A media data processing method, device and equipment

Abstract

The present application discloses a media data processing method, device and equipment, the method comprising: first obtaining the media delivery data of the target clue advertisement including the basic information data, planned delivery information data and historical delivery information data of the target advertiser, then performing feature extraction on these data to obtain the feature vector corresponding to the target clue advertisement, and determining the constraint feature vector with a positive or negative correlation with the corresponding conversion quantity; then inputting the feature vector and the constraint feature vector corresponding to the target clue advertisement into a pre-built conversion quantity estimation model to estimate the conversion quantity of the target clue advertisement. In this way, by using the pre-built conversion quantity estimation model to automatically estimate the feature vector corresponding to the media delivery data of the target clue advertisement, effective processing such as automatic estimation can be performed, which can facilitate users to perceive the delivery effect in advance, and adjust the delivery resource configuration more reasonably accordingly, thereby optimizing the delivery effect and improving the user's delivery experience.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device and equipment for processing media data. Background Art

[0002] With the rapid development of the Internet and short video platforms, more and more advertisers are placing lead ads on short video platforms, and more and more media data needs to be processed. For many advertisers, the estimated results of the number of lead ad conversions are closely related to the direct benefits of their lead ads.

[0003] However, there is currently no good media data processing method to accurately estimate the number of lead ads conversions to meet advertisers' estimation needs. Usually, they simply provide some empirical data as a reference for estimation, or advertisers pre-build a delivery plan and start delivery. After a period of delivery, by observing the actual conversion data, the number of conversions that may be obtained in the future delivery time is manually estimated, so that advertisers can determine whether to adjust the pre-built delivery plan in the future delivery time. However, these processing solutions cannot accurately estimate the number of conversions after the lead ads are delivered during the process of advertisers building delivery plans and before the lead ads are delivered, nor can they provide advertisers with reference guidance for building delivery plans, resulting in a poor delivery experience for advertisers. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a media data processing method, device and equipment, which can facilitate users to perceive the delivery effect in advance through effective processing of media delivery data related to clue advertisements, and adjust the delivery resource configuration more reasonably and accurately based on this, thereby optimizing the delivery effect and improving the user's delivery experience.

[0005] The present application provides a method for processing media data, including:

[0006] Acquire media delivery data of the target lead advertisement to be processed; the media delivery data of the target lead advertisement includes basic information data, planned delivery information data and historical delivery information data of the target advertiser;

[0007] Extracting features from the basic information data, planned delivery information data, and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement;

[0008] Based on the media delivery data of the target clue advertisement, determining a constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement;

[0009] The feature vector corresponding to the target clue advertisement and the constraint feature vector are input into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0010] In a possible implementation, extracting features from the basic information data, planned delivery information data, and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement includes:

[0011] Converting the monetary data of different units contained in different amount characteristics in the basic information data, planned delivery information data, and historical delivery information data of the target advertiser into monetary data of a unified preset unit;

[0012] splicing the character string data corresponding to multiple elements contained in different category features in the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a single character string data corresponding to each category feature;

[0013] The amounts of the currency data in a uniform preset unit contained in the different amount features are concatenated with the single character string data corresponding to each category feature, and the obtained concatenated feature vector is used as the feature vector corresponding to the target clue advertisement.

[0014] In a possible implementation, before inputting the feature vector corresponding to the target lead advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement, the method further includes:

[0015] In response to the target advertiser selecting a bidding method, determining whether to use a tolerance cost as input data;

[0016] If yes, the feature vector corresponding to the target lead advertisement and the constraint feature vector are input into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement, including:

[0017] The feature vector corresponding to the tolerance cost, the feature vector corresponding to the target clue advertisement, and the constraint feature vector are inputted into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0018] In a possible implementation, the conversion quantity estimation model includes a binary classification estimation model and a multi-classification estimation model; the step of inputting the feature vector corresponding to the target clue advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement includes:

[0019] Inputting the feature vector corresponding to the target clue advertisement and the constraint feature vector into the binary classification prediction model to estimate whether the conversion quantity corresponding to the target clue advertisement is greater than zero;

[0020] If so, the feature vector corresponding to the target clue advertisement and the constraint feature vector are input into the multi-classification prediction model to estimate the preset classification interval to which the conversion quantity corresponding to the target clue advertisement belongs.

[0021] In a possible implementation, the basic information data of the target advertiser includes the account balance and the industry to which the target advertiser belongs; the planned delivery information data includes the type of delivery user, planned budget information, bid information, and delivery cycle.

[0022] In a possible implementation, the conversion quantity estimation model is constructed as follows:

[0023] Acquire media delivery data of sample clue advertisements; the media delivery data of sample clue advertisements includes basic information data of sample advertisers, planned sample delivery information data and historical sample delivery information data;

[0024] Extracting features from the basic information data, planned sample delivery information data, and historical sample delivery information data of the sample advertiser to obtain a sample feature vector corresponding to the sample lead advertisement;

[0025] Based on the media delivery data of the sample clue advertisement, determine a sample constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the sample clue advertisement;

[0026] The initial conversion quantity estimation model is trained according to the sample feature vectors, sample constraint feature vectors and conversion quantity estimation classification labels corresponding to the sample lead advertisements to generate the conversion quantity estimation model.

[0027] In a possible implementation, the initial conversion quantity estimation model is a tree model LightGBM.

