CN114565408B - Bidding prediction method and system for advertisement putting - Google Patents
Bidding prediction method and system for advertisement putting Download PDFInfo
- Publication number
- CN114565408B CN114565408B CN202210193539.8A CN202210193539A CN114565408B CN 114565408 B CN114565408 B CN 114565408B CN 202210193539 A CN202210193539 A CN 202210193539A CN 114565408 B CN114565408 B CN 114565408B
- Authority
- CN
- China
- Prior art keywords
- program
- bidding
- exposure
- advertisement
- public opinion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0273—Determination of fees for advertising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a bidding prediction method and a bidding prediction system for advertisement putting, which comprise the following steps: s1, an exposure measurement model is constructed by an advertisement delivery party, and potential delivery income of the advertisement delivery party and a lower limit interval of bid pricing of an advertisement bid inviting party are calculated according to the exposure obtained by measurement; s2, constructing a public opinion risk measurement model by the advertisement delivery party, and calculating potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measurement and calculation; and S3, calculating a lower limit interval of the advertisement putting party putting income based on the potential putting income and the potential putting loss, and constructing a decision model taking the success rate and the putting income as double high optimization targets based on the lower limit interval of the putting income and the lower limit interval of the bid pricing of the advertisement tenderer. The invention realizes automatic bidding prediction in the process of advertisement bidding, effectively improves the precision and efficiency of bidding prediction, and simultaneously ensures the bidding success rate and the release income.
Description
Technical Field
The invention relates to the technical field of advertisement bidding, in particular to a bidding prediction method and a bidding prediction system for advertisement putting.
Background
With the rapid development of internet applications, advertising on the internet is becoming a mainstream way. The method for distributing advertisements via the internet has the advantages of wide coverage, strong initiative and the like, so that the method for distributing advertisements via the internet is more and more favored by various merchants, and thus, a traffic type platform for providing contents for an intelligent terminal is gradually developed, and when a user terminal requests to acquire the platform contents, advertisement delivery or pushing to the user terminal becomes one of the main profitable means of the platform.
In the existing application of internet advertisement, an advertiser obtains the advertisement flow provided by a platform in a bidding manner, that is, obtains an opportunity for advertisement display on an intelligent terminal used by a platform user through bidding. The bidding mode is that who bids more, corresponding advertisement flow can be obtained. For example, a browser APP that is popular and used by users may provide an ad slot, such as the top of the home page, through which advertisers want to bid for an ad. The current popular internet advertisement delivery mode is mainly the delivery form of real-time bidding advertisement.
However, the existing bidding methods manually propose the price for purchasing the advertisement space, and the proposed purchase price is higher, that is, the bid price is higher, in order to win the advertisement space. However, after a huge advertising fee is invested, the ultimate profit after the winning advertisement exhibition opportunity is converted into the actual purchasing behavior of the commodity is determined by manual experience, and the accuracy and the efficiency are limited.
Disclosure of Invention
The invention aims to provide a bidding prediction method for advertisement putting, which aims to solve the technical problems that in the prior art, the final profit after the advertisement display opportunity is converted into the actual purchasing behavior of the commodity is determined by manual experience in the bidding process, and the accuracy and the efficiency are limited.
In order to solve the technical problems, the invention specifically provides the following technical scheme:
a bid prediction method for advertisement placement, comprising the steps of:
s1, an advertisement delivery party constructs an exposure measurement model for measuring and calculating the exposure of a bid inviting program of the advertisement delivery party, and calculates the potential delivery income of the advertisement delivery party and a lower limit interval of bid inviting pricing of the advertisement delivery party according to the measured and calculated exposure;
s2, the advertisement delivery party constructs a public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program, and calculates the potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measuring and calculating;
and S3, calculating a lower limit interval of the advertisement putting income of the advertisement putting party based on the potential putting income and the potential putting loss, and constructing a decision model taking the success rate and the putting income as double high optimization targets based on the lower limit interval of the putting income and the lower limit interval of the bid pricing of the advertisement putting party so as to obtain a bid prediction result of the bid program.
As a preferred aspect of the present invention, the constructing an exposure measurement model by the advertisement delivery party includes:
acquiring program attribute characteristics and program exposure of historical programs at an advertising tenderer, and quantizing each historical program into a single longitudinal sample in sequence, wherein the program attribute characteristics are subjected to semantic vectorization to obtain a program attribute characteristic vector as a longitudinal sample characteristic, and the program exposure is subjected to data vectorization to obtain a program exposure vector as a longitudinal sample label;
taking the longitudinal sample characteristics as the input of a BP neural network, taking a longitudinal sample label as the output of the BP neural network, and carrying out model training on the BP neural network by using all longitudinal samples to obtain an exposure measurement model;
preferably, the longitudinal sample balance adjustment is performed before model training is performed on the BP neural network by all longitudinal samples to improve the training efficiency and the training precision of the exposure measurement model under the condition of limited number of longitudinal samples, and the method comprises the following steps:
extracting program attribute features of the bidding program, and performing semantic vectorization on the program attributes of the bidding program to obtain a program attribute feature vector of the bidding program;
sequentially calculating the feature similarity of each longitudinal sample and the bidding program, wherein the feature similarity is measured by using Euclidean distances between the features of the longitudinal samples and the feature vectors of the program attributes of the bidding program, and the calculation formula of the feature similarity is as follows:
in the formula, p i Characterizing the similarity of the ith vertical sample to the features of the bidding document, X i The characteristic is the longitudinal sample characteristic of the ith longitudinal sample, Y is the program attribute characteristic vector of the bidding program, i is a metering constant and has no substantial meaning;
setting an equalization adjustment threshold, wherein,
if the feature similarity p i If the value is larger than or equal to the balance adjustment threshold value, performing vertical sample retention on the vertical sample i to improve the vertical sample concentration with the attribute characteristics similar to those of the bidding program;
if the feature similarity p i If the sample number is less than the balance adjustment threshold, longitudinal sample elimination is carried out on the longitudinal sample i so as to reduce the concentration of the longitudinal sample with different attribute characteristics from the bidding program;
the vertical sample concentration is characterized by the proportion of vertical samples with different attribute characteristics with the bidding program or vertical samples with similar attribute characteristics with the bidding program in all the vertical samples, and the aspect of the balance adjustment of the vertical samples comprises the increase or decrease of the vertical sample concentration.
