WO2017157159A1 - 一种推送增值业务信息的方法、装置及电子设备 - Google Patents

一种推送增值业务信息的方法、装置及电子设备 Download PDF

Info

Publication number
WO2017157159A1
WO2017157159A1 PCT/CN2017/075063 CN2017075063W WO2017157159A1 WO 2017157159 A1 WO2017157159 A1 WO 2017157159A1 CN 2017075063 W CN2017075063 W CN 2017075063W WO 2017157159 A1 WO2017157159 A1 WO 2017157159A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
added service
service information
ecpm
correction coefficient
Prior art date
Application number
PCT/CN2017/075063
Other languages
English (en)
French (fr)
Inventor
张旭东
段全盛
Original Assignee
北京金山安全软件有限公司
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 北京金山安全软件有限公司 filed Critical 北京金山安全软件有限公司
Publication of WO2017157159A1 publication Critical patent/WO2017157159A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR 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
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241Advertisements
    • G06Q30/0251Targeted advertisements
    • G06Q30/0255Targeted advertisements based on user history
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR 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
    • G06Q30/00Commerce
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR 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
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0283Price estimation or determination

Definitions

  • the present invention relates to value-added service application technologies, and in particular, to a method, device and electronic device for pushing value-added service information.
  • value-added service operators have seen opportunities to place value-added service information based on electronic devices. For example, value-added service operators push value-added service information to mobile phones through value-added service platforms provided by Internet operators. Wide, easy to use, and high efficiency of value-added services.
  • the value-added service platform can deliver value-added service information of multiple charging types based on multiple types of value-added service charging types, where the charging type includes but is not limited to: per thousand times Show how to charge (CPM, Cost-Per-Mille), how to charge per click (CPC, Cost-Per-Click), how to charge per installation (CPI, Cost-Per-Install) , according to the method of charging per action (CPA, Cost-Per-Action), charging according to each purchase (CPS, Cost-Per-Sale), charging according to each video broadcast (CPV, Cost) -Per-Video)etc.
  • the value-added service platform proposes a method for sorting and displaying the value-added service information to be pushed, and separately sets the value-added service information of different charging types.
  • the value-added service information of different charging types uses different algorithms to calculate eCPM. The following briefly describes the process of calculating eCPM:
  • Step 1 training a corresponding value-added service information model for calculating CTR/CVR/ICVR according to a history log of each value-added service information;
  • Step 2 Selecting a candidate value-added service information set that meets the targeting condition from the value-added service information set according to the orientation condition in the value-added service information request, for example, the target condition of the country, gender, and age required to be delivered.
  • Step 3 Extract, for each value-added service information in the candidate value-added service information set, user attributes (eg, gender, age, country, language, device type, and the like) of the value-added service information, and value-added service information attributes (eg, material size, Effect class, conversion class, industry classification), context attribute (for example, media, value-added service information bit), generate model value-added service feature information;
  • user attributes eg, gender, age, country, language, device type, and the like
  • value-added service information attributes eg, material size, Effect class, conversion class, industry classification
  • context attribute for example, media, value-added service information bit
  • Step 4 input the generated model value-added service feature information into the trained value-added service information model, and calculate the CTR/CVR/ICVR of the value-added service information;
  • Step 5 After calculating the CTR/CVR/ICVR of the value-added service information, calculate the eCPM according to the eCPM calculation formula corresponding to the charging type of the value-added service information, and then use the product of the eCPM and the supporting coefficient to the candidate value-added service.
  • the value-added service information in the information set is sorted, and the top-ranked value-added service information is returned to the user.
  • CTR, CVR, ICVR and other indicators are usually trained on different data sets using different models, for example, using Logistic Regression (LR, Logistic Regression) model, gradient lifting decision tree (GBDT, Gradient) Boosting Decision Tree) model.
  • LR Logistic Regression
  • GBDT Gradient
  • Boosting Decision Tree Gradient Boosting Decision Tree
  • the effect of the type-added value-added service information is not good, so that the "premium" value-added service information of a certain type of billing is lower than the eCPM of the "low-quality" value-added service information under the other billing type.
  • the value-added service information pushed is not efficient; further, the support factor is predetermined, so that when the eCPM distribution between the value-added service information of different charging types changes greatly, the supporting coefficient cannot be satisfied. Calculation needs; moreover, the value of the support factor depends on the type of billing, without considering other factors such as user factors, advertising factors and context And other factors, resulting in a lower calculation of the eCPM, push value-added business information is not efficient.
  • the embodiments of the present invention provide a method, a device, and an electronic device for pushing value-added service information, which improve the push efficiency of value-added service information.
  • an embodiment of the present invention provides a method for pushing value-added service information, including:
  • Parsing the received value-added service information delivery request obtaining the feature information included in the value-added service information delivery request, and selecting a candidate value-added service information set that meets the feature information from the preset value-added service information set;
  • the service characteristic information of the extracted value-added service information is written into a preset universal dimension group to obtain a dimension group of the value-added service information;
  • mapping relationship between the dimension group and the eCPM correction coefficient includes:
  • the value of the dimension other than the value-added service information group identifier in the common dimension group is fixed, the value-added service information group identifier is traversed, and the eCPM correction coefficient is calculated.
  • the average estimated eCPM is calculated using the following formula:
  • eCPM HISTORY (dims(n)) is the average estimated eCPM
  • C(dims(n)) is a collection of historical logs
  • Ac(i) is the behavior type of the i-th value-added service information, wherein the behavior type includes: presentation, click, and installation, the presentation identifier is imp, the click identifier is clk, and the installation identifier is ins;
  • eCPM PREDICT (i) is the estimated eCPM value of the i-th value-added service information.
  • the eCPM correction coefficient is calculated by using the following formula:
  • Delta(b, v 2 ,..., v n ) is the eCPM correction coefficient
  • b i is the i-th accounting type
  • b is the specified billing type
  • m is the type of billing type included in the dimension group.
  • the method further includes:
  • the historical true eCPM is calculated based on the pre-stored presentation history log.
  • the historical real eCPM is calculated using the following formula:
  • eCPM REAL (dims(n)) is historical real eCPM
  • TOTAL_COST(dims(n)) is the total revenue corresponding to the pre-stored presentation history log
  • TOTAL_IMPRESSION(dims(n)) is the total amount of impressions corresponding to the pre-stored presentation history log.
  • the eCPM correction coefficient is calculated by using the following formula:
  • the modified eCPM is calculated using the following formula:
  • the method further includes:
  • the user behavior information of the pushed value-added service information is recorded in the history log.
  • an embodiment of the present invention provides an apparatus for pushing value-added service information, including: a candidate value-added service information set acquisition module, a charging coefficient calculation module, an eCPM calculation module, a dimension group acquisition module, an eCPM correction coefficient query module, and a value-added service.
  • Information push module wherein
  • the candidate value-added service information collection module is configured to parse the received value-added service information delivery request, obtain the feature information included in the value-added service information delivery request, and select the qualified information from the preset value-added service information set.
  • Candidate value-added service information set is configured to parse the received value-added service information delivery request, obtain the feature information included in the value-added service information delivery request, and select the qualified information from the preset value-added service information set.
  • a charging coefficient calculation module configured to sequentially extract service characteristic information of each value-added service information in the candidate value-added service information set, input the pre-trained value-added service information model, and calculate a charging coefficient for each value-added service information
  • the eCPM calculation module is configured to calculate, according to the charging coefficient corresponding to the charging type of the value-added service information, a charging eCPM displayed by the expected value-added service information per thousand times;
  • a dimension group obtaining module configured to write the service characteristic information of the extracted value-added service information into a preset universal dimension group, to obtain a dimension group of the value-added service information
  • the eCPM correction coefficient querying module is configured to query the mapping relationship between the pre-stored charging type eCPM correction coefficient centralized dimension group and the eCPM correction coefficient according to the charging type of the value-added service information, and obtain the eCPM of the dimensional group mapping of the value-added service information. Correction factor;
  • the value-added service information pushing module is configured to apply the eCPM correction coefficient to the preset eCPM correction formula to obtain a modified eCPM for each value-added service information, and sort and push the sorted value-added service information according to the calculated corrected eCPM.
  • the eCPM correction coefficient query module includes: an average estimated eCPM calculation unit, a mapping relationship construction unit, and an eCPM correction coefficient query unit, where
  • the average estimated eCPM calculation unit is configured to calculate an average estimated eCPM according to the pre-stored presentation history log;
  • a mapping relationship construction unit configured to fix values of other dimensions of the universal dimension group except the value-added service information group identifier, traverse the value-added service information group identifier, and calculate corresponding dimension groups corresponding to the value-added service information
  • the eCPM correction coefficient is placed in the charging type eCPM correction coefficient set;
  • the eCPM correction coefficient querying unit is configured to query the mapping relationship construction unit according to the charging type of the value-added service information, and acquire an eCPM correction coefficient of the dimension group mapping of the value-added service information.
  • the average estimated eCPM is calculated using the following formula:
  • eCPM HISTORY (dims(n)) is the average estimated eCPM
  • C(dims(n)) is a collection of historical logs
  • Ac(i) is the behavior type of the i-th value-added service information, wherein the behavior type includes: presentation, click, and installation, the presentation identifier is imp, the click identifier is clk, and the installation identifier is ins;
  • eCPM PREDICT (i) is the estimated eCPM value of the i-th value-added service information.
  • the eCPM correction coefficient is calculated by using the following formula:
  • Delta(b, v 2 ,..., v n ) is the eCPM correction coefficient
  • b i is the i-th accounting type
  • b is the specified billing type
  • m is the type of billing type included in the dimension group.
  • the eCPM correction coefficient query module further includes:
  • a historical real eCPM calculation unit for calculating a historical true eCPM based on a pre-stored presentation history log.
  • the historical real eCPM is calculated using the following formula:
  • eCPM REAL (dims(n)) is historical real eCPM
  • TOTAL_COST(dims(n)) is the total revenue corresponding to the pre-stored presentation history log
  • TOTAL_IMPRESSION(dims(n)) is the total amount of impressions corresponding to the pre-stored presentation history log.
  • the eCPM correction coefficient is calculated by using the following formula:
