: 19903911

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 01/10/2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19903911

Country of ref document: EP

Kind code of ref document: A1