As a preferred aspect of the present invention, the exposure measurement model measures the exposure of a program of a bidding program, and includes:
inputting the program attribute feature vector of the bidding program into the exposure measurement model to obtain the program exposure vector of the bidding program, and converting the program exposure vector of the bidding program into the program exposure.
As a preferred embodiment of the present invention, the calculating the potential revenue of the advertisement delivery party and the lower limit interval of bid pricing of the advertisement bid inviting party according to the advertisement exposure amount obtained by the measurement and calculation includes:
the calculation formula of the potential release income is as follows:
S＝W 1 a 1 Z+W 2 a 2 Z；
in the formula, S is characterized as potential release yield, a 1 、a 2 Respectively characterized by the conversion rate of the program exposure and the income of the entity commodity, the conversion rate of the program exposure and the income of the stock price, W 1 、W 2 Respectively representing the potential influence weight of the program exposure on the entity commodity income, the potential influence weight of the program exposure and the share price income, and Z representing the program exposure;
commodity conversion and stock price conversion;
the calculation formula of the lower limit interval of the bid pricing is as follows:
T＝bZ；
in the formula, T is characterized as a lower limit interval of bid pricing, and b is characterized as an interval of a program exposure pricing coefficient;
preferably, the potential impact weight is determined by a financial report revenue structure of the advertising sponsor, wherein,
in the formula, Q 1 、Q 2 Respectively representing the entity commodity income and the stock price income in the financial newspaper;
the method for determining the upper limit and the lower limit of the interval in the interval of the program exposure pricing coefficient comprises the following steps:
extracting the program exposure and the bid price in the advertisement of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the bid price in the advertisement and the bid price time sequence expansion coefficient by the product of the program exposure and the exposure time sequence expansion coefficient to obtain a program exposure pricing coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity;
selecting all historical programs of all advertising tenders as transverse samples, calculating the feature similarity of all transverse samples and the tendered programs, selecting the program exposure of the historical programs corresponding to the transverse sample with the maximum feature similarity and the bidding price in the advertisements, and dividing the product of the bidding price in the advertisements and the bidding price time sequence expansion coefficient by the program exposure and the exposure time sequence expansion coefficient to obtain the program pricing exposure coefficient of the historical programs corresponding to the transverse sample with the maximum feature similarity;
respectively taking the maximum value and the minimum value in the program exposure pricing coefficient of the historical program corresponding to the longitudinal sample and the program exposure pricing coefficient of the historical program corresponding to the transverse sample as an upper interval limit and a lower interval limit;
the price marking time sequence expansion coefficient is the currency expansion rate of the time sequence of the bidding program and the time sequence of the historical program, and the exposure time sequence expansion coefficient is the ratio of audience to people of the time sequence of the bidding program to the time sequence of the historical program.
As a preferred scheme of the present invention, the constructing a public opinion risk measurement model by the advertisement delivery party includes:
extracting historical public sentiment events from a participation main body related to a crisis public affair event in the Internet;
extracting public opinion risk factors based on a historical public opinion event and a crisis public affair workflow rule base to construct an air risk factor set, constructing an expert group, expressing the influence level of the risk factor set by using a phrase set by an expert, and collecting expression data of the influence level to form a risk factor influence level data set;
the method comprises the steps that a three-layer structure of an initial public opinion risk measurement model is built by utilizing an analytic hierarchy process, and each level node is defined, wherein the public opinion risk measurement model consists of a target layer, a criterion layer and an index layer, wherein the target layer determines that a main body of evaluation is public opinion risk, the criterion layer defines four standard criteria of a participation main body, event attributes, public opinion state and public affair cost based on fishbone map analysis of historical public opinion events, and the index layer consists of risk factors and influence levels corresponding to the risk factors;
constructing an evaluation matrix based on the risk factor influence level data set, and calculating by using an extended good and bad solution distance method to synthesize a multi-dimensional evaluation matrix to obtain the influence levels of each risk factor;
determining the arrangement order of each layer element in the initial public opinion risk measurement model by using an analytic hierarchy process based on the influence levels of each risk factor to obtain an optimal public opinion risk measurement model with a complete system structure;
public sentiment is carried out by taking a participating subject of a bidding program as a keyword in the Internet.