  • the modified eCPM is calculated using the following formula:
  • the device further includes:
  • a recording module configured to record user behavior information of the pushed value-added service information into the history log.
  • an embodiment of the present invention provides an electronic device, including: a housing, a processor, a memory, a circuit board, and a power supply circuit, wherein the circuit board is disposed in a space enclosed by the housing
  • the processor and the memory are disposed on the circuit board; the power circuit is configured to supply power to each circuit or device of the electronic device; the memory is used to store executable program code; and the processor reads the executable program code stored in the memory.
  • the program corresponding to the executable program code is executed to execute the method for pushing the value-added service information according to any of the foregoing.
  • an embodiment of the present invention provides a computer program product, when the instruction processor in the computer program product executes, performing a method for pushing value-added service information, where the method includes:
  • Parsing the received value-added service information delivery request obtaining the feature information included in the value-added service information delivery request, and selecting a candidate value-added service information set that meets the feature information from the preset value-added service information set;
  • the service characteristic information of the extracted value-added service information is written into a preset universal dimension group to obtain a dimension group of the value-added service information;
  • an embodiment of the present invention provides a storage medium, when the instructions in the storage medium are executed by a processor of a server, to enable the server to perform a method for pushing value-added service information, where the method includes:
  • the service characteristic information of the extracted value-added service information is written into a preset universal dimension group to obtain a dimension group of the value-added service information;
  • the method, device, and electronic device for pushing the value-added service information provided by the embodiment of the present invention, parsing the received value-added service information delivery request, and acquiring the feature information included in the value-added service information delivery request, from the preset value-added service information set, Selecting a candidate value-added service information set that meets the feature information; sequentially extracting service feature information of each value-added service information in the candidate value-added service information set, inputting the pre-trained value-added service information model, and calculating for each value-added service
  • the charging coefficient of the information calculating the charging fee of the expected value per thousand times of the value-added service information according to the charging coefficient corresponding to the charging type of the value-added service information; writing the preset universal dimension according to the service characteristic information of the extracted value-added service information
  • the group obtains the dimension group of the value-added service information; and according to the charging type of the value-added service information, queries the pre-stored charging type eCPM correction
  • the eCPM difference between the types of value-added service information makes the overall estimated eCPM of the value-added service information more accurate, thereby optimizing the overall ranking effect of the value-added service information and improving the push efficiency of the value-added service information.
  • FIG. 1 is a schematic flowchart of a method for pushing value-added service information according to an embodiment of the present invention
  • FIG. 2 is a schematic flowchart of obtaining a mapping relationship between a dimension group and an eCPM correction coefficient according to an embodiment of the present invention
  • FIG. 3 is a schematic structural diagram of an apparatus for pushing value-added service information according to an embodiment of the present invention
  • FIG. 4 is a schematic structural view of an embodiment of an electronic device according to the present invention.
  • FIG. 1 is a schematic flowchart of an apparatus for pushing value-added service information according to an embodiment of the present invention. Referring to Figure 1, the method includes:
  • Step 11 parsing the received value-added service information delivery request, and obtaining the value-added service information delivery
  • the feature information included in the request is selected from the preset value-added service information set to select a candidate value-added service information set that meets the feature information;
  • the feature information includes, but is not limited to, one of a value-added service information group identifier, a user attribute, a value-added service information attribute, and a context attribute, or any combination thereof, where
  • the value-added service information group also known as the value-added service information series, refers to a set of value-added service information with the same budget, billing type, bidding, delivery period and targeting conditions.
  • User attributes include, but are not limited to, gender, age, country, language, device type, etc.
  • Value-added service information attributes include but are not limited to: material size, effect class, conversion class, industry classification;
  • Context attributes include, but are not limited to, media, value-added service information bits.
  • the user or the value-added service operator may initiate a value-added service information delivery request through the client/browser, and the value-added service information delivery request includes feature information such as user attributes and context information.
  • the feature information may be an orientation condition including a user attribute.
  • Step 12 sequentially extract service characteristic information of each value-added service information in the candidate value-added service information set, input the pre-trained value-added service information model, and calculate a charging coefficient for each value-added service information;
  • the service feature information of the value-added service information is extracted, including: user attributes (eg, gender, age, country, language, device type, etc.), value-added service information. Attributes (eg, material size, effect class, conversion class, industry classification), context attributes (eg, media, value-added service information bits), or any combination thereof, generate model value-added service feature information, and add value to the generated model
  • user attributes eg, gender, age, country, language, device type, etc.
  • Attributes eg, material size, effect class, conversion class, industry classification
  • context attributes eg, media, value-added service information bits
  • the CTR is calculated by inputting the service characteristic information of the value-added service information into the value-added service information model including the CPC value-added service information;
  • the value-added service information of the type CPA/CPI/CPS is calculated by inputting the service characteristic information of the value-added service information into the value-added service information model including the CPA/CPI/CPS value-added service information, thereby calculating CTR and CVR; or ICVR.
  • the service feature information includes, but is not limited to, one of a user attribute, a value-added service information attribute, and a context attribute, or any combination thereof.
  • the charging coefficient includes but is not limited to: CTR, CVR, and ICVR.
  • Step 13 Calculate, according to the charging coefficient corresponding to the charging type of the value-added service information, calculate a charging fee that is expected to be displayed per thousand times of value-added service information;
  • the expected charging per thousand times of value-added service information is calculated by using the following formula:
  • CTR Click-Through-Rate, which is the ratio of the value-added business information hits to the value-added business information.
  • Bid is the operator’s bid for thousand impressions.
  • CVR is the conversion rate (Conversion-Rate), which is the ratio of the purchase/installation amount generated by value-added service information to the value-added service information click volume.
  • the ICVR is the installation-conversion-Rate, which is the ratio of the purchase/installation amount generated by the value-added service information to the value-added service information.
  • Step 14 The service feature information of the extracted value-added service information is written into a preset universal dimension group to obtain a dimension group of the value-added service information.
  • the universal dimension group is a dimension space generated according to the service feature information included in the value-added service information group or the value-added service information creation.
  • the generic dimension group is expressed as follows:
  • the offer_ID is a value-added service information group identifier, and identifies a unique value-added service information group or a value-added service information creation.
  • the value-added service information creation refers to a collection of information including a title, description, and image link of a value-added service information, which is a mandatory option;
  • n is the general dimension group dimension.
  • other dimensions may be selected from user attributes, and/or value-added service information attributes, and/or information contained in the context attributes, and the number of selected dimensions may be set according to actual needs.
  • the universal dimension group includes all the service feature information in the value-added service information group or the value-added service information creation. For example, the following:
  • the juvenile's value can be set to 0, the youth value is 1, the middle-aged value is 2, and the old-aged value is 3, so that a common dimension of different values can be formed. group.
  • the dimension of the universal dimension group is 4, for example,
  • the dimension of the value-added service information Groups can be:
  • Step 15 Query a mapping relationship between the pre-stored charging type eCPM correction coefficient concentration dimension group and the eCPM correction coefficient according to the charging type of the value-added service information, and obtain an eCPM correction coefficient of the dimension group mapping of the value-added service information;
  • FIG. 2 is a schematic flowchart of obtaining a mapping relationship between a dimension group and an eCPM correction coefficient according to an embodiment of the present invention. Referring to Figure 2, the process includes:
  • Step 21 calculating an average estimated eCPM according to the pre-stored presentation history log
  • the average pre-estimation eCPM under each value of the common dimension group dims(n) is calculated from the display history log of each value-added service information group stored in advance.
  • the average estimated eCPM is calculated by using the following formula:
  • eCPM HISTORY (dims(n)) is the average estimated eCPM
  • C(dims(n)) is a collection of historical log records, and each history log contains a real-time estimated eCPM of a value-added service information behavior type;
  • Ac(i) is the behavior type of the i-th value-added service information, wherein the behavior type includes: presentation, click, and installation, the presentation identifier is imp, the click identifier is clk, and the installation identifier is ins;
  • eCPM PREDICT (i) is the estimated eCPM value of the i-th value-added service information.
  • each value-added service information corresponds to a common dimension group of values
  • the general-purpose dimension group includes value-added service information of multiple charging types.
  • the numerator is recorded in all value-added service information
  • the behavior type is the sum of the real-time estimated eCPM values of the value-added service information presented
  • the denominator is all the value-added service information recorded
  • the behavior type is the value-added service presented. The sum of the number of pieces of information.
  • the history log may be selected one day before or before.
  • the log of the day can also use the hour-level log, depending on the amount of value-added service information log for each value of dims(n).
  • the method may further include:
  • the historical true eCPM is calculated based on the pre-stored presentation history log.
  • eCPM REAL (dims(n)) is historical real eCPM
  • TOTAL_COST(dims(n)) is the total revenue corresponding to the pre-stored presentation history log
  • TOTAL_IMPRESSION(dims(n)) is the total amount of impressions corresponding to the pre-stored presentation history log.
  • the eCPM correction coefficient may be calculated according to the historical true eCPM in the following.
  • Step 22 Fix values of other dimensions except the value-added service information group identifier in the universal dimension group, traverse the value-added service information group identifier, and calculate an eCPM correction coefficient.
  • dim2 v2
  • dim3 v3
  • a presentation history log does not contain a feature of a certain dimension, it is set to its default value, that is, a default value is predefined for the feature corresponding to each dimension in the common dimension group.
