Advertisement click rate prediction method and system based on multivariate data feature fusion

Abstract

The invention discloses an advertisement click-through rate prediction method and system based on multivariate data feature fusion, and the method comprises the steps: determining an extraction time period of data for prediction calculation, and extracting the carrier features of an advertisement putting carrier based on the extraction time period of the data, the carrier characteristics of the advertisement putting carrier comprise a same-category advertisement click rate data set of the advertisement putting carrier in an extraction time period; calculating a predicted base value of the advertisement click rate according to the carrier characteristics of the advertisement putting carrier; the content features of the advertisement content are extracted, the content features of the advertisement content comprise brands, prices and image-text information related to the advertisement content, and a brand influence coefficient, a price influence coefficient and an image-text influence coefficient are calculated; further calculating a correction coefficient of advertisement click rate prediction according to content characteristics of the advertisement content; and finally, correcting the prediction base value of the advertisement click-through rate based on the correction coefficient to obtain a prediction value of the advertisement click-through rate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, in particular, to an advertisement click rate prediction method and system based on multi-element data feature fusion. BACKGROUND

[0002] With the development of the era of self-media, the self-media platform has become the core field of advertisement placement. The user flow of the self-media platform has the characteristics of periodicity and suddenness. The traditional advertisement prediction method usually adopts a fixed data period prediction calculation method, which has a large error when used in the advertisement placement of the self-media platform. In order to improve the effect of advertisement placement, it is of great significance to design an advertisement click rate prediction method suitable for the high variability characteristics of the self-media platform. SUMMARY

[0003] In order to solve the technical problems proposed in the background art, the purpose of the present application is to provide an advertisement click rate prediction method and system based on multi-element data feature fusion. The present application adopts a dynamic period algorithm to adapt to the characteristics of periodicity and suddenness of the user flow of the self-media platform, and combines the multi-element data features of the advertisement placement carrier and the advertisement content to further improve the accuracy of prediction.

[0004] In order to achieve the above purpose, the present application provides an advertisement click rate prediction method based on multi-element data feature fusion. In the technical solution of the present application, the advertisement click rate prediction method based on multi-element data feature fusion comprises the following steps: determining the extraction time period of the data used for prediction calculation; extracting the carrier features of the advertisement placement carrier based on the extraction time period of the data, including the same category advertisement click rate data set of the advertisement placement carrier in the extraction time period; calculating the prediction base value of the advertisement click rate according to the carrier features of the advertisement placement carrier; extracting the content features of the advertisement content, including the brand, price and image-text involved in the advertisement content, judging whether it has brand attraction, price attraction and image-text attraction, and determining the brand influence coefficient, price influence coefficient and image-text influence coefficient; calculating the correction coefficient of the advertisement click rate prediction according to the content features of the advertisement content; correcting the prediction base value of the advertisement click rate based on the correction coefficient to obtain the prediction value of the advertisement click rate.

[0005] Further, in the technical solution of the present application, the extraction time period of the data used for prediction calculation adopts a dynamic period algorithm, comprising the following steps: Step M1: determining the minimum extraction time period and the maximum extraction time period ,in, This represents the minimum extraction time period. To predict the day before sky, This represents the maximum extraction time period. To predict the day before sky; Step M2: Set the extraction time period Extracting advertising placement carriers within the extraction time period Dataset of click-through rates for ads within the same product category: , ,in, This represents the extraction time period. This indicates the time period for extracting information from the advertising platform. The first Dataset of click-through rates for similar product category ads over a day; Step M3: Calculate the advertising delivery vehicle within the extraction time period. Average daily click-through rate data for ads of the same product category: ,in This indicates the time period for extracting information from the advertising platform. The first Average click-through rate of ads in the same product category over the past day; Step M4: Extract the advertising platform within the specified time period. The average daily click-through rate data for ads of the same product category was linearly fitted to obtain the fitting function: ,in, This indicates the time period for extracting information from the advertising platform. Inner The linear fit value for the day, Represented as a constant term, Represented as a trend coefficient; Step M5: Calculate the coefficient of determination based on the fitted function. Set the coefficient of determination Reliability threshold The coefficient of determination will be calculated. With reliability threshold Comparison: when and Then, return to step M2 and set the extraction time period. The value is Calculate again in the order of the steps; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the maximum extraction time period. ; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the minimum extraction time period. ; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is .

[0006] Furthermore, in the technical solution of the present invention, calculating the predicted baseline value of the ad click-through rate based on the carrier characteristics of the ad placement carrier includes the following steps: Step N1: Based on the advertising platform during the extraction time period Click-through rate dataset for similar product category ads , Calculate the advertising placement carrier during the extraction time period Average daily click-through rate of ads in the same product category ; Step N2: Extract the advertising platform within the specified time period. Average daily click-through rate of ads in the same product category A linear fit is performed to obtain the fitted function: ; Step N3: Calculate the predicted baseline value of the ad click-through rate for the predicted day and subsequent days based on the fitted function, i.e., the predicted day is the [missing value]. The base value for the predicted day is [value]. .

[0007] Furthermore, in the technical solution of this invention, determining whether a brand has appeal, price has appeal, and graphic appeal, and determining the brand influence coefficient, price influence coefficient, and graphic influence coefficient specifically includes: Determining brand appeal and brand influence includes: obtaining information on the frequency of brand mentions on internet platforms and setting a threshold for brand influence assessment. When a brand is mentioned more times on internet platforms than a certain threshold, it is considered a valid reason. To determine brand attractiveness and establish a brand influence coefficient. ,in, When a brand is mentioned less than a certain threshold on internet platforms. And when the advertising content does not involve the brand, the brand influence coefficient is determined as follows: This means judging that the brand lacks appeal and disregards brand influence; Determining whether an ad is price-attractive and determining its price impact factor includes: obtaining the discount range involved in the ad content and setting a price impact threshold. When the discount offered in the advertisement exceeds the threshold for judgment. Determine if it has price attractiveness and determine the price influence coefficient. ,in, When the discount offered by the advertisement is less than the threshold value... And when the advertising content does not involve price, the price influence coefficient is determined as follows: This means judging that the price is not attractive, i.e., not considering the price effect; Determining the appeal of an advertisement and its impact factor involves: obtaining image and text information from the advertisement content; matching the similarity of image and text information to determine the frequency of mentions of similar images and text information on internet platforms; and setting an impact threshold for the advertisement. When the total number of times the image and text information involved in the advertisement is mentioned exceeds the judgment threshold. Determine the attractiveness of the graphic and textual content and determine its influence coefficient. ,in, When the total number of times the image and text information involved in the advertisement is mentioned is less than the judgment threshold. And when the advertisement content does not involve images and text, the image and text influence coefficient is determined as follows: This means judging that the text and images are not attractive, or that the influence of the text and images is not considered.

[0008] Furthermore, in the technical solution of the present invention, in calculating the correction coefficient for predicting the ad click-through rate based on the content characteristics of the ad content, the correction coefficient is the product of the brand influence coefficient, the price influence coefficient, and the image and text influence coefficient.

[0009] Furthermore, in the technical solution of the present invention, the method of correcting the predicted base value of the ad click-through rate based on the correction coefficient to obtain the predicted value of the ad click-through rate includes: The predicted value for calculating ad click-through rate is: ,in, This is represented as a correction factor. This is expressed as a predicted value for ad click-through rate.

[0010] In the technical solution of the present invention, an advertising click-through rate prediction system based on multi-data feature fusion is also provided, for implementing the advertising click-through rate prediction method based on multi-data feature fusion as described above. The advertising click-through rate prediction system based on multi-data feature fusion includes: The data acquisition module includes: The carrier feature acquisition submodule acquires the carrier features of the advertising carrier, including the click-through rate dataset of the same category of advertisements within the extraction time period. The content feature acquisition submodule acquires the content features of the advertising content, including the brand, price, and images and text involved in the advertising content; The data calculation module includes: The extraction time period calculation submodule calculates and determines the extraction time period of the data used for prediction calculations; The prediction baseline calculation submodule calculates the prediction baseline value of the ad click-through rate based on the characteristics of the ad delivery platform. The correction coefficient calculation submodule calculates the correction coefficient for ad click-through rate prediction based on the content characteristics of the ad content. The prediction calculation submodule calculates the predicted click-through rate (CTR) based on the correction coefficient, adjusting the predicted base value of the CTR to obtain the predicted CTR value.

[0011] Beneficial Effects: In summary, this invention provides a method and system for predicting ad click-through rates based on multi-source data feature fusion. In the technical solution of this invention, a dynamic periodic algorithm is used to calculate the data extraction time period for prediction calculation. Different data extraction time periods can be calculated and set according to different data characteristics of the ad placement carrier in different time domains to adapt to the periodic and sudden changes in user traffic on self-media platforms. At the same time, this invention uses the historical click-through rate data trend of the same category of ads on the ad placement carrier for current prediction and uses the content characteristics of the ad content to correct the prediction base value, further improving the accuracy of prediction.

[0012] Other features and advantages of the present invention will be set forth in the following description. Attached Figure Description

[0013] To more clearly illustrate the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating an advertising click-through rate prediction method based on multi-data feature fusion according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating a dynamic period algorithm for calculating the extraction time period according to an embodiment of the present invention. Figure 3 This is a schematic diagram of a process for calculating the baseline value for predicting ad click-through rate according to an embodiment of the present invention; Figure 4This is a schematic diagram of an advertising click-through rate prediction system based on multi-data feature fusion according to an embodiment of the present invention. Detailed Implementation

[0015] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0016] The core of this invention is to provide an advertising click-through rate prediction method and system based on multi-data feature fusion. This embodiment adopts a dynamic periodic algorithm to adapt to the periodic and sudden changes in user traffic on self-media platforms, and combines multi-data features of advertising platforms and advertising content to further improve the accuracy of prediction.

[0017] This embodiment provides a method for predicting ad click-through rates based on the fusion of multiple data features. Figure 1 This is a flowchart illustrating an advertising click-through rate prediction method based on multi-data feature fusion according to an embodiment of the present invention. Figure 1 As shown, in this embodiment, the ad click-through rate prediction method based on multi-data feature fusion includes the following steps: Determine the time period for extracting data used in predictive calculations; Based on the data extraction time period, the carrier characteristics of the advertising carrier are extracted, including the click-through rate dataset of the same category of advertisements on the advertising carrier within the extraction time period; Calculate the predicted baseline value of ad click-through rate based on the characteristics of the ad delivery platform; Extract the content features of the advertisement, including the brand, price, and images and text involved in the advertisement, and determine whether it has brand appeal, price appeal, and image and text appeal, and determine the brand influence coefficient, price influence coefficient, and image and text influence coefficient; Calculate the correction factor for ad click-through rate prediction based on the content characteristics of the ad content; The predicted click-through rate (CTR) is obtained by correcting the base value of the CTR prediction based on the correction factor.

[0018] Specifically, in this embodiment, Figure 2 This is a flowchart illustrating a dynamic period algorithm for calculating the extraction time period according to an embodiment of the present invention, as shown below. Figure 2 As shown, the extraction time period for the data used in the prediction calculation is determined. The dynamic periodic algorithm includes the following steps: Step M1: Determine the minimum extraction time period and maximum extraction time period ,in, This represents the minimum extraction time period. To predict the day before sky, This represents the maximum extraction time period. To predict the day before Days, among which, determine the minimum extraction time period To ensure sufficient data volume and reduce data errors, while determining the maximum extraction time period. Prevent continuous looping; Step M2: Set the extraction time period Extracting advertising placement carriers within the extraction time period Dataset of click-through rates for ads within the same product category: , ,in, This represents the extraction time period. This indicates the time period for extracting information from the advertising platform. The first This dataset contains click-through rates for similar product category ads over a given period of time. The amount and value of the click-through rate data for the same product category on different days within the same dataset are all variable; Step M3: Calculate the advertising delivery vehicle within the extraction time period. Average daily click-through rate data for ads of the same product category: ,in This indicates the time period for extracting information from the advertising platform. The first Average click-through rate of ads in the same product category over the past day; Step M4: Extract the advertising platform within the specified time period. The average daily click-through rate data for ads of the same product category was linearly fitted to obtain the fitting function: ,in, This indicates the time period for extracting information from the advertising platform. Inner The linear fit value for the day, Represented as a constant term, Represented as a trend coefficient; Step M5: Calculate the coefficient of determination based on the fitted function. Among them, the coefficient of determination Based on existing calculation methods, a determination coefficient is set. Reliability threshold The coefficient of determination will be calculated. With reliability threshold Comparison: when and Then, return to step M2 and set the extraction time period. The value is The reliability threshold is obtained by repeating the calculation in the order of steps, i.e., by iterative calculation. Maximum extraction time period The value of is increased to improve the sample size while ensuring the accuracy of the fit, so as to further improve the prediction accuracy. when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the maximum extraction time period. That is, within the maximum extraction time period Within the specified time frame, click-through rate data for ads of the same category across all advertising platforms can be used; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the minimum extraction time period. That is, within the minimum extraction time period Within a given timeframe, the click-through rate data for ads of the same product category across different advertising platforms fluctuates significantly, indicating a large prediction error in this situation. when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is That is, the reliability threshold is obtained through iterative calculation. Maximum extraction time period The value is the extraction time period calculated last time. The value is .

[0019] Specifically, in this embodiment, Figure 3 This is a schematic diagram of a process for calculating the predicted baseline value of ad click-through rate according to an embodiment of the present invention, such as... Figure 3 As shown, calculating the predicted baseline value of ad click-through rate based on the characteristics of the ad delivery platform includes the following steps: Step N1: Based on the advertising platform during the extraction time period Click-through rate dataset for similar product category ads , Calculate the advertising placement carrier during the extraction time period Average daily click-through rate of ads in the same product category ; Step N2: Extract the advertising platform within the specified time period. Average daily click-through rate of ads in the same product category A linear fit is performed to obtain the fitted function: ; Step N3: Calculate the predicted baseline value of the ad click-through rate for the predicted day and subsequent days based on the fitted function, i.e., the predicted day is the [missing value]. The base value for the predicted day is [value]. .

[0020] Specifically, in this embodiment, determining whether a brand has appeal, price has appeal, and graphic appeal, and determining the brand influence coefficient, price influence coefficient, and graphic influence coefficient specifically includes: Determining brand appeal and brand influence includes: obtaining information on the frequency of brand mentions on internet platforms and setting a threshold for brand influence assessment. When a brand is mentioned more times on internet platforms than a certain threshold, it is considered a valid reason. To determine brand attractiveness and establish a brand influence coefficient. ,in, When a brand is mentioned less than a certain threshold on internet platforms. And when the advertising content does not involve the brand, the brand influence coefficient is determined as follows: This means judging that the brand lacks appeal and disregards brand influence; Determining whether an ad is price-attractive and determining its price impact factor includes: obtaining the discount range involved in the ad content and setting a price impact threshold. When the discount offered in the advertisement exceeds the threshold for judgment. Determine if it has price attractiveness and determine the price influence coefficient. ,in, When the discount offered by the advertisement is less than the threshold value... And when the advertising content does not involve price, the price influence coefficient is determined as follows: This means judging that the price is not attractive, i.e., not considering the price effect; Determining the appeal of an advertisement and its impact factor involves: obtaining image and text information from the advertisement content; matching the similarity of image and text information to determine the frequency of mentions of similar images and text information on internet platforms; and setting an impact threshold for the advertisement. When the total number of times the image and text information involved in the advertisement is mentioned exceeds the judgment threshold. Determine the attractiveness of the graphic and textual content and determine its influence coefficient. ,in, When the total number of times the image and text information involved in the advertisement is mentioned is less than the judgment threshold. And when the advertisement content does not involve images and text, the image and text influence coefficient is determined as follows: This means judging that the text and images are not attractive, or that the influence of the text and images is not considered.

[0021] Specifically, in this embodiment, in calculating the correction coefficient for predicting the ad click-through rate based on the content characteristics of the ad content, the correction coefficient is the product of the brand influence coefficient, the price influence coefficient, and the image and text influence coefficient.

[0022] Specifically, in this embodiment, the predicted value of the ad click-through rate is obtained by correcting the predicted base value of the ad click-through rate based on the correction coefficient, including: The predicted value for calculating ad click-through rate is: ,in, This is represented as a correction factor. This is expressed as a predicted value for ad click-through rate.

[0023] This embodiment also provides an advertising click-through rate prediction system based on multi-data feature fusion, used to implement the advertising click-through rate prediction method based on multi-data feature fusion as described above. Figure 4 This is a schematic diagram of an advertising click-through rate prediction system based on multi-data feature fusion according to an embodiment of the present invention, as shown below. Figure 4 As shown, in this embodiment, the ad click-through rate prediction system includes: The data acquisition module includes: The carrier feature acquisition submodule acquires the carrier features of the advertising carrier, including the click-through rate dataset of the same category of advertisements within the extraction time period. The content feature acquisition submodule acquires the content features of the advertising content, including the brand, price, and images and text involved in the advertising content; The data calculation module includes: The extraction time period calculation submodule calculates and determines the extraction time period of the data used for prediction calculations; The prediction baseline calculation submodule calculates the prediction baseline value of the ad click-through rate based on the characteristics of the ad delivery platform. The correction coefficient calculation submodule calculates the correction coefficient for ad click-through rate prediction based on the content characteristics of the ad content. The prediction calculation submodule calculates the predicted click-through rate (CTR) based on the correction coefficient, adjusting the predicted base value of the CTR to obtain the predicted CTR value.

[0024] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting ad click-through rate based on multi-source data feature fusion, characterized in that, Includes the following steps: Determine the time period for extracting data used in predictive calculations; Based on the data extraction time period, the carrier characteristics of the advertising carrier are extracted, including the click-through rate dataset of the same category of advertisements on the advertising carrier within the extraction time period; Calculate the predicted baseline value of ad click-through rate based on the characteristics of the ad delivery platform; Extract the content features of the advertisement, including the brand, price, and images and text involved in the advertisement, and determine whether it has brand appeal, price appeal, and image and text appeal, and determine the brand influence coefficient, price influence coefficient, and image and text influence coefficient; Calculate the correction factor for ad click-through rate prediction based on the content characteristics of the ad content; The predicted click-through rate (CTR) is obtained by correcting the base value of the CTR prediction based on the correction factor.

2. The ad click-through rate prediction method based on multi-source data feature fusion according to claim 1, characterized in that, The dynamic period algorithm is used to determine the extraction time period of the data used for prediction calculations, and includes the following steps: Step M1: Determine the minimum extraction time period and maximum extraction time period ,in, This represents the minimum extraction time period. To predict the day before sky, This represents the maximum extraction time period. To predict the day before sky; Step M2: Set the extraction time period Extracting advertising placement carriers within the extraction time period Dataset of click-through rates for ads within the same product category: , ,in, This represents the extraction time period. This indicates the time period for extracting information from the advertising platform. The first Dataset of click-through rates for similar product category ads over a day; Step M3: Calculate the advertising delivery vehicle within the extraction time period. Average daily click-through rate data for ads of the same product category: ,in This indicates the time period for extracting information from the advertising platform. The first Average click-through rate of ads in the same product category over the past day; Step M4: Extract the advertising platform within the specified time period. The average daily click-through rate data for ads of the same product category was linearly fitted to obtain the fitting function: ,in, This indicates the time period for extracting information from the advertising platform. Inner The linear fit value for the day, Represented as a constant term, Represented as a trend coefficient; Step M5: Calculate the coefficient of determination based on the fitted function. Set the coefficient of determination Reliability threshold The coefficient of determination will be calculated. With reliability threshold Comparison: when and Then, return to step M2 and set the extraction time period. The value is Calculate again in the order of the steps; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the maximum extraction time period. ; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is the minimum extraction time period. ; when and At that time, the extraction time period for the data used in the prediction calculation is obtained. The value is .

3. The ad click-through rate prediction method based on multi-source data feature fusion according to claim 2, characterized in that, Calculating the predicted baseline value of ad click-through rate based on the characteristics of the ad delivery platform includes the following steps: Step N1: Based on the advertising platform during the extraction time period Click-through rate dataset for similar product category ads , Calculate the advertising placement carrier during the extraction time period Average daily click-through rate of ads in the same product category ; Step N2: Extract the advertising platform within the specified time period. Average daily click-through rate of ads in the same product category A linear fit is performed to obtain the fitted function: ; Step N3: Calculate the predicted baseline value of the ad click-through rate for the predicted day and subsequent days based on the fitted function, i.e., the predicted day is the [missing value]. The base value for the predicted day is [value]. .

4. The ad click-through rate prediction method based on multi-source data feature fusion according to claim 3, characterized in that, To determine whether a brand possesses brand appeal, price appeal, and graphic appeal, the specific steps for determining the brand influence coefficient, price influence coefficient, and graphic influence coefficient include: Determining brand appeal and brand influence includes: obtaining information on the frequency of brand mentions on internet platforms and setting a threshold for brand influence assessment. When a brand is mentioned more times on internet platforms than a certain threshold, it is considered a valid reason. To determine brand attractiveness and establish a brand influence coefficient. ,in, When a brand is mentioned less than a certain threshold on internet platforms. And when the advertising content does not involve the brand, the brand influence coefficient is determined as follows: This means judging that the brand lacks appeal and disregards brand influence; Determining whether an ad is price-attractive and determining its price impact factor includes: obtaining the discount range involved in the ad content and setting a price impact threshold. When the discount offered in the advertisement exceeds the threshold for judgment. Determine if it has price attractiveness and determine the price influence coefficient. ,in, When the discount offered by the advertisement is less than the threshold value... And when the advertising content does not involve price, the price influence coefficient is determined as follows: This means judging that the price is not attractive, i.e., not considering the price effect; Determining the appeal of an advertisement and its impact factor involves: obtaining image and text information from the advertisement content; matching the similarity of image and text information to determine the frequency of mentions of similar images and text information on internet platforms; and setting an impact threshold for the advertisement. When the total number of times the image and text information involved in the advertisement is mentioned exceeds the judgment threshold. Determine the attractiveness of the graphic and textual content and determine its influence coefficient. ,in, When the total number of times the image and text information involved in the advertisement is mentioned is less than the judgment threshold. And when the advertisement content does not involve images and text, the influence coefficient of images and text is determined as follows: This means judging that the text and images are not attractive, or that the influence of the text and images is not considered.

5. The ad click-through rate prediction method based on multi-source data feature fusion according to claim 4, characterized in that, In the calculation of the correction coefficient for predicting ad click-through rate based on the content characteristics of the ad content, the correction coefficient is the product of the brand influence coefficient, the price influence coefficient, and the image and text influence coefficient.

6. The method for predicting ad click-through rate based on multi-source data feature fusion according to claim 5, characterized in that, The predicted click-through rate (CTR) values ​​are obtained by correcting the baseline value of the CTR prediction based on the correction factor, including: The predicted value for calculating ad click-through rate is: ,in, This is represented as a correction factor. This is expressed as a predicted value for ad click-through rate.

7. An advertising click-through rate prediction system based on multi-source data feature fusion, used to implement the advertising click-through rate prediction method based on multi-source data feature fusion as described in any one of claims 1-6, characterized in that, include: The data acquisition module includes: The carrier feature acquisition submodule acquires the carrier features of the advertising carrier, including the click-through rate dataset of the same category of advertisements within the extraction time period. The content feature acquisition submodule acquires the content features of the advertising content, including the brand, price, and images and text involved in the advertising content; The data calculation module includes: The extraction time period calculation submodule calculates and determines the extraction time period of the data used for prediction calculations; The prediction baseline calculation submodule calculates the prediction baseline value of the ad click-through rate based on the characteristics of the ad delivery platform. The correction coefficient calculation submodule calculates the correction coefficient for ad click-through rate prediction based on the content characteristics of the ad content. The prediction calculation submodule calculates the predicted click-through rate (CTR) based on the correction coefficient, adjusting the predicted base value of the CTR to obtain the predicted CTR value.

Images