As a preferable aspect of the present invention, the public opinion risk calculating model for calculating public opinion risk of a bidding program includes:
collecting public opinion keywords of each participating subject in the bidding program, and evaluating risk factors in the optimal public opinion risk measurement model based on the public opinion keywords to obtain the risk weight of each risk factor in the index layer;
integrating the risk factor influence levels in the index layer and the risk weights, and calculating by using an optimal public opinion risk calculation model to obtain a public opinion risk value of each participating main body;
and based on the role weights of the participating subjects, carrying out weighted summation on the public opinion risk values of the participating subjects according to the role weights to obtain the public opinion risk comprehensive values of all the participating subjects to be used as the public opinion risk comprehensive values of the bidding programs.
As a preferred aspect of the present invention, the calculating a potential delivery loss of an advertisement delivery party according to the public opinion risk obtained by the calculation includes:
the calculation formula of the potential release loss is as follows:
R＝rh；
in the formula, R represents potential delivery loss, R represents a delivery loss coefficient, and h represents a public opinion risk comprehensive value;
the method for determining the release loss coefficient comprises the following steps:
and extracting the fact putting loss and the public opinion risk comprehensive value of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the fact putting loss and the winning price time sequence expansion coefficient by the public opinion risk comprehensive value to obtain the putting loss coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity.
As a preferred embodiment of the present invention, the calculating a lower limit interval of advertisement putting revenue based on the potential putting revenue and the potential putting loss includes:
the calculation formula of the lower limit interval of the advertisement putting party putting income is as follows:
E＝S-R；
in the formula, E represents the advertisement putting party putting income.
As an optimal scheme of the present invention, the constructing of the decision model with success rate and advertisement revenue as dual high optimization objectives based on the lower limit interval of the advertisement revenue and the lower limit interval of the bid pricing of the advertisement tenderer includes:
constructing a first optimization function with the lowest success rate of all advertisement putting parties, wherein the first optimization function is as follows:
in the formula, P is characterized bySuccess rates of all advertising sponsors, d j Characterized as the bid price of the jth advertising sponsor, E j The characteristic is the lower limit interval of the delivery income of the jth advertisement delivery party, j is a metering constant and has no substantial function, and n is the total number of the advertisement delivery parties participating in the advertisement delivery and the bid of the bid inviting program;
constructing a second optimization function with the highest actual profit of the advertisement putting party, wherein the second optimization function is as follows:
M j ＝E j -d j ，d j ≥{d k (k∈[1，n]∩k＝j)}；
in the formula, M j Characterized as the factual revenue of the jth advertiser, d k Characterized as the actual revenue of all advertising sponsors except j;
setting a function expression of a decision model based on the first optimization function and the second optimization function as follows:
in the equation, min is characterized as the minimization operator.
As a preferred aspect of the present invention, there is provided a bid prediction system according to the bid prediction method for advertisement placement, including:
the exposure measuring and calculating model building unit is used for building an exposure measuring and calculating model for the advertising sponsor to measure and calculate the exposure of the bidding programs of the advertising sponsor, and calculating the potential release income of the advertising sponsor and the lower limit interval of bidding pricing of the advertising sponsor according to the measured and calculated exposure;
the public opinion risk measuring and calculating model building unit is used for building a public opinion risk measuring and calculating model for measuring and calculating public opinion risks of the bidding program by the advertising delivery party and calculating potential delivery loss of the advertising delivery party according to the public opinion risks obtained by the measuring and calculating;
and the decision model construction unit is used for calculating a lower limit interval of the advertising revenue of the advertising sponsor based on the potential advertising revenue and the potential advertising loss, and constructing a decision model taking success rate and the advertising revenue as double high optimization targets based on the lower limit interval of the advertising revenue and the lower limit interval of the bidding pricing of the advertising sponsor so as to obtain a bidding prediction result of the bidding program.
Compared with the prior art, the invention has the following beneficial effects:
the method constructs the exposure measuring and calculating model for measuring and calculating the exposure of the bidding program of the advertising bidding party, constructs the public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program and constructs the decision model to obtain the bidding prediction result of the bidding program, realizes automatic bidding prediction in the advertising bidding process, effectively improves the accuracy and efficiency of bidding prediction, and simultaneously ensures the bidding success rate and the release income.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary, and that other embodiments can be derived from the drawings provided by those of ordinary skill in the art without inventive effort.
FIG. 1 is a flow chart of a bid prediction method for advertisement placement according to an embodiment of the present invention;
fig. 2 is a block diagram of a bidding prediction system according to an embodiment of the present invention.
The reference numerals in the drawings denote the following, respectively:
1-an exposure measurement model construction unit; 2-public opinion risk measurement model construction unit; 3-a decision model construction unit.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the present invention provides a bid prediction method for advertisement placement, comprising the following steps:
s1, an advertisement delivery party constructs an exposure measurement model for measuring and calculating the exposure of a bid inviting program of the advertisement delivery party, and calculates the potential delivery income of the advertisement delivery party and a lower limit interval of bid inviting pricing of the advertisement delivery party according to the measured and calculated exposure;
the exposure directly concerns the profit conversion of advertisement putting parties, no matter the sale profit of actual commodities or the stock expansion profit of stock markets, so the exposure of bidding programs needs to be measured firstly.