  • the dimension group of the value-added service information is: (gender 1, country 1, material size 1, default age, default language, default device, ...); for the second value-added service information, the value-added service information
  • the dimension groups are: (gender 2, default country, default material size, age 2, language 2, device 2, ).
  • the corresponding eCPM correction coefficient may be different in different value-added service information request requests because different feature information included in different value-added service information delivery requests is different.
  • the eCPM correction factor is calculated using the following formula:
  • Delta (b, v 2 , ..., v n ) is the eCPM correction coefficient. Under different value-added service information delivery requests, the eCPM correction coefficient corresponding to the same value-added service information may be different;
  • b i is the i-th accounting type
  • b is the specified billing type
  • m is the type of billing type included in the dimension group.
  • the eCPM correction coefficient may also be calculated by using the following formula:
  • the delta value of the dims(n) various value combinations may be stored in an offline file or a database.
  • Step 16 Apply the eCPM correction coefficient to the preset eCPM correction formula to obtain a modified eCPM for each value-added service information, sort and update the sorted value-added service information according to the calculated corrected eCPM.
  • each dimension in the dimension group dims(n) corresponding to each value-added service information in the value-added service information push request is first parsed out.
  • the corrected eCPM is calculated by using the following formula:
  • 0 ⁇ ⁇ ⁇ 1, which is a constant. The larger the value, the stronger the correction strength. In practical applications, ⁇ is the same for all value-added service information and all value-added service information delivery requests.
  • the above formula is for the value-added service information with a lower historical eCPM. Increase its current estimated eCPM; otherwise it will lower its current estimated eCPM.
  • a plurality of value-added service information (the value-added service information of the top N bits before sorting) ranked first in the score may be selected and returned to the client/browser.
  • the method may further include:
  • the user behavior information of the pushed value-added service information is recorded in the history log.
  • the feature information included in the value-added service information delivery request is obtained by parsing the received value-added service information delivery request, and the candidate value-added service conforming to the feature information is selected from the preset value-added service information set.
  • An information set sequentially extracting service characteristic information of each value-added service information in the candidate value-added service information set, inputting the pre-trained value-added service information model, and calculating a charging coefficient for each value-added service information;
  • the charging coefficient corresponding to the charging type is used to calculate the expected charging per thousand times of value-added service information; and the pre-set general dimension group is written according to the service characteristic information of the extracted value-added service information to obtain the dimension group of the value-added service information;
  • the pre-stored charging type eCPM correction coefficient set, the mapping relationship between the dimension group and the eCPM correction coefficient, and the eCPM correction coefficient of the dimension group mapping of the value-added service information are obtained;
  • the eCPM is Correction factor applied to pre-set eCPM correction Type, each corrected eCPM value-added service information sorted according to the calculated correction value-added service information and pushes the eCPM sorted.
  • the deviation of the historical distribution of the estimated eCPM of the value-added service information of different charging types (eCPM correction coefficient) is calculated, and the current pre-dot on the line is used.
  • the eCPM value is adjusted to reduce the eCPM difference between the value-added service information of different charging types, so that the historical eCPM factor of the value-added service information is considered in the sorting, so that the value-added service information of different charging types can be more fair.
  • the sorting effect improves the push efficiency of the value-added service information; further, the historical eCPM correction coefficient is dynamically calculated according to the historical log, and the timeliness is better; and the set of universal dimension groups can support multiple dimensions, which can better adapt to different situations.
  • FIG. 3 is a schematic structural diagram of an apparatus for pushing value-added service information according to an embodiment of the present invention.
  • the device includes: a candidate value-added service information collection module 31, a charging coefficient calculation module 32, an eCPM calculation module 33, a dimension group acquisition module 34, an eCPM correction coefficient query module 35, and a value-added service information pushing module 36, where
  • the candidate value-added service information collection module 31 is configured to parse the received value-added service information delivery request, obtain the feature information included in the value-added service information delivery request, and select the qualified information from the preset value-added service information set.
  • the feature information includes, but is not limited to, one of a value-added service information group identifier, a user attribute, a value-added service information attribute, and a context attribute, or any combination thereof.
  • the charging coefficient calculation module 32 is configured to sequentially extract the service feature information of each value-added service information in the candidate value-added service information set, input the pre-trained value-added service information model, and calculate a charging coefficient for each value-added service information. ;
  • the service feature information includes, but is not limited to, one of a user attribute, a value-added service information attribute, and a context attribute, or any combination thereof.
  • the charging coefficient includes but is not limited to: CTR, CVR, and ICVR. among them,
  • the CTR is calculated; and for the value-added service information of the charging type CPA/CPI/CPS, the CTR and the CVR are calculated; or, the ICVR.
  • the eCPM calculation module 33 is configured to calculate, according to the charging coefficient corresponding to the charging type of the value-added service information, a charging eCPM displayed by the expected value-added service information per thousand times;
  • the expected charging per thousand times of value-added service information is calculated by using the following formula:
  • CTR is the click rate
  • Bid is the operator’s bid for thousand impressions.
  • CVR is the conversion rate
  • ICVR is the installation conversion rate.
  • the dimension group obtaining module 34 is configured to write the service feature information of the extracted value-added service information into a preset universal dimension group to obtain a dimension group of the value-added service information.
  • the universal dimension group is expressed as follows:
  • the offer_ID is a value-added service information group identifier, which is mandatory;
  • n is the general dimension group dimension.
  • other dimensions may be selected from user attributes, and/or value-added service information attributes, and/or information contained in the context attributes, and each included information corresponds to one dimension.
  • the eCPM correction coefficient querying module 35 is configured to query a mapping relationship between the pre-stored charging type eCPM correction coefficient concentration dimension group and the eCPM correction coefficient according to the charging type of the value-added service information, and obtain the dimension group mapping of the value-added service information.
  • eCPM correction factor ;
  • the eCPM correction coefficient query module 35 includes: an average estimated eCPM calculation unit, a mapping relationship construction unit, and an eCPM correction coefficient query unit (not shown), where
  • the average estimated eCPM calculation unit is configured to calculate an average estimated eCPM according to the pre-stored presentation history log;
  • a mapping relationship construction unit configured to fix values of other dimensions of the universal dimension group except the value-added service information group identifier, traverse the value-added service information group identifier, and calculate corresponding dimension groups corresponding to the value-added service information
  • the eCPM correction coefficient is placed in the charging type eCPM correction coefficient set;
  • the eCPM correction coefficient querying unit is configured to query the mapping relationship construction unit according to the charging type of the value-added service information, and acquire an eCPM correction coefficient of the dimension group mapping of the value-added service information.
  • the average estimated eCPM is calculated by using the following formula:
  • eCPM HISTORY (dims(n)) is the average estimated eCPM
  • C(dims(n)) is a collection of historical logs
  • Ac(i) is the behavior type of the i-th value-added service information, wherein the behavior type includes: presentation, click, and installation, the presentation identifier is imp, the click identifier is clk, and the installation identifier is ins;
  • eCPM PREDICT (i) is the estimated eCPM value of the i-th value-added service information.
  • the eCPM correction coefficient is calculated using the following formula:
  • Delta(b, v 2 ,..., v n ) is the eCPM correction coefficient
  • b i is the i-th accounting type
  • b is the specified billing type
  • m is the type of billing type included in the dimension group.
  • the eCPM correction coefficient query module 35 may further include:
  • a historical real eCPM calculation unit for calculating a historical true eCPM based on a pre-stored presentation history log.
  • the historical real eCPM is calculated by using the following formula:
  • eCPM REAL (dims(n)) is historical real eCPM
  • TOTAL_COST(dims(n)) is the total revenue corresponding to the pre-stored presentation history log
  • TOTAL_IMPRESSION(dims(n)) is the total amount of impressions corresponding to the pre-stored presentation history log.
  • the eCPM correction coefficient is calculated using the following formula:
  • the value-added service information pushing module 36 is configured to apply the eCPM correction coefficient to the preset eCPM correction formula to obtain a modified eCPM for each value-added service information, and sort and push the sorted value-added service information according to the calculated corrected eCPM.
  • the modified eCPM is calculated by using:
  • the apparatus further includes:
  • the recording module 37 is configured to record user behavior information of the pushed value-added service information into the history log.
  • An embodiment of the present invention further provides an electronic device, where the electronic device includes the device described in any of the foregoing embodiments.
  • the electronic device may include a housing 41, a processor 42, a memory 43, a circuit board 44, and a power supply circuit 45, wherein the circuit board 44 is disposed in the housing 41.
  • the processor 42 and the memory 43 are disposed on the circuit board 44; the power supply circuit 45 is configured to supply power to the respective circuits or devices of the electronic device; the memory 43 is used to store executable program code; and the processor 42 is configured to read The executable program code stored in the memory 43 is configured to execute a program corresponding to the executable program code for performing the method of pushing the value-added service information described in any of the foregoing embodiments.
  • FIG. 1-3 of the present invention The description of the embodiment shown in FIG. 1-3 of the present invention is omitted, and the details are not described herein.
  • the electronic device exists in a variety of forms including, but not limited to:
  • Mobile communication devices These devices are characterized by mobile communication functions and are mainly aimed at providing voice and data communication.
  • Such terminals include: smart phones (such as iPhone), multimedia phones, functional phones, and low-end phones.
  • Ultra-mobile personal computer equipment This type of equipment belongs to the category of personal computers, has computing and processing functions, and generally has mobile Internet access.
  • Such terminals include: PDAs, MIDs, and UMPC devices, such as the iPad.
  • Portable entertainment devices These devices can display and play multimedia content. Such devices include: audio, video players (such as iPod), handheld game consoles, e-books, and smart toys and portable car navigation devices.
  • the server consists of a processor, a hard disk, a memory, a system bus, etc.
  • the server is similar to a general-purpose computer architecture, but because of the need to provide highly reliable services, processing power and stability High reliability in terms of reliability, security, scalability, and manageability.
  • the machine can be read into a storage medium, and when executed, the program can include the flow of an embodiment of the methods as described above.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