[0028] In a possible implementation, the method further includes:

[0029] Acquire media delivery data for verifying clue advertisements; the media delivery data for verifying clue advertisements includes basic information data of verified advertisers, planned verification delivery information data, and historical verification delivery information data;

[0030] Extracting features from the basic information data, planned verification delivery information data, and historical verification delivery information data of the verification advertiser to obtain a verification feature vector corresponding to the verification clue advertisement;

[0031] Based on the media delivery data of the verification clue advertisement, determining a verification constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the verification clue advertisement;

[0032] Inputting the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement into the conversion quantity estimation model to obtain the conversion quantity estimation result of the verification clue advertisement;

[0033] When the conversion quantity estimation result of the verification clue advertisement is inconsistent with the conversion quantity marking result corresponding to the verification clue advertisement, the media delivery data of the verification clue advertisement is re-used as the media delivery data of the sample clue advertisement to update the conversion quantity estimation model.

[0034] The present application also provides a media data processing device, the device comprising:

[0035] The first acquisition unit is used to acquire media delivery data of the target clue advertisement to be processed; the media delivery data of the target clue advertisement includes basic information data, planned delivery information data and historical delivery information data of the target advertiser;

[0036] A first extraction unit is used to extract features from the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement;

[0037] A first determining unit is used to determine, based on the media delivery data of the target clue advertisement, a constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement;

[0038] The estimation unit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0039] In a possible implementation, the first extraction unit includes:

[0040] A conversion subunit, configured to convert the monetary data of different units contained in the different amount characteristics in the basic information data, the planned delivery information data and the historical delivery information data of the target advertiser into the monetary data of a unified preset unit;

[0041] A first concatenation subunit is used to concatenate character string data corresponding to multiple elements contained in different category features in the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a single character string data corresponding to each category feature;

[0042] The second splicing subunit is used to splice the amount of the currency data of the unified preset unit contained in the different amount features with the single character string data corresponding to each category feature, and use the obtained spliced ​​feature vector as the feature vector corresponding to the target clue advertisement.

[0043] In a possible implementation manner, the device further includes:

[0044] a determination unit, configured to determine whether to use the tolerable cost as input data in response to the target advertiser selecting a bidding method;

[0045] If yes, then the estimation unit is specifically used for:

[0046] The feature vector corresponding to the tolerance cost, the feature vector corresponding to the target clue advertisement, and the constraint feature vector are inputted into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0047] In a possible implementation, the conversion quantity estimation model includes a binary classification estimation model and a multi-classification estimation model; the estimation unit includes:

[0048] A first estimation subunit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the binary classification estimation model to estimate whether the conversion quantity corresponding to the target clue advertisement is greater than zero;

[0049] The second estimation sub-unit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the multi-classification estimation model if the estimated conversion number corresponding to the target clue advertisement is greater than zero, so as to estimate the preset classification interval to which the conversion number corresponding to the target clue advertisement belongs.

[0050] In a possible implementation, the basic information data of the target advertiser includes the account balance and the industry to which the target advertiser belongs; the planned delivery information data includes the type of delivery user, planned budget information, bid information, and delivery cycle.

[0051] In a possible implementation manner, the device further includes:

[0052] The second acquisition unit is used to acquire media delivery data of sample clue advertisements; the media delivery data of sample clue advertisements includes basic information data of sample advertisers, planned sample delivery information data and historical sample delivery information data;

[0053] A second extraction unit is used to extract features from the basic information data, planned sample delivery information data and historical sample delivery information data of the sample advertiser to obtain a sample feature vector corresponding to the sample lead advertisement;

[0054] A second determination unit is used to determine, based on the media delivery data of the sample clue advertisement, a sample constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the sample clue advertisement;

[0055] The training unit is used to train the initial conversion quantity estimation model according to the sample feature vector of the sample clue advertisement, the sample constraint feature vector and the conversion quantity estimation classification label corresponding to the sample clue advertisement to generate the conversion quantity estimation model.

[0056] In a possible implementation, the initial conversion quantity estimation model is a tree model LightGBM.

[0057] In a possible implementation manner, the device further includes:

[0058] The third acquisition unit is used to acquire media delivery data of the verification clue advertisement; the media delivery data of the verification clue advertisement includes basic information data of the verification advertiser, planned verification delivery information data and historical verification delivery information data;

[0059] A third extraction unit is used to extract features from the basic information data, planned verification delivery information data and historical verification delivery information data of the verification advertiser to obtain a verification feature vector corresponding to the verification clue advertisement;

[0060] A third determination unit is used to determine, based on the media delivery data of the verification clue advertisement, a verification constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the verification clue advertisement;

[0061] An obtaining unit, configured to input the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement into the conversion quantity estimation model to obtain a conversion quantity estimation result of the verification clue advertisement;

[0062] An updating unit is used to update the conversion quantity estimation model by reusing the media delivery data of the verification clue advertisement as the media delivery data of the sample clue advertisement when the conversion quantity estimation result of the verification clue advertisement is inconsistent with the conversion quantity marking result corresponding to the verification clue advertisement.

[0063] The embodiment of the present application also provides a media data processing device, including: a processor, a memory, and a system bus;

[0064] The processor and the memory are connected via the system bus;

[0065] The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes any one of the implementations of the above-mentioned media data processing method.

[0066] An embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any one of the implementations of the above-mentioned media data processing method.

[0067] A media data processing method, device and equipment provided in an embodiment of the present application first obtains the media delivery data of the target clue advertisement to be processed; wherein the media delivery data of the target clue advertisement includes the basic information data, planned delivery information data and historical delivery information data of the target advertiser, and then the basic information data, planned delivery information data and historical delivery information data of the target advertiser are feature extracted to obtain the feature vector corresponding to the target clue advertisement, and then, based on the media delivery data of the target clue advertisement, the constraint feature vector having a positive or negative correlation with the conversion number corresponding to the target clue advertisement is determined; and then the feature vector and the constraint feature vector corresponding to the target clue advertisement can be input into a pre-constructed conversion number estimation model to estimate the conversion number corresponding to the target clue advertisement. It can be seen that since the embodiment of the present application uses a pre-constructed conversion number estimation model to automatically estimate the feature vector corresponding to the media delivery data of the target clue advertisement, so as to quickly and accurately estimate the conversion number after the lead advertisement is delivered, it can provide a strong reference basis for users when delivering advertisements, so that they can perceive the delivery effect in advance, and adjust the delivery resource configuration more accurately and reasonably accordingly, so as to improve the user's delivery experience while optimizing the delivery effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0069] Figure 1 A flowchart of a media data processing method provided in an embodiment of the present application;

[0070] Figure 2An example diagram of a process for inputting a feature vector and a constraint feature vector corresponding to a target clue advertisement into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement provided in an embodiment of the present application;

[0071] Figure 3 A schematic diagram of the composition of a media data processing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0072] With the rapid development of short video platforms, placing lead ads on short video platforms has become the main advertising method adopted by advertisers, and the number of lead ad conversions is closely related to the advertiser's final revenue. For example, for automobile manufacturers, they usually choose to place test drive ads to platform users on short video platforms such as TikTok, and the amount of form data containing all personal information and expected test drive information filled in by users is the number of conversions of the test drive ads placed for them. Based on these feedback form data, automobile manufacturers can formulate more appropriate production plans and sales plans for the vehicles they produce in order to increase their final revenue. Therefore, it is particularly important to improve the accuracy of the estimated results of the number of lead ad conversions so that advertisers can build reasonable delivery plans.

[0073] However, when advertisers create lead ads on advertising platforms, they cannot perceive in advance what kind of conversion benefits the currently created delivery plan can bring. Especially for some advertisers who are placing ads on short video platforms such as TikTok for the first time, they often do not know how to build a delivery plan because they have no experience in placement. In other words, there is currently no better method for processing media delivery data to accurately estimate the number of lead ad conversions to meet the advertiser's estimation needs. Usually, it is simply based on the historical delivery results of other advertisers to provide some experience data as a reference for estimation, or the advertiser pre-builds a delivery plan and starts to deliver, and then after a period of delivery, by observing the actual conversion data, manually estimates the number of conversions that may be obtained in the future delivery time, so that the advertiser can determine whether to adjust the pre-built delivery plan in the future delivery time. However, these processing solutions cannot accurately estimate the number of conversions after the lead ad is delivered during the process of the advertiser building the delivery plan and before the lead ad is delivered, nor can they provide advertisers with reference guidance for building a delivery plan, resulting in a poor delivery experience for advertisers.

[0074] To address the above-mentioned defects, the present application provides a media data processing method, device and equipment, which first obtains the media delivery data of the target lead advertisement to be processed; wherein the media delivery data of the target lead advertisement includes the basic information data, planned delivery information data and historical delivery information data of the target advertiser, and then performs feature extraction on the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement, and then, based on the media delivery data of the target lead advertisement, determines a constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the target lead advertisement; and then, the feature vector and the constraint feature vector corresponding to the target lead advertisement can be input into a pre-constructed conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement. It can be seen that since the embodiment of the present application utilizes a pre-built conversion quantity estimation model to perform automatic estimation and other effective processing on the feature vectors corresponding to the media delivery data of the target lead advertisement, the number of conversions after the lead advertisement is delivered can be quickly and accurately estimated, thereby providing users with a strong reference basis when delivering advertisements, so that they can perceive the delivery effect in advance and adjust the delivery resource allocation more accurately and reasonably accordingly, thereby optimizing the delivery effect while improving the user's delivery experience.

[0075] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0076] First embodiment

[0077] See also Figure 1 , is a flow chart of a media data processing method provided in this embodiment, the method comprising the following steps:

[0078] S101: Acquire media delivery data of a target lead advertisement to be processed; wherein the media delivery data of the target lead advertisement includes basic information data, planned delivery information data, and historical delivery information data of a target advertiser.

[0079] In this embodiment, the lead advertisement for which the conversion quantity estimation is required is defined as a target lead advertisement, and the advertiser who places the target lead advertisement is defined as a target advertiser. The target lead advertisement will include a contact form for users to fill in their personal details to express their interest in the goods or services displayed in the target lead advertisement, so as to facilitate the target advertiser to contact these users for accurate sales of goods or services.

[0080] In order to more quickly and accurately estimate the number of conversions after the target lead advertisement is delivered, it is first necessary to obtain the media delivery data of the target lead advertisement to be processed in order to execute subsequent steps S102 and S103. It should be noted that this embodiment does not limit the method of obtaining the media delivery data of the target lead advertisement, but it is necessary to ensure that the media delivery data of the target lead advertisement is obtained with the consent of the target advertiser and in compliance with legal regulations. At the same time, the media delivery data of the target lead advertisement needs to include the basic information data, planned delivery information data and historical delivery information data of the target advertiser.

[0081] Among them, the basic information data of the target advertiser includes but is not limited to the target advertiser's account balance and the industry to which it belongs, such as the target advertiser's account balance is 10 million US dollars, the target advertiser belongs to the automobile industry and the type of the line advertisement it places is automobile advertisement, etc. The planned delivery information data includes but is not limited to the target advertiser's delivery user type, planned budget information, bid information, delivery cycle and other characteristic data, such as the target advertiser plans to target the target line advertisement to young people aged 20 to 40, the target advertiser's planned budget is 1 million US dollars, the target advertiser's bid is about 200 US dollars and the delivery cycle is 10 days, etc. The historical delivery information data includes but is not limited to the characteristic data indicating the effect of the target advertiser's past delivery of other line advertisements, such as 100 (conversion number) feedback form data obtained after a certain line advertisement was placed in a certain period of time in the past.

[0082] It should be noted that in the embodiments of the present application, the basic information data, planned delivery information data, and historical delivery information data of the target advertiser generally do not involve sensitive information of the target advertiser, and these data are obtained and used after authorization by the target advertiser. In one example, before obtaining these data, the corresponding interface displays prompt information related to obtaining data use authorization, and the target advertiser determines whether to agree to the authorization based on the prompt information.

[0083] S102: Extract features from the basic information data, planned delivery information data, and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement.

[0084] In this embodiment, in order to more quickly and accurately estimate the number of conversions after the target lead advertisement is delivered, so that users can perceive the delivery effect in advance, after the media delivery data of the target lead advertisement is obtained through step S101, the basic information data, planned delivery information data and historical delivery information data of the target advertiser contained therein can be further processed using existing or future data feature extraction methods to extract feature vectors that can represent their data information as feature vectors corresponding to the target lead advertisement, so as to execute the subsequent step S104.

[0085] Specifically, an optional implementation manner is that the implementation process of step S102 may specifically include the following steps AC:

[0086] Step A: converting the monetary data of different units contained in the different amount characteristics in the basic information data, planned delivery information data and historical delivery information data of the target advertiser into monetary data of a unified preset unit.

[0087] In this implementation, in order to improve the accuracy of the processing results, it is necessary to convert the monetary data of different units contained in the media delivery data of the target clue advertisement into monetary data of a unified preset unit before it can be processed and used. Among them, the specific value of the unified preset unit can be set according to the actual situation, and the embodiment of the present application does not limit this. For example, the unified preset unit can be set to US dollars, and then the monetary data of different units such as euros and pounds can be converted into US dollars to perform the subsequent step C.

[0088] For example, for the "budget" and "bid" contained in the planned delivery information data, which may contain amount characteristic data of different units of currency data, if the "budget" and "bid" correspond to two different units of currency data, namely, Euro and British Pound, the currency amounts of the two can be converted into US dollars to unify the currency amount units for executing the subsequent step C.

[0089] Step B: concatenate the character string data corresponding to multiple elements contained in different category features in the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a single character string data corresponding to each category feature.

[0090] In this implementation, in order to improve the accuracy of the estimation results, it is also necessary to concatenate the character strings corresponding to the different category feature data containing multiple elements in the media delivery data of the target lead advertisement, so that each category feature corresponds to a unique string data, which is used to characterize the corresponding category feature and to execute the subsequent step C.

[0091] For example, for different categories of feature data such as "delivery targeting" and "traffic position" included in the planned delivery information data, which may contain multiple elements, if "delivery targeting" corresponds to two elements, and the character strings of these two elements are 0 and 1 respectively, and 0 indicates that the target line advertisement is delivered in country A, and 1 indicates that the target line advertisement is delivered in country B, then the two elements can be spliced ​​together, and the splicing result (such as 2) can be used as a single character string data corresponding to the category feature of "delivery targeting" to execute the subsequent step C.

[0092] Step C: concatenate the amounts of the currency data of the unified preset units contained in the different amount features with the single character string data corresponding to each category feature, and use the concatenated feature vector obtained as the feature vector corresponding to the target clue advertisement.

[0093] In this implementation, after obtaining the amount of monetary data in a unified preset unit contained in different amount features through step A, and obtaining the single string data corresponding to each category feature through step B, further, in order to improve the accuracy of the estimation result, the two can be spliced ​​to obtain a feature vector as the overall feature vector corresponding to the target lead advertisement, which carries all relevant delivery data information of the target lead advertisement.

[0094] It should be noted that, in actual applications, some characteristic data in the media delivery data of the target lead advertisements need to be further processed in combination with actual conditions so as to meet the actual operating conditions. For example, for the amount characteristic data of "budget", assuming that the amount in the "planned budget information" of the target advertiser is smaller than its "account balance", then the corresponding effective "budget" cannot exceed the "account balance". Specifically, the value of the effective budget can be obtained by combining the campaign, ad group budget and the account balance of the target advertiser. The specific calculation process is consistent with the existing method and will not be repeated here.

[0095] S103: Based on the media delivery data of the target clue advertisement, determine a constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement.

[0096] In this embodiment, in order to improve the accuracy of the processing results, so that users can perceive the delivery effect in advance. After the media delivery data of the target clue advertisement is obtained through step S101, it is further possible to screen out feature data that has a positive or negative correlation with the number of conversions corresponding to the target clue advertisement based on the media delivery data of the target clue advertisement. For example, under normal circumstances, the estimated result of the target line advertisement conversion number should be positively correlated with the feature data such as the budget and bid set by the target advertiser, that is, when the target advertiser increases the budget or bid, the estimated number of leads collected should not decrease. However, in the actual training process, the results of the model estimation may not necessarily meet the above requirements. Therefore, after screening out the feature data that has a positive or negative correlation with the number of conversions corresponding to the target clue advertisement, it is necessary to further determine the feature vector corresponding to the entire feature data, and define it as a constraint feature vector as input data to execute the subsequent step S104 to improve the accuracy of the final estimation result.

[0097] Among them, the constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the target lead advertisement can be determined by manually specifying feature correlations, but the specific determination method is not limited in this application. It only needs to ensure that the determined constraint feature vector can make the change of the final model prediction result more logical and more accurate.

[0098] S104: Inputting the feature vector and constraint feature vector corresponding to the target lead advertisement into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement.

[0099] In this embodiment, after extracting the feature vector corresponding to the target lead advertisement through step S102 and determining the constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target lead advertisement through step S103, these two can be further input as input data into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement.

[0100] Specifically, an optional implementation method is that the conversion quantity estimation model may include a binary classification estimation model and a multi-classification estimation model. Thus, after determining the feature vector and constraint feature vector corresponding to the target lead advertisement through steps S102 and S103, in order to accurately estimate the number of conversions that can be collected after the target lead advertisement is released, the two can first be input into the binary classification estimation model, such as Figure 2 As shown, a classification result of 0 or 1 is output, where 0 indicates that the conversion number of the target lead advertisement is 0, and 1 indicates that the conversion number of the target lead advertisement is greater than 0.

[0101] On this basis, after determining that the number of conversions of the target lead advertisement is greater than 0 through the output results of the binary classification estimation model, further, such as Figure 2 As shown, the feature vector and constraint feature vector corresponding to the target lead advertisement can be further input into the multi-classification prediction model to output a set of vectors representing the classification of the conversion quantity of the target lead advertisement. The value of each dimension in the vector can be a value in the interval [0,1], and the value of each dimension represents the probability value that the conversion quantity of the target lead advertisement belongs to each preset quantity interval. At this time, the quantity interval corresponding to the maximum probability value can be used as the classification interval to which the conversion quantity of the target lead advertisement belongs.

[0102] For example: Assume that there are four preset quantity intervals of "0, (0, 20], (20, 50], (50, inf]", and assume that the output vector of the model is s = [0.001, 0.03, 0.9, 0.06]. It can be seen that the value of the third dimension, 0.9, is the highest. Therefore, the quantity interval corresponding to this dimension is the classification interval to which the conversion number of the target lead advertisement belongs, that is, the conversion number of the target lead advertisement is in (20, 50].

[0103] It should be noted that, in actual applications, the target advertiser may set some restrictions on the cost of receiving each conversion data before launching the target line advertisement, that is, there is a tolerance cost, rather than the default of spending all the budget. For example, assuming that the target advertiser limits the cost of receiving each conversion data to no more than 2 yuan before launching the target line advertisement, then when the budget is 100 yuan, the condition is met only when it is estimated that at least 50 feedback form data can be received.

[0104] Therefore, an optional implementation method is that before executing this step S104, it is necessary to respond to the operation of the target advertiser selecting a bidding method (that is, the operation of selecting whether to limit the cost paid for receiving each conversion data) and determine whether to use the tolerance cost as input data; if so, it is necessary to first determine the feature vector corresponding to the tolerance cost, and then input the feature vector corresponding to the tolerance cost and the feature vector corresponding to the target lead advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to more accurately estimate the conversion quantity corresponding to the target lead advertisement.

[0105] It should be noted that, in order to implement the above step S104, a conversion quantity estimation model needs to be constructed in advance. Next, this embodiment will introduce the construction process of the conversion quantity estimation model. The process includes the following steps (1)-(4):

[0106] Step (1): obtaining media delivery data of sample lead advertisements; wherein the media delivery data of sample lead advertisements includes basic information data of sample advertisers, planned sample delivery information data and historical sample delivery information data.

[0107] In this embodiment, in order to build a conversion quantity estimation model, a lot of preparatory work needs to be done in advance. First, it is necessary to collect the media delivery data of historical lead ads. For example, the media delivery data corresponding to 100 line ads can be collected on the TikTok platform in advance, and the media delivery data of each line ad collected can be used as the media delivery data of the sample lead ad. The quantity classification intervals of the conversion quantities corresponding to these sample lead ads are manually marked in advance to train the conversion quantity estimation model.

[0108] The media delivery data of the sample lead advertisement may include basic information data of the sample advertiser, planned sample delivery information data and historical sample delivery information data.

[0109] Step (2): Extract features from the basic information data of the sample advertiser, the planned sample delivery information data, and the historical sample delivery information data to obtain a sample feature vector corresponding to the sample lead advertisement.

[0110] In this embodiment, after the media delivery data of the sample lead advertisement is obtained through step (1), it cannot be directly used to train and generate a conversion quantity estimation model. Instead, it is necessary to perform feature extraction on the basic information data of the sample advertiser, the planned sample delivery information data, and the historical sample delivery information data contained therein to obtain a sample feature vector corresponding to the sample lead advertisement. The specific extraction process can be found in the above step S102, which only requires replacing the target data with the sample data. The implementation process will not be repeated here.

[0111] Step (3): Based on the media delivery data of the sample lead advertisement, determine the sample constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the sample lead advertisement.

[0112] In this embodiment, after the media delivery data of the sample clue advertisement is obtained through step (1), it cannot be directly used to train and generate a conversion quantity estimation model. Not only is it necessary to obtain the sample feature vector corresponding to the sample clue advertisement through step (2), but it is also necessary to determine the sample constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the sample clue advertisement based on the media delivery data of the sample clue advertisement. The specific extraction process can be referred to the above step S103, and only the target data therein is replaced with the sample data. The implementation process will not be repeated here. Then, the sample feature vector and sample constraint feature vector corresponding to the extracted sample clue advertisement can be used to train a conversion quantity estimation model.

[0113] Step (4): Train the initial conversion quantity estimation model based on the sample feature vector of the sample lead advertisement, the sample constraint feature vector, and the conversion quantity estimation classification label corresponding to the sample lead advertisement to generate a conversion quantity estimation model.

[0114] In this embodiment, after the sample feature vector and sample constraint feature vector of the sample lead advertisement are determined respectively through steps (2) and (3), the initial conversion quantity estimation model can be further trained according to the sample feature vector, sample constraint feature vector and the conversion quantity estimation classification labeling results corresponding to the sample lead advertisement, thereby generating a conversion quantity estimation model.

[0115] Among them, an optional implementation method is that the initial conversion quantity estimation model can be a tree model LightGBM.

[0116] Through the above steps (1)-(4), a conversion quantity estimation model can be generated based on the media delivery data of the sample lead advertisement. Furthermore, the generated conversion quantity estimation model can be verified using the media delivery data of the verification lead advertisement. The specific verification process may include the following steps ①-⑤:

[0117] Step ①: Acquire media delivery data for verifying lead ads; wherein the media delivery data for verifying lead ads includes basic information data of verified advertisers, planned verification delivery information data, and historical verification delivery information data.

[0118] In this embodiment, in order to verify the conversion quantity estimation model, it is first necessary to obtain the media delivery data of the verification lead ads. For example, the media delivery data corresponding to 100 line ads can be collected from the TikTok platform as the media delivery data for the verification lead ads. The media delivery data for the verification lead ads refers to the media data information that can be used to verify the conversion quantity estimation model. The media delivery data for the verification lead ads includes the basic information data of the verification advertiser, the planned verification delivery information data, and the historical verification delivery information data. After obtaining the media delivery data of these verification lead ads and the conversion quantity classification label corresponding to each verification lead ad, the subsequent step ② can be continued.

[0119] Step ②: Extract features from the basic information data of the verified advertiser, the planned verified delivery information data, and the historical verified delivery information data to obtain a verification feature vector corresponding to the verified lead advertisement.

[0120] In actual applications, after obtaining the media delivery data of the verification lead advertisement through step ①, it cannot be directly used to verify the conversion quantity estimation model. Instead, it is necessary to perform feature extraction on the basic information data of the verification advertiser, the planned verification delivery information data, and the historical verification delivery information data contained in the media delivery data of the verification lead advertisement to obtain the verification feature vector corresponding to the verification lead advertisement, so as to execute the subsequent step ④.

[0121] Step 3: Based on the media delivery data of the verification lead advertisement, determine the verification constraint feature vector that has a positive or negative correlation with the conversion quantity corresponding to the verification lead advertisement.

[0122] In actual applications, after obtaining the media delivery data of the verification clue advertisement through step ①, it cannot be directly used to verify the conversion quantity estimation model. It is necessary not only to obtain the verification feature vector corresponding to the verification clue advertisement through step ②, but also to determine the verification constraint feature vector that is positively or negatively correlated with the conversion quantity corresponding to the verification clue advertisement based on the media delivery data of the verification clue advertisement, so as to execute the subsequent step ④.

[0123] Step ④: Input the verification feature vector and the verification constraint feature vector corresponding to the verification lead advertisement into the conversion quantity estimation model to obtain the conversion quantity estimation result of the verification lead advertisement.

[0124] In the specific implementation process, after determining the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement respectively through steps ② and ③, further, the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement can be input into the conversion quantity estimation model to obtain the conversion quantity estimation result of the verification clue advertisement, and then step ⑤ can be continued.

[0125] Step ⑤: When the conversion quantity estimation result of the verification lead advertisement is inconsistent with the conversion quantity marking result corresponding to the verification lead advertisement, the media delivery data of the verification lead advertisement is used as the media delivery data of the sample lead advertisement to update the conversion quantity estimation model.

[0126] After obtaining the estimated conversion quantity result of the verification lead advertisement through step ④, if the estimated conversion quantity result of the verification lead advertisement is inconsistent with the actual classification result of the conversion quantity corresponding to the verification lead advertisement (such as the manually marked quantity classification marking result), the media delivery data of the verification lead advertisement can be used again as the media delivery data of the sample lead advertisement to update the parameters of the conversion quantity estimation model.

[0127] Through the above-mentioned embodiments, the conversion quantity estimation model can be effectively verified by utilizing the media delivery data of the verification lead advertisement. When the conversion quantity estimation classification result of the verification lead advertisement is inconsistent with the actual classification result of the conversion quantity corresponding to the verification lead advertisement (such as the manually labeled quantity classification marking result), the conversion quantity estimation model can be adjusted and updated in time, thereby helping to improve the estimation precision and accuracy of the estimation model.

[0128] In summary, a media data processing method provided in this embodiment first obtains the media delivery data of the target clue advertisement to be processed; wherein, the media delivery data of the target clue advertisement includes the basic information data, planned delivery information data and historical delivery information data of the target advertiser, and then the basic information data, planned delivery information data and historical delivery information data of the target advertiser are feature extracted to obtain the feature vector corresponding to the target clue advertisement, and then, based on the media delivery data of the target clue advertisement, the constraint feature vector having a positive or negative correlation with the conversion number corresponding to the target clue advertisement is determined; and then the feature vector and the constraint feature vector corresponding to the target clue advertisement can be input into the pre-constructed conversion number estimation model to estimate the conversion number corresponding to the target clue advertisement. It can be seen that, since the embodiment of the present application utilizes the pre-constructed conversion number estimation model to automatically estimate the feature vector corresponding to the media delivery data of the target clue advertisement, so as to quickly and accurately estimate the conversion number after the lead advertisement is delivered, it can provide a strong reference basis for users when delivering advertisements, so that they can perceive the delivery effect in advance, and adjust the delivery resource configuration more accurately and reasonably accordingly, so as to improve the user's delivery experience while optimizing the delivery effect.

[0129] Second embodiment

[0130] This embodiment will introduce a media data processing device. For related content, please refer to the above method embodiment.

[0131] See also Figure 3 , is a schematic diagram of the composition of a media data processing device provided in this embodiment, the device 300 includes:

[0132] The first acquisition unit 301 is used to acquire media delivery data of the target clue advertisement to be processed; the media delivery data of the target clue advertisement includes basic information data, planned delivery information data and historical delivery information data of the target advertiser;

[0133] The first extraction unit 302 is used to extract features from the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement;

[0134] A first determining unit 303 is used to determine, based on the media delivery data of the target clue advertisement, a constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement;

[0135] The estimation unit 304 is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0136] In an implementation of this embodiment, the first extracting unit 302 includes:

[0137] A conversion subunit, configured to convert the monetary data of different units contained in the different amount characteristics in the basic information data, the planned delivery information data and the historical delivery information data of the target advertiser into the monetary data of a unified preset unit;

[0138] A first concatenation subunit is used to concatenate character string data corresponding to multiple elements contained in different category features in the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a single character string data corresponding to each category feature;

[0139] The second splicing subunit is used to splice the amount of the currency data of the unified preset unit contained in the different amount features with the single character string data corresponding to each category feature, and use the obtained spliced ​​feature vector as the feature vector corresponding to the target clue advertisement.

[0140] In an implementation of this embodiment, the device further includes:

[0141] a determination unit, configured to determine whether to use the tolerable cost as input data in response to the target advertiser selecting a bidding method;

[0142] If yes, the estimation unit 304 is specifically configured to:

[0143] The feature vector corresponding to the tolerance cost, the feature vector corresponding to the target clue advertisement, and the constraint feature vector are inputted into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

[0144] In one implementation of this embodiment, the conversion quantity estimation model includes a binary classification estimation model and a multi-classification estimation model; the estimation unit 304 includes:

[0145] A first estimation subunit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the binary classification estimation model to estimate whether the conversion quantity corresponding to the target clue advertisement is greater than zero;

[0146] The second estimation sub-unit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the multi-classification estimation model if the estimated conversion number corresponding to the target clue advertisement is greater than zero, so as to estimate the preset classification interval to which the conversion number corresponding to the target clue advertisement belongs.

[0147] In an implementation of this embodiment, the basic information data of the target advertiser includes the account balance and the industry to which the target advertiser belongs; the planned delivery information data includes the type of delivery user, planned budget information, bid information, and delivery cycle.

[0148] In an implementation of this embodiment, the device further includes:

[0149] The second acquisition unit is used to acquire media delivery data of sample clue advertisements; the media delivery data of sample clue advertisements includes basic information data of sample advertisers, planned sample delivery information data and historical sample delivery information data;

[0150] A second extraction unit is used to extract features from the basic information data, planned sample delivery information data and historical sample delivery information data of the sample advertiser to obtain a sample feature vector corresponding to the sample lead advertisement;

[0151] A second determination unit is used to determine, based on the media delivery data of the sample clue advertisement, a sample constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the sample clue advertisement;

[0152] The training unit is used to train the initial conversion quantity estimation model according to the sample feature vector of the sample clue advertisement, the sample constraint feature vector and the conversion quantity estimation classification label corresponding to the sample clue advertisement to generate the conversion quantity estimation model.

[0153] In one implementation of this embodiment, the initial conversion quantity estimation model is a tree model LightGBM.

[0154] In an implementation of this embodiment, the device further includes:

[0155] The third acquisition unit is used to acquire media delivery data of the verification clue advertisement; the media delivery data of the verification clue advertisement includes basic information data of the verification advertiser, planned verification delivery information data and historical verification delivery information data;

[0156] A third extraction unit is used to extract features from the basic information data, planned verification delivery information data and historical verification delivery information data of the verification advertiser to obtain a verification feature vector corresponding to the verification clue advertisement;

[0157] A third determination unit is used to determine, based on the media delivery data of the verification clue advertisement, a verification constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the verification clue advertisement;

[0158] An obtaining unit, configured to input the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement into the conversion quantity estimation model to obtain a conversion quantity estimation result of the verification clue advertisement;

[0159] An updating unit is used to update the conversion quantity estimation model by reusing the media delivery data of the verification clue advertisement as the media delivery data of the sample clue advertisement when the conversion quantity estimation result of the verification clue advertisement is inconsistent with the conversion quantity marking result corresponding to the verification clue advertisement.

[0160] Furthermore, an embodiment of the present application also provides a media data processing device, including: a processor, a memory, and a system bus;

[0161] The processor and the memory are connected via the system bus;

[0162] The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes any one of the implementation methods of the above-mentioned media data processing method.

[0163] Furthermore, an embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any one of the implementation methods of the above-mentioned media data processing method.

[0164] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., including several instructions for enabling a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0165] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0166] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0167] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A media data processing method, characterized in that: The method comprises: Acquire media delivery data of the target lead advertisement to be processed; the media delivery data of the target lead advertisement includes basic information data, planned delivery information data and historical delivery information data of the target advertiser, the planned delivery information data includes the type of the target advertiser's delivery user, planned budget information, bidding information and one or more characteristic data in the delivery cycle, and the historical delivery information data includes characteristic data indicating the effect of other lead advertisements delivered by the target advertiser in the past; Extracting features from the basic information data, planned delivery information data, and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement; Based on the media delivery data of the target clue advertisement, determining a constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement; Inputting the feature vector corresponding to the target clue advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement; The conversion quantity estimation model includes a binary classification estimation model and a multi-classification estimation model; the feature vector corresponding to the target clue advertisement and the constraint feature vector are input into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement, including: Inputting the feature vector corresponding to the target clue advertisement and the constraint feature vector into the binary classification prediction model to estimate whether the conversion quantity corresponding to the target clue advertisement is greater than zero; If so, the feature vector corresponding to the target clue advertisement and the constraint feature vector are input into the multi-classification prediction model to estimate the preset classification interval to which the conversion quantity corresponding to the target clue advertisement belongs.

2. The method according to claim 1, characterized in that: The feature extraction of the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement includes: Converting the monetary data of different units contained in different amount characteristics in the basic information data, planned delivery information data, and historical delivery information data of the target advertiser into monetary data of a unified preset unit; splicing the character string data corresponding to multiple elements contained in different category features in the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a single character string data corresponding to each category feature; The amounts of the currency data in a uniform preset unit contained in the different amount features are concatenated with the single character string data corresponding to each category feature, and the obtained concatenated feature vector is used as the feature vector corresponding to the target clue advertisement.

3. The method according to claim 1, characterized in that Before inputting the feature vector corresponding to the target lead advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement, the method further includes: In response to the target advertiser selecting a bidding method, determining whether to use a tolerance cost as input data; If yes, the feature vector corresponding to the target lead advertisement and the constraint feature vector are input into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target lead advertisement, including: The feature vector corresponding to the tolerance cost, the feature vector corresponding to the target clue advertisement, and the constraint feature vector are inputted into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement.

4. The method according to claim 1, characterized in that: The basic information data of the target advertiser includes the account balance and the industry to which the target advertiser belongs; the planned delivery information data includes the type of delivery user, planned budget information, bid information, and delivery cycle.

5. The method according to claim 1, characterized in that The conversion quantity estimation model is constructed as follows: Acquire media delivery data of sample clue advertisements; the media delivery data of sample clue advertisements includes basic information data of sample advertisers, planned sample delivery information data and historical sample delivery information data; Extracting features from the basic information data, planned sample delivery information data, and historical sample delivery information data of the sample advertiser to obtain a sample feature vector corresponding to the sample lead advertisement; Based on the media delivery data of the sample clue advertisement, determine a sample constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the sample clue advertisement; The initial conversion quantity estimation model is trained according to the sample feature vectors, sample constraint feature vectors and conversion quantity estimation classification labels corresponding to the sample lead advertisements to generate the conversion quantity estimation model.

6. The method according to claim 5, characterized in that The initial conversion quantity estimation model is the tree model LightGBM.

7. The method according to claim 5 or 6, characterized in that: The method further comprises: Acquire media delivery data for verifying clue advertisements; the media delivery data for verifying clue advertisements includes basic information data of verified advertisers, planned verification delivery information data, and historical verification delivery information data; Extracting features from the basic information data, planned verification delivery information data, and historical verification delivery information data of the verification advertiser to obtain a verification feature vector corresponding to the verification clue advertisement; Based on the media delivery data of the verification clue advertisement, determining a verification constraint feature vector having a positive or negative correlation with the number of conversions corresponding to the verification clue advertisement; Inputting the verification feature vector and the verification constraint feature vector corresponding to the verification clue advertisement into the conversion quantity estimation model to obtain the conversion quantity estimation result of the verification clue advertisement; When the conversion quantity estimation result of the verification clue advertisement is inconsistent with the conversion quantity marking result corresponding to the verification clue advertisement, the media delivery data of the verification clue advertisement is re-used as the media delivery data of the sample clue advertisement to update the conversion quantity estimation model.

8. A media data processing device, characterized in that: The device comprises: The first acquisition unit is used to acquire media delivery data of a target lead advertisement to be processed; the media delivery data of the target lead advertisement includes basic information data, planned delivery information data and historical delivery information data of a target advertiser, the planned delivery information data includes the type of delivery user of the target advertiser, planned budget information, bidding information and one or more characteristic data in the delivery cycle, and the historical delivery information data includes characteristic data indicating the effect of other lead advertisements delivered by the target advertiser in the past; A first extraction unit is used to extract features from the basic information data, planned delivery information data and historical delivery information data of the target advertiser to obtain a feature vector corresponding to the target lead advertisement; A first determining unit is used to determine, based on the media delivery data of the target clue advertisement, a constraint feature vector having a positive or negative correlation with the conversion quantity corresponding to the target clue advertisement; An estimation unit, used for inputting the feature vector corresponding to the target clue advertisement and the constraint feature vector into a pre-built conversion quantity estimation model to estimate the conversion quantity corresponding to the target clue advertisement; The conversion quantity estimation model includes a binary classification estimation model and a multi-classification estimation model; the estimation unit includes: A first estimation subunit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the binary classification estimation model to estimate whether the conversion quantity corresponding to the target clue advertisement is greater than zero; The second estimation sub-unit is used to input the feature vector corresponding to the target clue advertisement and the constraint feature vector into the multi-classification estimation model if the estimated conversion number corresponding to the target clue advertisement is greater than zero, so as to estimate the preset classification interval to which the conversion number corresponding to the target clue advertisement belongs.

9. A media data processing device, characterized in that: include: Processor, memory, system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the method according to any one of claims 1 to 7.

Images