The method for constructing the exposure measurement model by the advertisement putting party comprises the following steps:
acquiring program attribute characteristics and program exposure of historical programs at an advertising tenderer, and quantizing each historical program into a single longitudinal sample in sequence, wherein the program attribute characteristics are subjected to semantic vectorization to obtain a program attribute characteristic vector as a longitudinal sample characteristic, and the program exposure is subjected to data vectorization to obtain a program exposure vector as a longitudinal sample label;
taking the longitudinal sample characteristics as the input of a BP neural network, taking a longitudinal sample label as the output of the BP neural network, and carrying out model training on the BP neural network by using all longitudinal samples to obtain an exposure measurement model;
preferably, the longitudinal sample balance adjustment is performed before model training is performed on the BP neural network by all longitudinal samples to improve the training efficiency and the training precision of the exposure measurement model under the condition of limited number of longitudinal samples, and the method comprises the following steps:
extracting program attribute features of the bidding program, and performing semantic vectorization on the program attributes of the bidding program to obtain a program attribute feature vector of the bidding program;
sequentially calculating the feature similarity of each longitudinal sample and the bidding program, wherein the feature similarity is measured by using Euclidean distances of the longitudinal sample features and the program attribute feature vector of the bidding program, and the calculation formula of the feature similarity is as follows:
in the formula, p i Characterizing feature similarity, X, between the ith vertical sample and the tender program i The characteristic is the longitudinal sample characteristic of the ith longitudinal sample, Y is the program attribute characteristic vector of the bidding program, i is a metering constant and has no substantial meaning;
setting an equalization adjustment threshold, wherein,
if the feature similarity p i If the value is larger than or equal to the balance adjustment threshold value, performing vertical sample retention on the vertical sample i to improve the vertical sample concentration with the attribute characteristics similar to those of the bidding program;
if the feature similarity p i If the sample number is less than the balance adjustment threshold, longitudinal sample elimination is carried out on the longitudinal sample i so as to reduce the concentration of the longitudinal sample with different attribute characteristics from the bidding program;
the method has the advantages that the concentration of the longitudinal samples with similar attribute characteristics to the bidding program is improved, the concentration of the longitudinal samples with different attribute characteristics to the bidding program is reduced, the balance of the samples can be changed, the longitudinal samples with similar attribute characteristics to the bidding program are biased, the models can rapidly and accurately learn the attribute characteristics of the longitudinal samples with similar attribute characteristics to the bidding program, and the exposure of the bidding program can be rapidly and accurately calculated by the trained models.
The vertical sample concentration is characterized by the proportion of vertical samples with different attribute characteristics with the bidding program or vertical samples with similar attribute characteristics with the bidding program in all the vertical samples, and the appearance of the balance adjustment of the vertical samples comprises the increase or decrease of the vertical sample concentration.
The exposure measurement model measures and calculates the program exposure of the bidding program, and comprises the following steps:
and inputting the program attribute feature vector of the bidding program into the exposure measurement model to obtain the program exposure vector of the bidding program, and converting the program exposure vector of the bidding program into the program exposure.
Calculating the potential delivery income of the advertisement delivery party and the lower limit interval of the bid pricing of the advertisement bid inviting party according to the measured and calculated advertisement exposure, and comprising the following steps of:
the calculation formula of the potential release yield is as follows:
S＝W 1 a 1 Z+W 2 a 2 Z；
in the formula, S is characterized as potential release yield, a 1 、a 2 Respectively characterized by the conversion rate of the program exposure and the income of the entity commodity, the conversion rate of the program exposure and the income of the stock price, W 1 、W 2 Respectively representing the potential influence weight of the program exposure on the entity commodity income, the potential influence weight of the program exposure and the share price income, and Z representing the program exposure;
commodity conversion and stock price conversion;
the calculation formula of the lower limit interval of bid pricing is as follows:
T＝bZ；
in the formula, T is characterized as a lower limit interval of bid pricing, and b is characterized as an interval of a program exposure pricing coefficient;
preferably, the potential impact weight is determined by a financial report revenue structure of the advertising sponsor, wherein,
in the formula, Q 1 、Q 2 Respectively representing the entity commodity income and the stock price income in the financial newspaper;
the method for determining the interval upper limit and the interval lower limit in the interval of the program exposure pricing coefficient comprises the following steps:
extracting the program exposure and the bid price in the advertisement of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the bid price in the advertisement and the bid price time sequence expansion coefficient by the product of the program exposure and the exposure time sequence expansion coefficient to obtain a program exposure pricing coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity;
selecting all historical programs of all advertising bidding parties as transverse samples, calculating the feature similarity between all the transverse samples and the bidding programs, selecting the program exposure and the bidding price in the advertising corresponding to the transverse sample with the maximum feature similarity, and dividing the product of the bidding price in the advertising and the bidding price time sequence expansion coefficient by the program exposure and the exposure time sequence expansion coefficient to obtain the program exposure pricing coefficient of the historical program corresponding to the transverse sample with the maximum feature similarity;
respectively taking the maximum value and the minimum value in the program exposure pricing coefficient of the historical program corresponding to the longitudinal sample and the program exposure pricing coefficient of the historical program corresponding to the transverse sample as an upper interval limit and a lower interval limit;
the expansion coefficient of the price marking time sequence is the currency expansion rate of the time sequence of the bidding program and the time sequence of the historical program, and the expansion coefficient of the exposure time sequence is the ratio of the audience of the time sequence of the bidding program to the audience of the time sequence of the historical program.
S2, the advertisement delivery party constructs a public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program, and calculates the potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measuring and calculating;
the biggest risk factor in the bidding program comes from crisis public affairs of the entertainers, and once the entertainers have a malignant public opinion event, the entertainers will have the influence of rectification and stop broadcasting on the bidding program and will also lose the advertising party, so that the public opinion risk calculation is carried out aiming at the main body of the participating entertainers to calculate the public opinion risk of the bidding program.