Abstract

一种推送增值业务信息的方法、装置及电子设备。方法包括:依据增值业务信息投放请求获取候选增值业务信息集;利用每一增值业务信息的业务特征信息以及增值业务信息模型,计算对应的收费系数;依据所述收费系数计算eCPM;依据提取的业务特征信息,写入通用维度组,得到该增值业务信息的维度组;依据增值业务信息的计费类型,查询维度组与eCPM修正系数的映射关系,获取增值业务信息的维度组映射的eCPM修正系数;将eCPM修正系数应用于eCPM修正公式,得到每一增值业务信息的修正eCPM,并按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。方法可以提升增值业务信息的推送效率。

Description

一种推送增值业务信息的方法、装置及电子设备
相关申请的交叉引用
本申请要求北京金山安全软件有限公司于2016年3月18日提交的、发明名称为“一种推送增值业务信息的方法、装置及电子设备”的、中国专利申请号“201610157453.4”的优先权。
技术领域
本发明涉及增值业务应用技术,尤其涉及一种推送增值业务信息的方法、装置及电子设备。
背景技术
随着计算机通信以及互联网技术的不断发展,智能移动电话、个人数字助理、掌上电脑、笔记本电脑等电子设备得到了越来越广泛的应用。不少增值业务运营商看到了基于电子设备投放增值业务信息的商机,例如,增值业务运营商通过互联网络运营商提供的增值业务平台,向移动电话推送增值业务信息,具有投放成本低、投放面广、用户易于、增值业务投放效率较高等特点。
在互联网络运营商提供的增值业务平台上,增值业务平台可以基于多种增值业务计费类型,投放多种计费类型的增值业务信息,其中,计费类型包括但不限于:按照每千次展现进行收费的方式(CPM,Cost-Per-Mille)、按照每次点击进行收费的方式(CPC,Cost-Per-Click)、按照每次安装进行收费的方式(CPI,Cost-Per-Install)、按照每次行动进行收费的方式(CPA,Cost-Per-Action)、按照每次购买进行收费的方式(CPS,Cost-Per-Sale)、按照每次视频播放进行收费的方式(CPV,Cost-Per-Video)等。
由于同一增值业务平台一次会投放多种应用对应的增值业务信息,不同的增值业务信息的设计水平参差不齐,为了衡量不同计费类型的增值业务信息的 水平或质量,以确定哪些增值业务信息排在靠前的位置展示给用户,增值业务平台提出一种对待推送的各增值业务信息进行排序展示的方法,对不同计费类型的增值业务信息分别设置相应的扶植系数w(bid_type),其中w(bid_type)>0,在计算出每一计费类型的统一的期望的每千次增值业务信息展现的收费(eCPM,expected-Cost-Per-Mille)后,将eCPM与扶植系数相乘得到修正eCPM,并按修正eCPM的高低对增值业务信息进行排序,并按照排序的序列,将增值业务信息展示给用户。其中,不同计费类型的增值业务信息,使用不同的算法计算出eCPM。下面简要描述计算eCPM的流程:
步骤1,根据各增值业务信息的历史日志,训练出计算CTR/CVR/ICVR的相应增值业务信息模型;
步骤2,根据增值业务信息请求中的定向条件,例如,要求投放的国别、性别、年龄等定向条件,从增值业务信息集中,选出符合定向条件的候选增值业务信息集。
步骤3,对于候选增值业务信息集中的每一增值业务信息,提取该增值业务信息的用户属性(例如,性别、年龄、国家、语言、设备类型等)、增值业务信息属性(例如,素材尺寸、效果类、转化类、行业分类)、上下文属性(例如,媒体、增值业务信息位),生成模型增值业务特征信息;
步骤4,将生成的模型增值业务特征信息输入训练出的相应增值业务信息模型中,计算出该增值业务信息的CTR/CVR/ICVR;
步骤5,在计算出该增值业务信息的CTR/CVR/ICVR后,按照该增值业务信息的计费类型对应的eCPM计算公式,计算出eCPM,然后,使用eCPM与扶植系数的乘积对候选增值业务信息集中的各增值业务信息进行排序,向用户返回排在最前的若干增值业务信息。
在上述计算eCPM时,由于CTR、CVR、ICVR等指标通常是在不同的数据集上使用不同模型训练得到的,例如,使用逻辑回归(LR,Logistic Regression)模型、梯度提升决策树(GBDT,Gradient Boosting Decision Tree) 模型。在用于模型训练的数据集中,对于不同的计费类型的增值业务信息数据,正样本(购买/点击行为)数据与负样本(没有后续购买/点击的展现)数据分布严重不均匀(前者远小于后者),导致通过模型预估出的增值业务信息CTR/CVR/ICVR与历史真实的增值业务信息CTR/ICVR有较大偏差,使得计算得到的eCPM的分数分布差距较大,导致不同计费类型的增值业务信息混排后效果不佳,使得某种计费类型的“优质”增值业务信息,由于其eCPM低于另一种计费类型下的“低质”增值业务信息的eCPM,得不到机会展现,使得推送的增值业务信息效率不高;进一步地,扶植系数是预先确定的,使得当不同计费类型的增值业务信息之间eCPM分布发生较大变动时,扶植系数不能满足计算需要;而且,扶植系数的取值取决于计费类型,没有考虑其他因素,如用户因素,广告因素以及上下文因素等,导致计算的eCPM较低,推送的增值业务信息效率不高。
发明内容
有鉴于此,本发明实施例提供一种推送增值业务信息的方法、装置及电子设备,提升增值业务信息的推送效率。
为达到上述目的,本发明的实施例采用如下技术方案:
第一方面,本发明实施例提供一种推送增值业务信息的方法,包括:
解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到 该增值业务信息的维度组;
依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
可选的,获取所述维度组与eCPM修正系数的映射关系包括:
依据预先存储的展现历史日志,计算平均预估eCPM;
固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算eCPM修正系数。
可选的,利用下式计算所述平均预估eCPM:
Figure PCTCN2017075063-appb-000001
式中,
eCPMHISTORY(dims(n))为平均预估eCPM;
C(dims(n))为展现历史日志集合;
ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
可选的,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000002
式中,
delta(b,v2,...,vn)为eCPM修正系数;
bi为第i种计费类型;
b为指定的计费类型;
m为维度组中包含的计费类型种类。
可选的,所述方法还包括:
依据预先存储的展现历史日志,计算历史真实eCPM。
可选的,利用下式计算所述历史真实eCPM:
Figure PCTCN2017075063-appb-000003
式中,
eCPMREAL(dims(n))为历史真实eCPM;
TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
可选的,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000004
可选的,利用下式计算所述修正eCPM:
Figure PCTCN2017075063-appb-000005
0≤α≤1,为常数。
可选的,所述方法还包括:
将推送的增值业务信息的用户行为信息记录至所述历史日志中。
第二方面,本发明实施例提供一种推送增值业务信息的装置,包括:候选增值业务信息集获取模块、收费系数计算模块、eCPM计算模块、维度组获取模块、eCPM修正系数查询模块以及增值业务信息推送模块,其中,
候选增值业务信息集获取模块,用于解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
收费系数计算模块,用于依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
eCPM计算模块,用于依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
维度组获取模块,用于将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
eCPM修正系数查询模块,用于依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
增值业务信息推送模块,用于将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
可选的,所述eCPM修正系数查询模块包括:平均预估eCPM计算单元、映射关系构建单元以及eCPM修正系数查询单元,其中,
平均预估eCPM计算单元,用于依据预先存储的展现历史日志,计算平均预估eCPM;
映射关系构建单元,用于固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算相应的所述增值业务信息的维度组对应的eCPM修正系数,置于计费类型eCPM修正系数集中;
eCPM修正系数查询单元,用于依据增值业务信息的计费类型,查询所述映射关系构建单元,获取所述增值业务信息的维度组映射的eCPM修正系数。
可选的,利用下式计算所述平均预估eCPM:
Figure PCTCN2017075063-appb-000006
式中,
eCPMHISTORY(dims(n))为平均预估eCPM;
C(dims(n))为展现历史日志集合;
ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
可选的,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000007
式中,
delta(b,v2,...,vn)为eCPM修正系数;
bi为第i种计费类型;
b为指定的计费类型;
m为维度组中包含的计费类型种类。
可选的,所述所述eCPM修正系数查询模块还包括:
历史真实eCPM计算单元,用于依据预先存储的展现历史日志,计算历史真实eCPM。
可选的,利用下式计算所述历史真实eCPM:
Figure PCTCN2017075063-appb-000008
式中,
eCPMREAL(dims(n))为历史真实eCPM;
TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
可选的,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000009
可选的,利用下式计算所述修正eCPM:
Figure PCTCN2017075063-appb-000010
式中,
0≤α≤1,为常数。
可选的,所述装置还包括:
记录模块,用于将推送的增值业务信息的用户行为信息记录至所述历史日志中。