The public opinion risk measurement model established by the advertisement delivery party comprises the following steps:
the historical public opinion events are extracted from the participating main bodies related to the crisis public relations events in the internet,
extracting public opinion risk factors based on a historical public opinion event and a crisis public affair workflow rule base to construct an air risk factor set, constructing an expert group, expressing the influence level of the risk factor set by using a phrase set by an expert, and collecting expression data of the influence level to form a risk factor influence level data set;
the method comprises the steps that a three-layer structure of an initial public opinion risk measurement model is built by utilizing an analytic hierarchy process, and each level node is defined, wherein the public opinion risk measurement model consists of a target layer, a criterion layer and an index layer, wherein the target layer determines that a main body of evaluation is public opinion risk, the criterion layer defines four standard criteria of a participation main body, event attributes, public opinion state and public affair cost based on fishbone graph analysis of historical public opinion events, and the index layer consists of risk factors and influence levels corresponding to the risk factors;
constructing an evaluation matrix based on the risk factor influence level data set, and calculating by using an extended good and bad solution distance method to synthesize a multi-dimensional evaluation matrix to obtain the influence levels of each risk factor;
and determining the arrangement order of each layer element in the initial public opinion risk measurement model by using an analytic hierarchy process based on the influence levels of each risk factor to obtain the optimal public opinion risk measurement model with a complete system structure.
Public sentiment is carried out by taking a participating subject of a bidding program as a keyword in the Internet.
The public opinion risk measuring and calculating model measures the public opinion risk of the bidding program, and comprises the following steps:
collecting public opinion keywords of each participating subject in the bidding program, and evaluating risk factors in the optimal public opinion risk measurement model based on the public opinion keywords to obtain the risk weight of each risk factor in the index layer;
integrating the risk factor influence grade and the risk weight in the index layer, and calculating by using an optimal public opinion risk calculation model to obtain a public opinion risk value of each participating subject;
and based on the role weights of the participating subjects, carrying out weighted summation on the public opinion risk values of the participating subjects according to the role weights to obtain the public opinion risk comprehensive values of all the participating subjects to be used as the public opinion risk comprehensive values of the bidding programs.
Calculate advertisement putting side's potential delivery loss according to calculating the public opinion risk who obtains, include:
the potential drop loss is calculated by the formula:
R＝rh；
in the formula, R is characterized as potential delivery loss, R is characterized as a delivery loss coefficient, and h is characterized as a public opinion risk comprehensive value;
the method for determining the release loss coefficient comprises the following steps:
and extracting the fact putting loss and the public opinion risk comprehensive value of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the fact putting loss and the winning price time sequence expansion coefficient by the public opinion risk comprehensive value to obtain the putting loss coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity.
Calculate the lower bound interval of advertisement putting side input profit based on potential input profit and potential input loss, include:
the calculation formula of the lower limit interval of the advertisement putting party putting income is as follows:
E＝S-R；
in the formula, E represents the advertisement putting party putting income.
And S3, calculating a lower limit interval of the advertisement putting profit of the advertisement putting party based on the potential putting profit and the potential putting loss, and constructing a decision model taking the success rate and the putting profit as double high optimization targets based on the lower limit interval of the putting profit and the lower limit interval of the bid pricing of the advertisement putting party so as to obtain a bid prediction result of the bid program.
Based on the lower limit interval of the release income and the lower limit interval of the bid pricing of the advertising tenderer, a decision model which takes the success rate and the release income as double high optimization targets is constructed, and the decision model comprises the following steps:
constructing a first optimization function with the lowest success rate of all advertisement putting parties, wherein the first optimization function is as follows:
in the formula, P is characterized by being broadInforming the success rate of the sponsor, d j Characterized as the bid price of the jth advertising sponsor, E j The characteristic is the lower limit interval of the delivery income of the jth advertisement delivery party, j is a metering constant and has no substantial function, and n is the total number of the advertisement delivery parties participating in the advertisement delivery and the bid of the bid inviting program;
constructing a second optimization function with the highest actual profit of the advertisement putting party, wherein the second optimization function is as follows:
M j ＝E j -d j ，d j ≥{d k (k∈[1，n]∩k＝j)}；
in the formula, M j Characterized as the factual revenue of the jth advertiser, d k Characterized as the actual revenue of all advertising sponsors except j;
setting a function expression of the decision model based on the first optimization function and the second optimization function as follows:
in the equation, min is characterized as the minimize operator.
And solving the decision model to obtain a bidding prediction result on advertisement delivery.
As shown in fig. 2, based on the bid prediction method for advertisement delivery, the present invention provides a bid prediction system, including:
the exposure measurement and calculation model construction unit 1 is used for constructing an exposure measurement and calculation model for the advertising sponsor to measure and calculate the exposure of the bidding program of the advertising sponsor, and calculate the potential release income of the advertising sponsor and the lower limit interval of the bidding pricing of the advertising sponsor according to the measured and calculated exposure;
the public opinion risk measuring and calculating model building unit 2 is used for building a public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program by the advertisement delivery party and calculating the potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measuring and calculating;
and the decision model building unit 3 is used for calculating a lower limit interval of the advertisement putting profit of the advertisement putting party based on the potential putting profit and the potential putting loss, and building a decision model taking the success rate and the putting profit as double high optimization targets based on the lower limit interval of the putting profit and the lower limit interval of the bid pricing of the advertisement putting party so as to obtain a bid prediction result of the bid program.