第三方面,本发明实施例提供一种电子设备,所述电子设备包括:壳体、处理器、存储器、电路板和电源电路,其中,电路板安置在壳体围成的空间内 部,处理器和存储器设置在电路板上;电源电路,用于为上述电子设备的各个电路或器件供电;存储器用于存储可执行程序代码;处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序,用于执行前述任一所述的推送增值业务信息的方法。
第四方面,本发明实施例提供一种计算机程序产品,当所述计算机程序产品中的指令处理器执行时,执行一种推送增值业务信息的方法,所述方法包括:
解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
第五方面,本发明实施例提供一种存储介质,当所述存储介质中的指令由服务器的处理器被执行时,使得服务器能够执行一种推送增值业务信息方法,所述方法包括:
解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包 含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
本发明实施例提供的推送增值业务信息的方法、装置及电子设备,解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;依次提取候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费;依据提取的增值业务信息的业务特征信息,写入预先设置的通用维度组,得到该增值业务信息的维度组;依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中,维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息 的修正eCPM,并按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。这样,根据增值业务信息的历史日志中记录的预估eCPM值,算出不同计费类型的增值业务信息的预估eCPM的eCPM修正系数,对当前预估eCPM值进行调整,可以有效减少不同计费类型的增值业务信息之间的eCPM差异,使得增值业务信息的总体预估eCPM更加准确,从而优化增值业务信息的整体排序效果,提升增值业务信息的推送效率。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其它的附图。
图1为本发明实施例推送增值业务信息的方法流程示意图;
图2为本发明实施例获取维度组与eCPM修正系数的映射关系的流程示意图;
图3为本发明实施例推送增值业务信息的装置结构示意图;
图4为本发明电子设备一个实施例的结构示意图。
具体实施方式
下面结合附图对本发明实施例进行详细描述。
应当明确,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其它实施例,都属于本发明保护的范围。
图1为本发明实施例推送增值业务信息的装置流程示意图。参见图1,该方法包括:
步骤11,解析接收的增值业务信息投放请求,获取所述增值业务信息投放 请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
本步骤中,作为一可选实施例,特征信息包含但不限于:增值业务信息组标识、用户属性、增值业务信息属性以及上下文属性中的一种或其任意组合,其中,
增值业务信息组又称增值业务信息系列,是指具有相同预算、计费类型、出价、投放时段和定向条件的一组增值业务信息。
用户属性包括但不限于:性别、年龄、国家、语言、设备类型等;
增值业务信息属性包括但不限于:素材尺寸、效果类、转化类、行业分类;
上下文属性包括但不限于:媒体、增值业务信息位。
本发明实施例中,用户或增值业务运营商可以通过客户端/浏览器发起增值业务信息投放请求,增值业务信息投放请求中包含有用户属性、上下文信息等特征信息。
本发明实施例中,作为一可选实施例,特征信息可以是包含用户属性的定向条件。
步骤12,依次提取候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
本步骤中,对于候选增值业务信息集中的每一增值业务信息,提取该增值业务信息的业务特征信息,包括:用户属性(例如,性别、年龄、国家、语言、设备类型等)、增值业务信息属性(例如,素材尺寸、效果类、转化类、行业分类)、上下文属性(例如,媒体、增值业务信息位)中的一种或其任意组合,生成模型增值业务特征信息,将生成的模型增值业务特征信息输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数。例如,对于计费类型为CPC的增值业务信息,则通过将增值业务信息的业务特征信息输入包含CPC增值业务信息的增值业务信息模型,计算出CTR;而对于计费类 型为CPA/CPI/CPS的增值业务信息,则通过将增值业务信息的业务特征信息输入包含CPA/CPI/CPS增值业务信息的增值业务信息模型,计算出CTR和CVR;或者,ICVR。
本发明实施例中,业务特征信息包含但不限于:用户属性、增值业务信息属性以及上下文属性中的一种或其任意组合。
本发明实施例中,收费系数包括但不限于:CTR、CVR以及ICVR。
步骤13,依据所述增值业务信息的计费类型对应的收费系数,计算期望的每千次增值业务信息展现的收费;
本步骤中,对于计费类型为CPC的增值业务信息,利用下式计算期望的每千次增值业务信息展现的收费:
eCPM=CTR*bid
式中,
eCPM为期望的每千次增值业务信息展现的收费;
CTR为点击率(Click-Through-Rate),即增值业务信息点击量与增值业务信息展现量的比值;
bid为运营商的每千次展示出价。
对于计费类型为CPA/CPI/CPS的增值业务信息,利用下式计算期望的每千次增值业务信息展现的收费:
eCPM=CTR*CVR*bid
式中,
CVR为转化率(Conversion-Rate),即增值业务信息产生的购买/安装量与增值业务信息点击量的比值。
或者,
eCPM=ICVR*bid
式中,
ICVR为安装转化率(Install-Conversion-Rate),即增值业务信息产生的购买/安装量与增值业务信息展现量的比值。
步骤14,将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
本步骤中,通用维度组为依据增值业务信息组或增值业务信息创意中包含的业务特征信息生成的维度空间。作为一可选实施例,通用维度组表示如下:
dims(n)=<offer_ID,dim2,dim3,...,dimn>
其中,
offer_ID为增值业务信息组标识,标识唯一的一增值业务信息组或增值业务信息创意,其中,增值业务信息创意是指包含一增值业务信息的标题、描述、图片链接的信息集合,为必选项;
dim2~dimn为其他维度;
n为通用维度组维数。
本发明实施例中,其他维度可以从用户属性、和/或,增值业务信息属性、和/或,上下文属性包含的信息中选取,选取的维度数可依据实际需要设置。
作为一可选实施例,通用维度组包含增值业务信息组或增值业务信息创意中所有的业务特征信息,例如,可以设置:
性别对应dim2,取值dim2=v2;年龄对应dim3,取值dim3=v3;国家对应dim4,取值dim4=v4;语言对应dim5,取值dim5=v5;设备类型对应dim6,取值dim6=v6等;素材尺寸对应dim7,取值dim7=v7等。其中,取值为变量,例如,依据性别的不同,可以设置男性的取值为1,女性的取值为0,也可以直接设置为性别,例如,设置dim2=v2=男。再例如,依据年龄段的不同,可以设置少年的取值为0,青年的取值为1,中年的取值为2,老年的取值为3等,从而可以形成不同取值的通用维度组。
作为另一可选实施例,如果增值业务信息组或增值业务信息创意中包含的所有业务特征信息的数量为4,则通用维度组的维数为4,例如,
dims(4)=<offer_ID=计费类型,dim2=性别,dim3=缺省,dim4=国家,dim5=缺省,dim6=缺省,dim7=素材尺寸>
则对于某一增值业务信息,如果缺省值设置为0,该增值业务信息的维度 组可以为:
dims(4)=<CPC广告,男,0,中国,0,0,大号尺寸>。
步骤15,依据所述增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
本发明实施例中,作为一可选实施例,图2为本发明实施例获取维度组与eCPM修正系数的映射关系的流程示意图。参见图2,该流程包括:
步骤21,依据预先存储的展现历史日志,计算平均预估eCPM;
本步骤中,从预先存储的各增值业务信息组的展现历史日志中,计算在通用维度组dims(n)的各取值下的平均预估计eCPM。
本发明实施例中,利用下式计算平均预估eCPM:
Figure PCTCN2017075063-appb-000011
式中,
eCPMHISTORY(dims(n))为平均预估eCPM;
C(dims(n))为展现历史日志集合,每条历史日志包含一次增值业务信息行为类型的实时预估eCPM;
ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
本发明实施例中,每一条增值业务信息对应一取值的通用维度组,在通用维度组中,包含有多种计费类型的增值业务信息。在上述计算公式中,分子为记录的所有增值业务信息中,行为类型为展现的增值业务信息的实时预估eCPM值之和,分母为记录的所有增值业务信息中,行为类型为展现的增值业务信息条数之和。
本发明实施例中,作为一可选实施例,历史日志可以选取前一天或之前多 天的日志,也可以使用小时级别的日志,取决于dims(n)的每一取值的增值业务信息日志量的多少。
本发明实施例中,作为一可选实施例,该方法还可以包括:
依据预先存储的展现历史日志,计算历史真实eCPM。
本步骤中,利用下式计算历史真实eCPM:
Figure PCTCN2017075063-appb-000012
式中,
eCPMREAL(dims(n))为历史真实eCPM;
TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
本发明实施例中,后续中可以依据历史真实eCPM计算eCPM修正系数。
步骤22,固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算eCPM修正系数。
本步骤中,固定dim2,…,dimn的取值,即令dim2=v2,dim3=v3,dimn=vn,。如果某一展现历史日志不包含某一维度的特征,则设置为其缺省值,即对于通用维度组中每一维度对应的特征,预先定义一缺省值。例如,如果有两条增值业务信息,第一增值业务信息仅包含性别1、国家1和素材尺寸1,第二增值业务信息仅包含性别2、年龄2、语言2和设备类型2,对于第一增值业务信息,该增值业务信息的维度组为:(性别1,国家1,素材尺寸1,缺省年龄,缺省语言,缺省设备,…);对于第二增值业务信息,该增值业务信息的维度组为:(性别2,缺省国家,缺省素材尺寸,年龄2,语言2,设备2,…)。
本发明实施例中,由于不同增值业务信息投放请求中包含的特征信息不同,同一增值业务信息在不同的增值业务信息投放请求中,对应的eCPM修正系数也可以不同。
对于每一计费类型的取值b,利用下式计算eCPM修正系数:
Figure PCTCN2017075063-appb-000013
式中,
delta(b,v2,...,vn)为eCPM修正系数,不同的增值业务信息投放请求下,同一增值业务信息对应的eCPM修正系数可以不同;
bi为第i种计费类型;
b为指定的计费类型;
m为维度组中包含的计费类型种类。
本发明实施例中,作为另一可选实施例,还可以利用下式计算eCPM修正系数:
Figure PCTCN2017075063-appb-000014
本发明实施例中,作为一可选实施例,可将dims(n)各种取值组合下的delta值存在离线文件或数据库中。
步骤16,将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
本步骤中,进行线上打分时,对于一次增值业务信息推送请求以及对应的候选增值业务信息,先解析出该增值业务信息推送请求中对应各增值业务信息的维度组dims(n)中各个维度的值bid_type=b,dim2=v2,dim3=v3,dimn=vn,并计算出eCPM。然后,查询计费类型eCPM修正系数集,找到delta(b,v2,...,vn)。
本发明实施例中,利用下式计算修正eCPM:
Figure PCTCN2017075063-appb-000015
0≤α≤1,为常数,数值越大,表示修正力度越强;实际应用中,对于所有增值业务信息以及所有增值业务信息投放请求,α均相同。
本发明实施例中,上述公式对于历史预估eCPM较低的增值业务信息,会 提升其当前预估eCPM;反之则会降低其当前预估eCPM。
本发明实施例中,可以选出分数排在最前面的若干增值业务信息(排序前N位的增值业务信息)并返回给客户端/浏览器。
本发明实施例中,作为一可选实施例,该方法还可以包括:
将推送的增值业务信息的用户行为信息记录至所述历史日志中。
本步骤中,通过将上述流程中产生的有用的用户行为信息,例如,增值业务信息的展现,用户点击增值业务信息,用户安装增值业务信息记录到历史日志中,从而可以不断扩充增值业务信息模型。
本发明实施例中,通过解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;依次提取候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费;依据提取的增值业务信息的业务特征信息,写入预先设置的通用维度组,得到该增值业务信息的维度组;依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中,维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,并按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。这样,根据增值业务信息的历史日志中记录的预估eCPM值,算出不同计费类型的增值业务信息的预估eCPM的历史分布的偏差(eCPM修正系数),使用该偏差对线上的当前预估eCPM值进行调整,以减少不同计费类型的增值业务信息之间的eCPM差异,从而在排序时,考虑增值业务信息的历史eCPM因素,使得不同计费类型的增值业务信息之间能更加公平的竞争,提升用于排序的eCPM的精确度,使得增值业务信息的总体预估eCPM更加准确,提升eCPM的精确度,从而优化增值业务信息的整体 排序效果,提升增值业务信息的推送效率;进一步地,根据历史日志,动态计算历史eCPM修正系数,时效性较好;而且,设置的通用维度组可以支持多个维度,能更好的适应不同情境下的增值业务信息eCPM分布。
图3为本发明实施例推送增值业务信息的装置结构示意图。参见图3,该装置包括:候选增值业务信息集获取模块31、收费系数计算模块32、eCPM计算模块33、维度组获取模块34、eCPM修正系数查询模块35以及增值业务信息推送模块36,其中,
候选增值业务信息集获取模块31,用于解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
本发明实施例中,作为一可选实施例,特征信息包含但不限于:增值业务信息组标识、用户属性、增值业务信息属性以及上下文属性中的一种或其任意组合。
收费系数计算模块32,用于依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
本发明实施例中,作为一可选实施例,业务特征信息包含但不限于:用户属性、增值业务信息属性以及上下文属性中的一种或其任意组合。
本发明实施例中,收费系数包括但不限于:CTR、CVR以及ICVR。其中,
对于计费类型为CPC的增值业务信息,计算CTR;而对于计费类型为CPA/CPI/CPS的增值业务信息,计算CTR和CVR;或者,ICVR。
eCPM计算模块33,用于依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
本发明实施例中,对于计费类型为CPC的增值业务信息,利用下式计算期望的每千次增值业务信息展现的收费:
eCPM=CTR*bid
式中,
eCPM为期望的每千次增值业务信息展现的收费;
CTR为点击率;
bid为运营商的每千次展示出价。
对于计费类型为CPA/CPI/CPS的增值业务信息,利用下式计算期望的每千次增值业务信息展现的收费:
eCPM=CTR*CVR*bid
式中,
CVR为转化率。
或者,
eCPM=ICVR*bid
式中,
ICVR为安装转化率。
维度组获取模块34,用于将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
本发明实施例中,作为一可选实施例,通用维度组表示如下:
dims(n)=<offer_ID,dim2,dim3,...,dimn>
其中,
offer_ID为增值业务信息组标识,为必选项;
dim2~dimn为其他维度;
n为通用维度组维数。
本发明实施例中,其他维度可以从用户属性、和/或,增值业务信息属性、和/或,上下文属性包含的信息中选取,每一包含的信息对应一维度。
eCPM修正系数查询模块35,用于依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
本发明实施例中,作为一可选实施例,eCPM修正系数查询模块35包括:平均预估eCPM计算单元、映射关系构建单元以及eCPM修正系数查询单元(图中未示出),其中,
平均预估eCPM计算单元,用于依据预先存储的展现历史日志,计算平均预估eCPM;
映射关系构建单元,用于固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算相应的所述增值业务信息的维度组对应的eCPM修正系数,置于计费类型eCPM修正系数集中;
eCPM修正系数查询单元,用于依据增值业务信息的计费类型,查询所述映射关系构建单元,获取所述增值业务信息的维度组映射的eCPM修正系数。
本发明实施例中,作为一可选实施例,利用下式计算所述平均预估eCPM:
Figure PCTCN2017075063-appb-000016
式中,
eCPMHISTORY(dims(n))为平均预估eCPM;
C(dims(n))为展现历史日志集合;
ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
相应地,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000017
式中,
delta(b,v2,...,vn)为eCPM修正系数;
bi为第i种计费类型;
b为指定的计费类型;
m为维度组中包含的计费类型种类。
作为另一可选实施例,eCPM修正系数查询模块35还可以包括:
历史真实eCPM计算单元,用于依据预先存储的展现历史日志,计算历史真实eCPM。
本发明实施例中,作为一可选实施例,利用下式计算所述历史真实eCPM:
Figure PCTCN2017075063-appb-000018
式中,
eCPMREAL(dims(n))为历史真实eCPM;
TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
相应地,利用下式计算所述eCPM修正系数:
Figure PCTCN2017075063-appb-000019
增值业务信息推送模块36,用于将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
本发明实施例中,作为一可选实施例,利用下式计算所述修正eCPM:
Figure PCTCN2017075063-appb-000020
0≤α≤1,为常数。
本发明实施例中,作为一可选实施例,该装置还包括:
记录模块37,用于将推送的增值业务信息的用户行为信息记录至所述历史日志中。
本发明实施例还提供一种电子设备,所述电子设备包含前述任一实施例所述的装置。
图4为本发明电子设备一个实施例的结构示意图,可以实现本发明图1-3 所示实施例的流程,如图4所示,上述电子设备可以包括:壳体41、处理器42、存储器43、电路板44和电源电路45,其中,电路板44安置在壳体41围成的空间内部,处理器42和存储器43设置在电路板44上;电源电路45,用于为上述电子设备的各个电路或器件供电;存储器43用于存储可执行程序代码;处理器42通过读取存储器43中存储的可执行程序代码来运行与可执行程序代码对应的程序,用于执行前述任一实施例所述的推送增值业务信息的方法。
处理器42对上述步骤的具体执行过程以及处理器42通过运行可执行程序代码来进一步执行的步骤,可以参见本发明图1-3所示实施例的描述,在此不再赘述。
该电子设备以多种形式存在,包括但不限于:
(1)移动通信设备:这类设备的特点是具备移动通信功能,并且以提供话音、数据通信为主要目标。这类终端包括:智能手机(例如iPhone)、多媒体手机、功能性手机,以及低端手机等。
(2)超移动个人计算机设备:这类设备属于个人计算机的范畴,有计算和处理功能,一般也具备移动上网特性。这类终端包括:PDA、MID和UMPC设备等,例如iPad。
(3)便携式娱乐设备:这类设备可以显示和播放多媒体内容。该类设备包括:音频、视频播放器(例如iPod),掌上游戏机,电子书,以及智能玩具和便携式车载导航设备。
(4)服务器:提供计算服务的设备,服务器的构成包括处理器、硬盘、内存、系统总线等,服务器和通用的计算机架构类似,但是由于需要提供高可靠的服务,因此在处理能力、稳定性、可靠性、安全性、可扩展性、可管理性等方面要求较高。
(5)其他具有数据交互功能的电子设备。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一计算 机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。其中,所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM)或随机存储记忆体(Random Access Memory,RAM)等。
以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,可轻易想到的变化或替换,都应涵盖在本发明的保护范围之内。因此,本发明的保护范围应以权利要求的保护范围为准。

Claims (21)

  1. 一种推送增值业务信息的方法,其特征在于,包括:
    解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
    依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
    依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
    将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
    依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
    将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
  2. 根据权利要求1所述的方法,其特征在于,获取所述维度组与eCPM修正系数的映射关系包括:
    依据预先存储的展现历史日志,计算平均预估eCPM;
    固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算eCPM修正系数。
  3. 根据权利要求2所述的方法,其特征在于,利用下式计算所述平均预估eCPM:
    Figure PCTCN2017075063-appb-100001
    式中,
    eCPMHISTORY(dims(n))为平均预估eCPM;
    C(dims(n))为展现历史日志集合;
    ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
    eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
  4. 根据权利要求3所述的方法,其特征在于,利用下式计算所述eCPM修正系数:
    Figure PCTCN2017075063-appb-100002
    式中,
    delta(b,v2,...,vn)为eCPM修正系数;
    bi为第i种计费类型;
    b为指定的计费类型;
    m为维度组中包含的计费类型种类。
  5. 根据权利要求2至4任一项所述的方法,其特征在于,所述方法还包括:
    依据预先存储的展现历史日志,计算历史真实eCPM。
  6. 根据权利要求5所述的方法,其特征在于,利用下式计算所述历史真实eCPM:
    Figure PCTCN2017075063-appb-100003
    式中,
    eCPMREAL(dims(n))为历史真实eCPM;
    TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
    TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
  7. 根据权利要求6所述的方法,其特征在于,利用下式计算所述eCPM修正系数:
    Figure PCTCN2017075063-appb-100004
  8. 根据权利要求1至7任一项所述的方法,其特征在于,利用下式计算所述修正eCPM:
    Figure PCTCN2017075063-appb-100005
    0≤α≤1,为常数。
  9. 根据权利要求1至7任一项所述的方法,其特征在于,所述方法还包括:
    将推送的增值业务信息的用户行为信息记录至所述历史日志中。
  10. 一种推送增值业务信息的装置,其特征在于,该装置包括:候选增值业务信息集获取模块、收费系数计算模块、eCPM计算模块、维度组获取模块、eCPM修正系数查询模块以及增值业务信息推送模块,其中,
    候选增值业务信息集获取模块,用于解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
    收费系数计算模块,用于依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
    eCPM计算模块,用于依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
    维度组获取模块,用于将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
    eCPM修正系数查询模块,用于依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获 取所述增值业务信息的维度组映射的eCPM修正系数;
    增值业务信息推送模块,用于将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
  11. 根据权利要求10所述的装置,其特征在于,所述eCPM修正系数查询模块包括:平均预估eCPM计算单元、映射关系构建单元以及eCPM修正系数查询单元,其中,
    平均预估eCPM计算单元,用于依据预先存储的展现历史日志,计算平均预估eCPM;
    映射关系构建单元,用于固定所述通用维度组中除增值业务信息组标识外的其他维度的取值,遍历所述增值业务信息组标识,计算相应的所述增值业务信息的维度组对应的eCPM修正系数,置于计费类型eCPM修正系数集中;
    eCPM修正系数查询单元,用于依据增值业务信息的计费类型,查询所述映射关系构建单元,获取所述增值业务信息的维度组映射的eCPM修正系数。
  12. 根据权利要求11所述的装置,其特征在于,利用下式计算所述平均预估eCPM:
    Figure PCTCN2017075063-appb-100006
    式中,
    eCPMHISTORY(dims(n))为平均预估eCPM;
    C(dims(n))为展现历史日志集合;
    ac(i)为第i条增值业务信息的行为类型,其中,行为类型包括:展现、点击以及安装,展现标识为imp,点击标识为clk,安装标识为ins;
    eCPMPREDICT(i)为第i条增值业务信息的预估eCPM值。
  13. 根据权利要求12所述的装置,其特征在于,利用下式计算所述eCPM修正系数:
    Figure PCTCN2017075063-appb-100007
    式中,
    delta(b,v2,...,vn)为eCPM修正系数;
    bi为第i种计费类型;
    b为指定的计费类型;
    m为维度组中包含的计费类型种类。
  14. 根据权利要求11至13任一项所述的装置,其特征在于,所述eCPM修正系数查询模块还包括:
    历史真实eCPM计算单元,用于依据预先存储的展现历史日志,计算历史真实eCPM。
  15. 根据权利要求14所述的装置,其特征在于,利用下式计算所述历史真实eCPM:
    Figure PCTCN2017075063-appb-100008
    式中,
    eCPMREAL(dims(n))为历史真实eCPM;
    TOTAL_COST(dims(n))为预先存储的展现历史日志对应的总收入;
    TOTAL_IMPRESSION(dims(n))为预先存储的展现历史日志对应的总展示量。
  16. 根据权利要求15所述的装置,其特征在于,利用下式计算所述eCPM修正系数:
    Figure PCTCN2017075063-appb-100009
  17. 根据权利要求10至16任一项所述的装置,其特征在于,利用下式计算所述修正eCPM:
    Figure PCTCN2017075063-appb-100010
    式中,
    0≤α≤1,为常数。
  18. 根据权利要求10至16任一项所述的装置,其特征在于,所述装置还包括:
    记录模块,用于将推送的增值业务信息的用户行为信息记录至所述历史日志中。
  19. 一种电子设备,其特征在于,所述电子设备包括:壳体、处理器、存储器、电路板和电源电路,其中,电路板安置在壳体围成的空间内部,处理器和存储器设置在电路板上;电源电路,用于为上述电子设备的各个电路或器件供电;存储器用于存储可执行程序代码;处理器通过读取存储器中存储的可执行程序代码来运行与可执行程序代码对应的程序,用于执行前述任一权利要求1-9所述的推送增值业务信息的方法。
  20. 一种计算机程序产品,其特征在于,当所述计算机程序产品中的指令处理器执行时,执行一种增值业务消耗预算方法,所述方法包括:
    解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
    依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
    依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
    将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
    依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映 射的eCPM修正系数;
    将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
  21. 一种存储介质,其特征在于,当所述存储介质中的指令由服务器的处理器被执行时,使得服务器能够执行一种增值业务消耗预算方法,所述方法包括:
    解析接收的增值业务信息投放请求,获取所述增值业务信息投放请求中包含的特征信息,从预先设置的增值业务信息集中,选出符合所述特征信息的候选增值业务信息集;
    依次提取所述候选增值业务信息集中的每一增值业务信息的业务特征信息,输入预先训练出的增值业务信息模型中,计算出针对每一增值业务信息的收费系数;
    依据所述增值业务信息的计费类型对应的收费系数计算期望的每千次增值业务信息展现的收费eCPM;
    将提取的增值业务信息的业务特征信息写入预先设置的通用维度组,得到该增值业务信息的维度组;
    依据增值业务信息的计费类型,查询预先存储的计费类型eCPM修正系数集中维度组与eCPM修正系数的映射关系,获取所述增值业务信息的维度组映射的eCPM修正系数;
    将所述eCPM修正系数应用于预先设置的eCPM修正公式,得到每一增值业务信息的修正eCPM,按照计算得到的修正eCPM进行排序并推送排序的增值业务信息。
PCT/CN2017/075063 2016-03-18 2017-02-27 一种推送增值业务信息的方法、装置及电子设备 WO2017157159A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610157453.4 2016-03-18
CN201610157453.4A CN107203892B (zh) 2016-03-18 2016-03-18 一种推送增值业务信息的方法、装置及电子设备

Publications (1)

Publication Number Publication Date
WO2017157159A1 true WO2017157159A1 (zh) 2017-09-21

Family

ID=59850426

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/075063 WO2017157159A1 (zh) 2016-03-18 2017-02-27 一种推送增值业务信息的方法、装置及电子设备

Country Status (2)

Country Link
CN (1) CN107203892B (zh)
WO (1) WO2017157159A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110990677A (zh) * 2019-12-13 2020-04-10 北京搜狐新媒体信息技术有限公司 信息存储量的询量方法及装置
CN111865933A (zh) * 2020-06-30 2020-10-30 上海瀚之友信息技术服务有限公司 一种增值服务管理系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101689273A (zh) * 2007-05-04 2010-03-31 谷歌公司 用于在线广告的量度转换
US20130262218A1 (en) * 2012-03-30 2013-10-03 Google Inc. Incorporating Delayed Feedback In Performance-Based Content Distribution
CN103914468A (zh) * 2012-12-31 2014-07-09 阿里巴巴集团控股有限公司 一种投放信息搜索的方法和装置
CN105184586A (zh) * 2015-09-18 2015-12-23 北京金山安全软件有限公司 一种控制信息展现的方法及装置

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103379161B (zh) * 2012-04-25 2018-05-22 腾讯科技(北京)有限公司 媒体信息展示方法、系统和装置
WO2015051025A1 (en) * 2013-10-01 2015-04-09 Vantage Media LLC Engagement mode marketplace with enhanced bidding and adjusted pricing based on engagement performance
US11017425B2 (en) * 2013-10-02 2021-05-25 Apple Inc. Optimization of promotional content campaigns
CN104636940A (zh) * 2013-11-07 2015-05-20 深圳市腾讯计算机系统有限公司 广告投放方法及系统、广告筛选方法及装置、服务器

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101689273A (zh) * 2007-05-04 2010-03-31 谷歌公司 用于在线广告的量度转换
US20130262218A1 (en) * 2012-03-30 2013-10-03 Google Inc. Incorporating Delayed Feedback In Performance-Based Content Distribution
CN103914468A (zh) * 2012-12-31 2014-07-09 阿里巴巴集团控股有限公司 一种投放信息搜索的方法和装置
CN105184586A (zh) * 2015-09-18 2015-12-23 北京金山安全软件有限公司 一种控制信息展现的方法及装置

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110990677A (zh) * 2019-12-13 2020-04-10 北京搜狐新媒体信息技术有限公司 信息存储量的询量方法及装置
CN110990677B (zh) * 2019-12-13 2023-03-31 北京搜狐新媒体信息技术有限公司 信息存储量的询量方法及装置
CN111865933A (zh) * 2020-06-30 2020-10-30 上海瀚之友信息技术服务有限公司 一种增值服务管理系统
CN111865933B (zh) * 2020-06-30 2022-02-22 上海瀚之友信息技术服务有限公司 一种增值服务管理系统

Also Published As

Publication number Publication date
CN107203892B (zh) 2020-07-31
CN107203892A (zh) 2017-09-26

Similar Documents

Publication Publication Date Title
US10445768B1 (en) User-based ad ranking
TWI648642B (zh) Data search processing method and system
TWI512508B (zh) Recommended methods and systems for recommending information
TWI603273B (zh) Method and device for placing information search
US8099412B2 (en) Query statistics provider
CN109409928A (zh) 一种素材推荐方法、装置、存储介质、终端
TW201520936A (zh) 使用者參與度上下文相關之非保證交付自動標價
CN104424210A (zh) 一种信息推荐方法、系统及服务器
US20150348097A1 (en) Autocreated campaigns for hashtag keywords
US20120246178A1 (en) Evaluating supply of electronic content relating to keywords
US8725566B2 (en) Predicting advertiser keyword performance indicator values based on established performance indicator values
WO2017157159A1 (zh) 一种推送增值业务信息的方法、装置及电子设备
CN106445965A (zh) 信息推广处理方法及装置
WO2017157130A1 (zh) 推送增值业务信息的方法、装置及电子设备
US11367110B2 (en) Personalized mobile application re-engagement
US20140172587A1 (en) Dynamic floor prices in second-price auctions
US8620925B1 (en) System and method for identifying advertising opportunities
CN104331823B (zh) 确定发布信息中关键词保留价格的方法及装置
US10497031B1 (en) Conditional bids in an auction
US20120059714A1 (en) Bundling User Lists with Ad Impressions
EP3743873B1 (en) Modeling distribution of digital components over a network
US20150332192A1 (en) Estimating Future Inventory
US20140214555A1 (en) Externalities in an auction
US20170085672A1 (en) Commercial-Interest-Weighted User Profiles
US10049386B1 (en) Adjusting content selection based on search results

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

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

Ref document number: 17765701

Country of ref document: EP

Kind code of ref document: A1

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 03/01/2019)

122 Ep: pct application non-entry in european phase

Ref document number: 17765701

Country of ref document: EP

Kind code of ref document: A1