The method constructs the exposure measuring and calculating model for measuring and calculating the exposure of the bidding program of the advertising bidding party, constructs the public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program and constructs the decision model to obtain the bidding prediction result of the bidding program, realizes automatic bidding prediction in the advertising bidding process, effectively improves the precision and efficiency of bidding prediction, and simultaneously ensures the bidding success rate and the delivery income.
The above embodiments are only exemplary embodiments of the present application, and are not intended to limit the present application, and the protection scope of the present application is defined by the claims. Various modifications and equivalents may be made by those skilled in the art within the spirit and scope of the present application and such modifications and equivalents should also be considered to be within the scope of the present application.
Claims (7)
1. A bid prediction method for advertisement placement, comprising the steps of:
s1, an advertisement delivery party constructs an exposure measurement model for measuring and calculating the exposure of a bid inviting program of the advertisement delivery party, and calculates the potential delivery income of the advertisement delivery party and a lower limit interval of bid inviting pricing of the advertisement delivery party according to the measured and calculated exposure;
s2, the advertisement delivery party constructs a public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program, and calculates the potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measuring and calculating;
s3, calculating a lower limit interval of the advertisement putting income of the advertisement putting party based on the potential putting income and the potential putting loss, and constructing a decision model taking success rate and putting income as double high optimization targets based on the lower limit interval of the putting income and the lower limit interval of the bid pricing of the advertisement putting party to obtain a bid prediction result of the bid program;
the method for constructing the exposure measurement model by the advertisement putting party comprises the following steps:
acquiring program attribute characteristics and program exposure of historical programs at an advertising tenderer, and quantizing each historical program into a single longitudinal sample in sequence, wherein the program attribute characteristics are subjected to semantic vectorization to obtain a program attribute characteristic vector as a longitudinal sample characteristic, and the program exposure is subjected to data vectorization to obtain a program exposure vector as a longitudinal sample label;
taking the longitudinal sample characteristics as the input of a BP neural network, taking the longitudinal sample label as the output of the BP neural network, and carrying out model training on the BP neural network by using all longitudinal samples to obtain an exposure measurement model;
the method for improving the training efficiency and the training precision of the exposure measurement model under the condition of limited number of longitudinal samples by performing longitudinal sample balance adjustment before model training is performed on the BP neural network by all the longitudinal samples comprises the following steps:
extracting program attribute features of the bidding program, and performing semantic vectorization on the program attributes of the bidding program to obtain a program attribute feature vector of the bidding program;
sequentially calculating the feature similarity of each longitudinal sample and the bidding program, wherein the feature similarity is measured by using Euclidean distances of the longitudinal sample features and the program attribute feature vector of the bidding program, and the calculation formula of the feature similarity is as follows:
in the formula, p i Characterizing the similarity of the ith vertical sample to the features of the bidding document, X i The characteristic is the longitudinal sample characteristic of the ith longitudinal sample, Y is the program attribute characteristic vector of the bidding program, i is a metering constant and has no substantial meaning;
setting an equalization adjustment threshold, wherein,
if the feature similarity p i Greater than or equal to the threshold for equality adjustment, then the vertical samples are sampledThe method includes the steps that (i) longitudinal sample preservation is conducted, so that the concentration of longitudinal samples with similar attribute characteristics to the bidding program is improved;
if the feature similarity p i If the sample number is less than the balance adjustment threshold, longitudinal sample elimination is carried out on the longitudinal sample i so as to reduce the concentration of the longitudinal sample with different attribute characteristics from the bidding program;
the longitudinal sample concentration is characterized by the proportion of longitudinal samples with different attribute characteristics with the bidding program or longitudinal samples with similar attribute characteristics with the bidding program in all longitudinal samples, and the aspect of the adjustment of the longitudinal sample equilibrium comprises the increase or decrease of the longitudinal sample concentration;
calculating the potential delivery income of the advertisement delivery party and the lower limit interval of the bid pricing of the advertisement bid inviting party according to the advertisement exposure obtained by measurement and calculation, wherein the lower limit interval comprises the following steps:
the potential release yield calculation formula is as follows:
S＝W 1 a 1 Z＋W 2 a 2 Z；
wherein S is characterized as a potential revenue for delivery,a 1 、a 2 respectively characterized by the conversion rate of the program exposure and the income of the entity commodity, the conversion rate of the program exposure and the income of the stock price, W 1 、W 2 Respectively representing the potential influence weight of the program exposure on the entity commodity income, the potential influence weight of the program exposure and the share price income, and Z representing the program exposure;
the calculation formula of the lower limit interval of bid pricing is as follows:
T＝bZ；
in the formula, T is characterized as a lower limit interval of bid pricing, and b is characterized as an interval of a program exposure pricing coefficient;
the potential impact weight is determined by the revenue structure of the advertising sponsor, wherein,
in the formula, Q 1 、Q 2 Respectively representing the entity commodity income and the stock price income in the financial newspaper;
the method for determining the interval upper limit and the interval lower limit in the interval of the program exposure pricing coefficient comprises the following steps:
extracting the program exposure and the bid price in the advertisement of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the bid price in the advertisement and the bid price time sequence expansion coefficient by the product of the program exposure and the exposure time sequence expansion coefficient to obtain a program exposure pricing coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity;
selecting all historical programs of all advertising bidding parties as transverse samples, calculating the feature similarity between all the transverse samples and the bidding programs, selecting the program exposure and the bidding price in the advertising corresponding to the transverse sample with the maximum feature similarity, and dividing the product of the bidding price in the advertising and the bidding price time sequence expansion coefficient by the program exposure and the exposure time sequence expansion coefficient to obtain the program exposure pricing coefficient of the historical program corresponding to the transverse sample with the maximum feature similarity;
respectively taking the maximum value and the minimum value in the program exposure pricing coefficient of the historical program corresponding to the longitudinal sample and the program exposure pricing coefficient of the historical program corresponding to the transverse sample as an upper interval limit and a lower interval limit;
the price marking time sequence expansion coefficient is the currency expansion rate of the time sequence of the bidding program and the time sequence of the historical program, and the exposure time sequence expansion coefficient is the ratio of audience to audience of the time sequence of the bidding program to the time sequence of the historical program;
based on the lower limit interval of the release income and the lower limit interval of the bid pricing of the advertising tenderer, a decision model which takes the success rate and the release income as double high optimization targets is constructed, and the decision model comprises the following steps:
constructing a first optimization function with the lowest success rate of all advertisement putting parties, wherein the first optimization function is as follows:
where P is characterized as the success rate of all advertising sponsors,d j characterized as the bid price of the jth advertising sponsor, E j The characteristic is the lower limit interval of the delivery income of the jth advertisement delivery party, j is a metering constant and has no substantial function, and n is the total number of the advertisement delivery parties participating in the advertisement delivery and the bid of the bid inviting program;
constructing a second optimization function with the highest actual profit of the advertisement putting party, wherein the second optimization function is as follows:
M j ＝E j －d j ,d j ≥ {d k （k∈[1,n]∩k＝j）}；
in the formula, M j Characterized as the factual revenue of the jth advertiser, d k Characterized as the actual revenue of all advertising sponsors except j;
setting a function expression of the decision model based on the first optimization function and the second optimization function as follows:
in the equation, min is characterized as the minimization operator.
2. The method of claim 1, wherein the method comprises: the exposure measurement model is used for measuring and calculating the program exposure of the bidding program and comprises the following steps:
inputting the program attribute feature vector of the bidding program into the exposure measurement model to obtain the program exposure vector of the bidding program, and converting the program exposure vector of the bidding program into the program exposure.
3. The method of claim 2, wherein the method comprises: advertisement putting side constructs public opinion risk and calculates model, includes:
extracting historical public sentiment events from a performance main body related to crisis public relations events in the Internet;
extracting public opinion risk factors based on a historical public opinion event and a crisis public affair workflow rule base to construct an risk factor set and construct an expert group, expressing the influence level of the risk factor set by using a phrase set by an expert, and collecting expression data of the influence level to form a risk factor influence level data set;
the method comprises the steps that a three-layer structure of an initial public opinion risk measurement model is built by utilizing an analytic hierarchy process, and each level node is defined, wherein the public opinion risk measurement model consists of a target layer, a criterion layer and an index layer, wherein the target layer determines that a main body of evaluation is public opinion risk, the criterion layer defines four standard criteria of a participation main body, event attributes, public opinion state and public affair cost based on fishbone map analysis of historical public opinion events, and the index layer consists of risk factors and influence levels corresponding to the risk factors;
constructing an evaluation matrix based on the risk factor influence level data set, and calculating by using an extended good and bad solution distance method to synthesize a multi-dimensional evaluation matrix to obtain the influence levels of each risk factor;
determining the arrangement order of each layer element in the initial public opinion risk measurement model by using an analytic hierarchy process based on the influence levels of each risk factor to obtain an optimal public opinion risk measurement model with a complete system structure;
and carrying out public opinion risk measurement and calculation by taking the participation main body of the bidding program as a keyword in the Internet.
4. The method of claim 3, wherein the method comprises: the public opinion risk calculating model calculates the public opinion risk of the bidding program, and comprises the following steps:
collecting public opinion keywords of each participating subject in the bidding program, and evaluating risk factors in the optimal public opinion risk measurement model based on the public opinion keywords to obtain the risk weight of each risk factor in the index layer;
fusing the risk factor influence grade and the risk weight in the index layer, and obtaining the public opinion risk value of each participating subject by the optimal public opinion risk measurement and calculation model;
and based on the role weights of the participating subjects, carrying out weighted summation on the public opinion risk values of the participating subjects according to the role weights to obtain the public opinion risk comprehensive values of all the participating subjects to be used as the public opinion risk comprehensive values of the bidding programs.
5. The method of claim 4, wherein the method comprises: the public opinion risk that obtains according to calculating calculates advertisement putting side's potential input loss, includes:
the calculation formula of the potential release loss is as follows:
R＝rh；
in the formula, R is characterized as potential delivery loss, R is characterized as a delivery loss coefficient, and h is characterized as a public opinion risk comprehensive value;
the method for determining the release loss coefficient comprises the following steps:
and extracting the fact putting loss and the public opinion risk comprehensive value of the historical program corresponding to the longitudinal sample with the maximum feature similarity, and dividing the product of the fact putting loss and the winning price time sequence expansion coefficient by the public opinion risk comprehensive value to obtain the putting loss coefficient of the historical program corresponding to the longitudinal sample with the maximum feature similarity.
6. The method of claim 5, wherein the calculating a lower bound interval of advertisement putting profit based on the potential putting profit and potential putting loss comprises:
the calculation formula of the lower limit interval of the advertisement putting party putting income is as follows:
E＝S－R；
in the formula, E represents the advertisement putting party putting income.
7. A bid prediction system of the bid prediction method for advertisement placement according to any one of claims 1 to 6, comprising:
the exposure measurement and calculation model construction unit (1) is used for constructing an exposure measurement and calculation model for the advertising sponsor to measure and calculate the exposure of the bidding program of the advertising sponsor, and calculate the potential release income of the advertising sponsor and the lower limit interval of the bidding pricing of the advertising sponsor according to the measured and calculated exposure;
the public opinion risk measuring and calculating model building unit (2) is used for building a public opinion risk measuring and calculating model for measuring and calculating the public opinion risk of the bidding program by the advertisement delivery party and calculating the potential delivery loss of the advertisement delivery party according to the public opinion risk obtained by measuring and calculating;
and the decision model building unit (3) is used for calculating a lower limit interval of the advertisement putting profit of the advertisement putting party based on the potential putting profit and the potential putting loss, and building a decision model taking the success rate and the putting profit as double high optimization targets based on the lower limit interval of the putting profit and the lower limit interval of the bid pricing of the advertisement putting party so as to obtain a bid prediction result of the bid program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210193539.8A CN114565408B (en) | 2022-03-01 | 2022-03-01 | Bidding prediction method and system for advertisement putting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210193539.8A CN114565408B (en) | 2022-03-01 | 2022-03-01 | Bidding prediction method and system for advertisement putting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114565408A CN114565408A (en) | 2022-05-31 |
| CN114565408B true CN114565408B (en) | 2023-03-24 |
Family
ID=81715510
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210193539.8A Active CN114565408B (en) | 2022-03-01 | 2022-03-01 | Bidding prediction method and system for advertisement putting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114565408B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116451139B (en) * | 2023-06-16 | 2023-09-01 | 杭州新航互动科技有限公司 | Live broadcast data rapid analysis method based on artificial intelligence |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184611A (en) * | 2015-09-08 | 2015-12-23 | 精硕世纪科技（北京）有限公司 | Advertising effect quantification method and display system |
| CN109472632A (en) * | 2018-09-25 | 2019-03-15 | 平安科技（深圳）有限公司 | Evaluate method, apparatus, medium and the electronic equipment of advertisement competition power |
| CN113761084B (en) * | 2020-06-03 | 2023-08-08 | 北京四维图新科技股份有限公司 | POI search ranking model training method, ranking device, method and medium |
| CN112734154B (en) * | 2020-11-16 | 2023-08-01 | 中山大学 | Multi-factor public opinion risk assessment method based on fuzzy number similarity |
| CN112396471B (en) * | 2020-12-10 | 2021-05-11 | 杭州次元岛科技有限公司 | Advertisement putting optimization method and device based on big data |
| CN113947435A (en) * | 2021-10-22 | 2022-01-18 | 北京明略软件系统有限公司 | Multi-dimensional advertisement effect evaluation method, system, electronic device and storage medium |
-
2022
- 2022-03-01 CN CN202210193539.8A patent/CN114565408B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114565408A (en) | 2022-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Knetsch et al. | Comparison of methods for recreation evaluation | |
| Ethier et al. | A comparison of hypothetical phone and mail contingent valuation responses for green-pricing electricity programs | |
| US7908238B1 (en) | Prediction engines using probability tree and computing node probabilities for the probability tree | |
| US20090216619A1 (en) | Method for determining fair market values of multimedia advertising spaces | |
| CN107977859A (en) | Advertisement placement method, device, computing device and storage medium | |
| US20030069822A1 (en) | Corporate value evaluation system | |
| CN106372959A (en) | Internet-based user access behavior digital marketing system and method | |
| CN110443687B (en) | Electronic commerce platform based on big data | |
| CN109615442B (en) | RTB real-time bidding method based on incentive video advertisement | |
| TWI652639B (en) | Recommended system and method of product promotion combination | |
| CN107526810A (en) | Establish method and device, methods of exhibiting and the device of clicking rate prediction model | |
| CN114565408B (en) | Bidding prediction method and system for advertisement putting | |
| EP2353137A2 (en) | Systems and methods for risk management of sports-associated businesses | |
| CN112163886B (en) | Real-time bidding advertisement resource allocation method based on reinforcement learning | |
| Wu | Matching value and market design in online advertising networks: An empirical analysis | |
| CN115689655A (en) | Intelligent efficient automatic global analysis SCRM marketing system | |
| Loomis et al. | A hybrid individual—zonal travel cost model for estimating the consumer surplus of golfing in Colorado | |
| CN114862464A (en) | Advertisement putting effect estimation method and device | |
| Khandker et al. | Price determination for 4G service using price sensitivity model in India | |
| CN101359997A (en) | Method for automatically computing network advertisement grade and displaying advertisement | |
| CN111401943A (en) | Multi-source advertisement bidding system and method | |
| CN114493724A (en) | Multi-task learning-based bidding keyword effect estimation model construction method | |
| CN108510353A (en) | A kind of e-commerce system that intelligence degree is high | |
| CN110428281A (en) | Combine the method and apparatus for determining peers amount for a variety of related products | |
| CN114358809A (en) | Data analysis system applying deep neural network to digital marketing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |