A method, device and equipment for determining conversion rate estimation and a storage medium

Abstract

This application discloses a method for determining conversion rate prediction, comprising: acquiring a set of parameters to be optimized, a target CPA, and a sample data set; based on the set of parameters to be optimized and the target CPA, determining the sum of the PCVR of effective clicks for each advertisement according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration; determining the average difference between the sum of the PCVR of effective clicks for each advertisement and the actual total conversions; training the set of parameters to be optimized with the objective of minimizing the average difference to obtain a target parameter set; and determining the calibrated PCVR of the target advertisement based on the target parameter set and the associated data of the target advertisement. This application also discloses an apparatus, device, and medium. This application introduces more data for parameter optimization, thereby improving the accuracy of PCVR.

Description

Technical Field

[0001] This application relates to the field of Internet application technology, and in particular to a method, apparatus, device, and storage medium for determining conversion rate prediction. Background Technology

[0002] In businesses that optimize cost per action (oCPX), the accuracy of the predicted conversion rate (PCVR) directly determines the product's positive impact. Therefore, it is necessary to adjust the PCVR to achieve higher accuracy.

[0003] Traditional statistical conversion rate predictions are not very accurate. To address this issue, the current PCVR calibration strategy involves calibrating the single-time PCVR of an ad during its peak exposure phase, based on the sum of the number of conversions and the PCVR corresponding to valid clicks.

[0004] However, the PCVR calibration strategy described above only utilizes data from the first and second periods of a particular advertisement. Even during the peak exposure phase of an advertisement, the amount of data it generates is still limited. Therefore, it cannot accurately describe the actual situation, resulting in relatively low accuracy of PCVR. Summary of the Invention

[0005] This application provides a method, apparatus, device, and storage medium for determining conversion rate prediction. It not only incorporates more data for parameter optimization but also offers higher reliability compared to human experience, better describing the actual situation and thus improving the accuracy of PCVR.

[0006] In view of this, this application provides a method for determining conversion rate prediction, including:

[0007] Obtain the set of parameters to be optimized, the target cost per action (CPA) value, and the sample data set of K ads. The sample data set includes sample data for each ad, and the sample data for each ad includes the actual total conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period, and K is an integer greater than 1.

[0008] Based on the set of parameters to be optimized and the target CPA, and according to the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, the sum of the PCVR of the effective clicks for each ad is determined.

[0009] Determine the average difference for K ads based on the sum of PCVR of valid clicks for each ad and the actual total conversions;

[0010] With the goal of minimizing the average difference, the set of parameters to be optimized is trained to obtain the target set of parameters, which includes at least one optimized parameter.

[0011] Based on the target parameter set, the PCVR of the target ad after calibration is determined according to the associated data of the target ad and the PCVR to be adjusted. The associated data of the target ad includes the total number of conversions of the target ad in the third period, the total PCVR of the target ad in the third period, the total number of conversions of the target ad in the fourth period, the duration of the target ad in the third period, and the ad spend value. The fourth period is the period preceding the third period.

[0012] Another aspect of this application provides a conversion rate prediction and determination apparatus, comprising:

[0013] The acquisition module is used to acquire the set of parameters to be optimized, the target cost per action (CPA) value, and the sample data set of K ads. The sample data set includes sample data for each ad, and the sample data for each ad includes the actual total number of conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period, and K is an integer greater than 1.

[0014] The determination module is used to determine the sum of the PCVR of the effective clicks for each ad based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration.

[0015] The determination module is also used to determine the average difference for K ads based on the sum of PCVR of valid clicks for each ad and the actual total conversions;

[0016] The training module is used to train the set of parameters to be optimized with the goal of minimizing the average difference, so as to obtain the target parameter set, wherein the target parameter set includes at least one optimized parameter;

[0017] The determination module is also used to determine the calibrated PCVR of the target ad based on the target parameter set, the associated data of the target ad, and the PCVR to be adjusted. The associated data of the target ad includes the total number of conversions of the target ad in the third period, the total PCVR of the target ad in the third period, the total number of conversions of the target ad in the fourth period, the duration of the target ad in the third period, and the ad spend value. The fourth period is the period preceding the third period.

[0018] In one possible design, in another implementation of another aspect of the embodiments of this application,

[0019] The determination module is specifically used to divide the parameter set based on the consumption and the target CPA, and to determine the factor parameters corresponding to each advertisement based on the consumption value corresponding to each advertisement. The consumption division parameters included in the consumption division parameter set are pre-set.

[0020] Based on the set of associated parameters, the smoothing parameters for each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period.

[0021] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0022] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0023] The training module is specifically used to train the set of associated parameters with the goal of minimizing the average difference, so as to obtain the target set of parameters.

[0024] In one possible design, in another implementation of another aspect of the embodiments of this application, the set of associated parameters includes a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter;

[0025] The determination module is specifically used for any one of the K ads. If the total conversions in the second period corresponding to any one ad are not zero, then based on the first association parameter, the second association parameter, the third association parameter, and the fourth association parameter, the smoothing parameter corresponding to any one ad is determined according to the duration in the first period and the total conversions in the second period corresponding to any one ad.

[0026] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter is determined to be the smoothing parameter corresponding to any one ad, where the default association parameter is preset.

[0027] The training module is specifically used to train the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter with the goal of minimizing the average difference, so as to obtain the target parameter set, wherein the target parameter set includes the optimized first correlation parameter, the optimized second correlation parameter, the optimized third correlation parameter, and the optimized fourth correlation parameter.

[0028] In one possible design, in another implementation of another aspect of the embodiments of this application, the set of associated parameters includes default associated parameters;

[0029] The determination module is specifically used for any one of the K ads. If the total conversions in the second period corresponding to any one ad are not zero, then based on the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter, the smoothing parameter corresponding to any one ad is determined according to the duration in the first period and the total conversions in the second period corresponding to any one ad. The first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter are preset.

[0030] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter will be set to the smoothing parameter corresponding to any one ad.

[0031] The training module is specifically used to train the default association parameters with the goal of minimizing the average difference, so as to obtain the target parameter set, which includes the optimized default association parameters.

[0032] In one possible design, in another implementation of another aspect of the embodiments of this application,

[0033] The determination module is specifically used to divide the parameter set and target CPA based on consumption, and to determine the factor parameters corresponding to each advertisement based on the consumption value corresponding to each advertisement.

[0034] Based on the set of associated parameters, the smoothing parameters corresponding to each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The associated parameters included in the set of associated parameters are pre-set.

[0035] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0036] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0037] The training module is specifically used to train the parameter set of consumption with the goal of minimizing the average difference, so as to obtain the target parameter set.

[0038] In one possible design, in another implementation of another aspect of the embodiments of this application, the consumption partitioning parameter set includes N consumption partitioning parameters, where N is an integer greater than or equal to 1;

[0039] The determination module is specifically used to determine the factor parameters corresponding to each advertisement based on the N value, N consumption division parameters and target CPA, and according to the consumption value corresponding to each advertisement. The N value is preset.

[0040] The training module is specifically used to train each of the N consumption partitioning parameters with the goal of minimizing the average difference, so as to obtain the target parameter set, which includes the optimized N consumption partitioning parameters.

[0041] In one possible design, in another implementation of another aspect of the embodiments of this application,

[0042] The determination module is specifically used to determine the smoothing parameters corresponding to each advertisement based on the set of associated parameters, according to the duration of each advertisement in the first period and the total number of conversions in the second period. The set of associated parameters includes the default associated parameters, the first associated parameters, the second associated parameters, the third associated parameters, and the fourth associated parameters.

[0043] Based on the consumption-based parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined. The consumption-based parameter set includes an N value and N consumption-based parameters, where N is an integer greater than or equal to 1.

[0044] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0045] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0046] The training module is specifically used to train the set of correlation parameters with the goal of minimizing the average difference, so as to obtain the target parameter set. The target parameter set includes the optimized first correlation parameter, the optimized second correlation parameter, the optimized third correlation parameter, the optimized fourth correlation parameter, the optimized default correlation parameter, the optimized N value, and the optimized N consumption partitioning parameters.

[0047] In one possible design, in another implementation of another aspect of the embodiments of this application, the sample data for each advertisement also includes advertiser information, the total number of conversions counted per unit time, and the total PCVR counted per unit time.

[0048] The determination module is specifically used to divide K advertisements into T1 advertisement sets based on the advertiser information corresponding to each advertisement, wherein each advertisement set includes at least one advertisement, and T1 is an integer greater than or equal to 1;

[0049] If the total spending value corresponding to each ad set in the T1 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set. The (N+1) spending intervals are determined based on the spending parameter set and the target CPA. Each ad subset includes at least one ad, and the same ad subset corresponds to the same spending interval. N is an integer greater than or equal to 1.

[0050] For each of at least one subset of advertisements, the first conversion and the first PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in each subset within a unit of time, wherein the current period includes at least one unit of time.

[0051] For each of the T1 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the second number of conversions and the second PCVR of each ad set in the current period.

[0052] Based on the first conversion count, the first PCVR, the second conversion count, and the second PCVR, determine the factor parameters corresponding to each advertisement.

[0053] In one possible design, in another implementation of another aspect of the embodiments of this application, the sample data of each advertisement also includes brand information;

[0054] The determination module is specifically used to divide the K ads into T2 ad sets based on the brand information corresponding to each ad if the total spending value of any ad set in the T1 ad sets is less than the spending threshold. Each ad set includes at least one ad, and T2 is an integer greater than or equal to 1.

[0055] If the total spending value corresponding to each ad set in the T2 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0056] For each of at least one subset of advertisements, the third conversion and the third PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in the current period.

[0057] For each of the T2 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the fourth number of conversions and the fourth PCVR of each ad set in the current period.

[0058] Based on the third conversion count, third PCVR, fourth conversion count, and fourth PCVR, determine the factor parameters corresponding to each advertisement.

[0059] In one possible design, in another implementation of another aspect of the embodiments of this application, the sample data of each advertisement also includes product information;

[0060] The determination module is specifically used to divide the K ads into T3 ad sets according to the product information corresponding to each ad if the total consumption value of any ad set in the T2 ad sets is less than the consumption threshold. Each ad set includes at least one ad, and T3 is an integer greater than or equal to 1.

[0061] If the total spending value corresponding to each of the T3 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0062] For each ad subset in at least one ad subset, determine the fifth conversion number and the fifth PCVR of each ad subset in the current period based on the total conversion number and the total PCVR count of each ad in each ad subset within the unit time.

[0063] For each of the T3 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the sixth conversion and the sixth PCVR of each ad set in the current period.

[0064] Based on the fifth conversion count, fifth PCVR, sixth conversion count, and sixth PCVR, determine the factor parameters corresponding to each advertisement.

[0065] In one possible design, in another implementation of another aspect of the embodiments of this application,

[0066] The determination module is specifically used to determine the target CPA range for any given ad, based on the ad's spend value.

[0067] Based on the target CPA range, the target expansion coefficient is determined according to the ad spend value and the total conversions in the first period.

[0068] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters of the advertisement in the first period, determine the calibration parameters to be adjusted for the advertisement.

[0069] The calibration parameters corresponding to the advertisement are determined based on the calibration parameters to be adjusted and the target magnification factor.

[0070] In one possible design, in another implementation of another aspect of the embodiments of this application, the sample data for each advertisement also includes the total number of conversions counted per unit time and the total PCVR counted per unit time;

[0071] The determination module is also used to determine the seventh conversion and the seventh PCVR of the ad in the current period for any ad if the total conversions in the first period corresponding to the ad are less than the conversion threshold and the ad spend is less than the target CPA threshold.

[0072] The determination module is also used to determine the calibration parameters corresponding to the advertisement based on the seventh conversion number of the advertisement in the current period, the seventh PCVR in the current period, and the preset expansion coefficient.

[0073] The determination module is also used to determine the sum of the PCVR of the effective clicks of an ad based on the PCVR before each click calibration and the calibration parameters.

[0074] Another aspect of this application provides a computer device, including: a memory, a processor, and a bus system;

[0075] The memory is used to store programs;

[0076] The processor is used to execute programs in memory, and the processor is used to execute the methods provided by the above aspects according to the instructions in the program code;

[0077] Bus systems are used to connect memory and processor to enable communication between them.

[0078] Another aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods described above.

[0079] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the above aspects.

[0080] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0081] This application provides a method for determining conversion rate prediction. First, it obtains a set of parameters to be optimized, a target cost-per-action (CPA) value, and a sample data set of K advertisements. This sample data set includes sample data for each advertisement, comprising the actual total conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period. Then, based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration for each advertisement, the sum of the PCVR of effective clicks for each advertisement is determined. Then, based on the sum of the PCVR of effective clicks for each advertisement and the actual total conversions, the average difference for the K advertisements is determined. Finally, with the goal of minimizing the average difference, the set of parameters to be optimized is trained to obtain the target parameter set. Based on this, the calibrated PCVR of the target ad can be determined according to the target parameter set, target CPA, and the correlation data of the target ad. In the above method, the PCVR calibration of the target ad utilizes the optimized target parameter set for calculation. Each optimized parameter in the target parameter set is trained based on sample data from multiple ads. Therefore, not only is more data used for parameter optimization, but it also has higher reliability compared to human experience values, better describing the actual situation and thus improving the accuracy of PCVR. Attached Figure Description

[0082] Figure 1 This is a schematic diagram of the architecture of the conversion rate prediction and determination system in the embodiments of this application;

[0083] Figure 2 This is a schematic diagram of the advertising stage in an embodiment of this application;

[0084] Figure 3 This is a flowchart illustrating a method for determining conversion rate prediction in an embodiment of this application.

[0085] Figure 4 This is a schematic diagram of the conversion rate prediction and determination device in the embodiments of this application;

[0086] Figure 5 This is a schematic diagram of the structure of a computer device in an embodiment of this application. Detailed Implementation

[0087] This application provides a method, apparatus, device, and storage medium for determining conversion rate prediction. It not only incorporates more data but also offers higher reliability compared to human experience, better describing the actual situation and thus improving the accuracy of PCVR.

[0088] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0089] Internet advertising is one of the most important business models, and improving advertising return on investment (ROI) involves attracting more users to place ads on the platform. This raises the issue of promotion efficiency. Ad performance is typically measured through impressions, clicks, and conversions. Most advertising systems, limited by data feedback, can only optimize based on impressions or clicks as the primary metrics. Some advertising platforms help advertisers monitor post-campaign conversion rates. By training a Predicted Conversion Rate (PCVR) model based on conversion data and incorporating a PCVR factor into ad ranking, advertising performance can be optimized, thereby improving return on investment (ROI).

[0090] To improve the accuracy of PCVR, this application proposes a PCVR determination method, which is applied to... Figure 1The PCVR determination system shown in the figure includes a server and a database, and may also include advertising screens and / or terminal devices. The client is deployed on the terminal device. The client can run on the terminal device via a browser or as a standalone application (APP). The specific presentation format of the client is not limited here. The server involved in this application can be a standalone physical server, a server cluster composed of multiple physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal device can be a smartphone, tablet, laptop, PDA, personal computer, smart TV, smartwatch, in-vehicle device, wearable device, etc., but is not limited to these. The terminal device and the server can be directly or indirectly connected via wired or wireless communication, which is not limited here. The number of servers and terminal devices is also not limited.

[0091] As disclosed in this application, the PCVR determination processing method can store sample data and parameters to be optimized on a blockchain.

[0092] Conversion rate (CVR) is a metric that measures the effectiveness of an advertisement based on cost per action (CPA). It represents the conversion rate from a user clicking on an ad on a device or advertising screen to becoming an active, registered, or even paying user. Devices or advertising screens can report user actions related to the ad to the server, such as whether they registered or made a payment. The server records this information for each ad in a database, facilitating subsequent statistical analysis.

[0093] Given that this application involves some technical terms, these terms will be introduced below for ease of understanding.

[0094] (1) Advertising stage: Please refer to Figure 2 , Figure 2 This is a schematic diagram of the advertising stage in an embodiment of this application. As shown in the figure, the entire advertising process can be summarized into four stages: exposure, click, conversion, and payment.

[0095] (2) Billing Point: This refers to the billing method used by the platform. If the billing point is at the impression, the platform charges the advertiser based on the number of times the ad is displayed. If the billing point is at the click, the platform charges the advertiser based on the number of clicks on the ad, and so on.

[0096] (3) Bidding Point: This refers to how the advertiser bids. If the bidding point is at the impression level, the advertiser bids based on the number of ad impressions. If the bidding point is at the click level, the advertiser bids based on the number of ad clicks. If both the billing point and the bidding point are focused on impressions, this type of advertising is a classic Cost Per Mille (CPM) ad. If both the billing point and the bidding point are focused on clicks, this type of advertising is a classic Cost Per Click (CPC) ad. If both the billing point and the bidding point are focused on conversions, this type of advertising is a CPA ad.

[0097] (4) CPA Advertising: Pay-per-conversion advertising, where conversion results include, but are not limited to, leaving sales leads in forms, downloading and installing applications, registering new users, claiming coupons, adding items to carts, and placing orders. While CPA advertising is advantageous, many platforms do not support this approach. This is because advertising conversion data relies on feedback from advertisers; if the CPA model is used, advertisers may not send or may send fewer conversion numbers. In such cases, more specialized methods are needed, such as optimized CPA (OCPA).

[0098] (5) OCPA: This is a comparative model where the billing point and the bid point are separated. The billing point is at the click, and the bid point is at the conversion. The advantage of this approach is that it utilizes data advantages to help advertisers bid on conversions (i.e., to estimate PCVR) while ensuring the platform's revenue.

[0099] (6) PCVR: Advertising conversion rate prediction, usually refers to the click conversion rate of an advertisement, that is, the probability that an advertisement will be converted after being clicked.

[0100] (7) PCVR calibration: refers to a post-processing adjustment of the output results of the PCVR prediction model. Targeted adjustments will be made for a specific industry, such as PCVR calibration for direct-to-consumer e-commerce.

[0101] (8) Industry Factors: This involves readjusting the estimated cost per mile (eCPM) of ads for a specific industry. This readjustment strategy can be based on a specific industry, adjusting PCVR accordingly, or it can be based on a specific demographic within that industry, enhancing the effectiveness for that demographic. In direct-to-consumer (DTC) e-commerce, industry factors mainly include PCVR compensation factors and high-conversion-user enhancement factors. The PCVR compensation factor is obtained by training a model in a specific domain to obtain the PCVR enhancement factor for that domain. The high-conversion-user enhancement factor identifies high-conversion-users and calculates enhancement factors for them, increasing their chances of seeing ads in that domain.

[0102] (9) Ad conversions: For CPC ads, conversions are the same as clicks. For CPM ads, conversions are the same as impressions. For CPA ads, conversions are the same as conversion results.

[0103] (10) Advertising costs: refers to the expenses incurred in placing advertisements, that is, the money that advertisers pay to the platform.

[0104] (11) Target CPA (target_cpa): refers to the cost per conversion of the target, that is, the cost that the advertiser will pay for each conversion as predetermined by the advertiser.

[0105] (12) Initial stage of advertising: This refers to the initial stage of advertising exposure at the beginning of the day when the exposure is insufficient. For example, the number of conversions of the advertisement is ≤2, or the cost of the advertisement is ≤2*target_cpa.

[0106] (13) Advertising maturity stage (i.e., the stage of sufficient advertising exposure): This refers to the stage where the advertising has moved beyond the initial stage and entered the advertising maturity stage. For example, the number of conversions for the advertising is greater than 2, and the cost of the advertising is greater than 2*target_cpa.

[0107] (14) Similar type of advertisements: refers to advertisements with the same advertiser, or advertisements with the same brand, or advertisements with the same product information, etc.

[0108] (15) Parameter optimization modeling: refers to establishing a parameter optimization model for the important parameters involved in the PCVR calibration strategy, starting from the data, to learn the appropriate parameters in real time, so as to achieve a better PCVR calibration effect.

[0109] Based on the above introduction, in this application, for the PCVR calibration strategy in stages with a high number of ad conversions, the calibration parameters of the ad can be calculated in the following way:

[0110] Formula 1

[0111] Here, calibration_rate represents the calibration parameter of the ad, history_PCVR_bias_factor represents the factor parameter of the ad, conversion_num represents the total number of conversions of the ad in the first period, smooth_base represents the smoothing parameter of the ad, and sum_PCVR represents the total PCVR of the ad in the first period.

[0112] The smoothing parameters for the displayed advertisement can be calculated as follows:

[0113] Formula 2

[0114] Here, `smooth_base` represents the smoothing parameter for the ad, `sum_conversion_yesterday` represents the total number of conversions in the second period, `H` represents the duration of the ad in the first period, `min()` represents taking the minimum value, `ceil()` represents rounding to positive infinity, and `e` represents the base of the natural logarithm function. P1, P2, P3, and P4 can be parameters to be optimized, or they can be manually set empirical values. For example, P1 can be set to 10, P2 to 5, P3 to 10, and P4 to 10. Based on this, if the total number of conversions (`sum_conversion_yesterday`) in the second period is 0, then the smoothing parameter (`smooth_base`) for the ad is 10.

[0115] Based on Formula 1, the advertising factor parameter (history_PCVR_bias_factor) is also introduced, and the advertising factor parameter can be calculated as follows:

[0116] Formula 3

[0117] Where history_PCVR_bias_factor represents the factor parameter of the ad, Conv_valid_p represents the number of conversions in the p-th consumption interval, PCVR_valid_p represents the PCVR of the p-th consumption interval, Conv_valid represents the number of conversions of similar ads, and PCVR_valid represents the PCVR of similar ads.

[0118] Therefore, the factors affecting advertising performance are related to Conv_valid_p, PCVR_valid_p, Conv_valid, and PCVR_valid. Conv_valid and PCVR_valid can be obtained based on daily data statistics of similar advertisements. Conv_valid_p and PCVR_valid_p, however, are related to the method of dividing the consumption interval; that is, dividing the ad into several consumption intervals, the threshold of each interval will affect the results of Conv_valid_p and PCVR_valid_p.

[0119] Suppose variables X1, X2, ..., XN divide the consumption range into (N+1) consumption intervals, namely cost≤X1*target_cpa, X1*target_cpa<cost≤X2*target_cpa, ..., cost>XN*target_cpa. In this way, more efficient results can be obtained by optimizing the values ​​of N and the variables X1, X2, ..., XN.

[0120] For ease of explanation, Formula 1 will be broken down and analyzed below.

[0121] The total conversions (conversion_num) of the ad in the first period refers to the total conversions for that day, which is a clearly defined statistical result. The total PCVR (sum_PCVR) of the ad in the first period refers to the sum of the PCVRs corresponding to all valid clicks for that day, which is also a clearly defined statistical result.

[0122] The following section will break down and analyze Formula 2.

[0123] The four parameters P1, P2, P3, and P4 in Formula 2 can influence the specific value of the smoothing parameter (smooth_base) and the calibration rate to some extent. Therefore, more effective results can be obtained by optimizing these four parameters. If an ad did not consume any data yesterday, meaning the total conversions (sum_conversion_yesterday) for that ad in the second period are 0, then the default value of the smoothing parameter (smooth_base) (i.e., the default correlation parameter (default_smooth)) is also a quantity that can be optimized; optimizing this value can also yield more effective results.

[0124] In summary, the calculation of the calibration rate is related to the set of parameters to be optimized, which may include P1, P2, P3, P4, default_smooth, N value, and one or more of X1, X2, ..., XN. If these parameters are determined empirically, it is difficult to ensure their accuracy, and it is impossible for humans to frequently change these parameters in real time. Therefore, the conversion rate prediction method provided in this application can train and continuously optimize these parameters, ensuring they remain reasonable over time. This application builds a model from a data perspective to solve for suitable parameters, which is closer to real-world data and yields more accurate parameters compared to empirical values. Furthermore, the model can be trained daily, obtaining the latest version of the parameters calculated from the latest data each day, making it more real-time.

[0125] Based on the above introduction, the method for determining the conversion rate estimate in this application will be described below. Please refer to [link / reference needed]. Figure 3 One embodiment of the PCVR determination method in this application includes:

[0126] 101. Obtain the set of parameters to be optimized, the target cost per action (CPA) value, and the sample data set of K ads. The sample data set includes sample data for each ad, and the sample data for each ad includes the actual total number of conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period, and K is an integer greater than 1.

[0127] In one or more embodiments, the PCVR determining device needs to acquire a set of parameters to be optimized, a target CPA, and a sample data set of K advertisements. The set of parameters to be optimized includes optimizable parameters used when calculating PCVR. The target CPA is a pre-set target cost per conversion. The sample data set includes sample data corresponding to each of the K advertisements, where K is an integer greater than 1.

[0128] Specifically, the sample data for the i-th ad includes the actual total conversions (sumConvi), the PCVR before the j-th effective click calibration (PCVRj), the total conversions in the first period (conversion_num), the total PCVR in the first period (sum_PCVR), the total conversions in the second period (sum_conversion_yesterday), the duration (H) in the first period, and the cost. Here, j is an integer greater than or equal to 1, and the sample data includes the PCVR before each effective click calibration. The second period is the period preceding the first period. For example, if the first period is today (e.g., May 20, 2021), then the second period is yesterday, i.e., May 19, 2021. The duration (H) of the ad in the first period can be the duration of the ad on that day, in hours. For example, if the current time is 8:05, then the duration (H) of the ad in the first period is 8.

[0129] For example, the total conversions (conversion_num) in the first period are the total conversions for the current day. For instance, if the current time is 3:00 PM on May 20, 2021, then the total conversions (conversion_num) in the first period are the sum of the conversions from 00:00 AM on May 20, 2021 to 3:00 PM on May 20, 2021.

[0130] For example, the total PCVR (sum_PCVR) in the first period is the total number of conversions on that day. For instance, if the current time is 3:00 PM on May 20, 2021, then the total PCVR (sum_PCVR) in the first period is the sum of the PCVR from 00:00 AM on May 20, 2021 to 3:00 PM on May 20, 2021.

[0131] It should be noted that the PCVR determination device can be deployed on a computer device, which can be a server or a terminal device, without limitation here.

[0132] 102. Based on the set of parameters to be optimized and the target CPA, and according to the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, determine the sum of the PCVR of the effective clicks for each ad.

[0133] In one or more embodiments, the set of parameters to be optimized includes at least one parameter to be optimized, and the initial values ​​of these parameters can be set based on human experience. Based on this, according to the set of parameters to be optimized and the target CPA, combined with the total conversions (conversion_num) in the first period, the total PCVR (sum_PCVR) in the first period, the total conversions (sum_conversion_yesterday) in the second period, the duration (H) in the first period, the cost, and the PCVR before each click calibration (PCVRj), the sum of the PCVR of the effective clicks for each ad is determined.

[0134] Specifically, combining Equations 1, 2, and 3 above, the following model is constructed:

[0135] Formula 4

[0136] Formula 5

[0137] Formula 6

[0138] Formula 7

[0139] Formula 8

[0140] Where K represents the total number of ads, sumConvi represents the actual total conversions of the i-th ad, sumPCVRi represents the sum of PCVRs of valid clicks for the i-th ad, calibration_rate represents the calibration parameters for the i-th ad, history_PCVR_bias_factor represents the factor parameters for the i-th ad, conversion_num represents the total conversions of the i-th ad in the first period, smooth_base represents the smoothing parameters for the i-th ad, sum_PCVR represents the total PCVR of the i-th ad in the first period, sum_conversion_yesterday represents the total conversions of the i-th ad in the second period, H represents the duration of the i-th ad in the first period, default_smooth represents the default association parameters, Conv_valid_p represents the conversions in the p-th consumption interval, PCVR_valid_p represents the PCVR in the p-th consumption interval, Conv_valid represents the conversions of similar ads, and PCVR_valid represents the PCVR of similar ads.

[0141] As described above, Conv_valid_p and PCVR_valid_p are related to the method of dividing the consumption interval. For clarity, please refer to Table 1, which illustrates the method of dividing the consumption interval.

[0142] Table 1

[0143] Class Consumption range The conversion numbers and PCVR obtained from statistics 1 cost≤X1*target_cpa Conv_category_1PCVR_category_1 2 X1*target_cpa<cost≤X2*target_cpa Conv_category_2PCVR_category_2 3 X2*target_cpa<cost≤X3*target_cpa Conv_category_3PCVR_category_3 … … … N+1 cost > XN * target_cpa Conv_category_(N+1)PCVR_category_(N+1)

[0144] Here, "category" can be advertiser information, product information, or brand information. "target_cpa" represents the target CPA.

[0145] 103. Determine the average difference for K ads based on the sum of PCVR of valid clicks for each ad and the actual total conversions;

[0146] In one or more embodiments, Equation 4 represents the overall optimization objective, aiming to make the sum of the PCVR of effective clicks for the K ads as close as possible to the actual total conversions. Based on this, combining Equations 4 to 8, and the consumption interval division method shown in Table 1, the following average difference for the K ads is obtained:

[0147] ;

[0148] Where E represents the average difference for K advertisements.

[0149] 104. With the goal of minimizing the average difference, train the set of parameters to be optimized to obtain the target parameter set, wherein the target parameter set includes at least one optimized parameter;

[0150] In one or more embodiments, a set of parameters to be optimized is trained with the objective of minimizing the average difference (E). The set of parameters to be optimized may include one or more of P1, P2, P3, P4, default_smooth, N, and X1, X2, ..., XN. After training, a target set of parameters is obtained, which includes at least one optimized parameter.

[0151] It should be noted that the methods for solving the set of parameters to be optimized include, but are not limited to, solving a single formula, or using a genetic algorithm, or using a particle swarm optimization algorithm, or using an ant colony optimization algorithm, etc.

[0152] It is understandable that, taking advertising in the direct-to-consumer e-commerce sector as an example, the sample dataset used is specific to the direct-to-consumer e-commerce sector, and the optimized target parameter set is applicable to advertising in this sector. In practical applications, data from other sectors can also be used for training; this is merely an illustration and should not be construed as a limitation of this application. Furthermore, the sample dataset used for training is typically data from within the last 5 days. If the data spans too long, its reliability is lower, which may reduce the accuracy of the target parameter set.

[0153] 105. Based on the target parameter set, determine the calibrated PCVR of the target ad according to the associated data of the target ad and the PCVR to be adjusted. The associated data of the target ad includes the total number of conversions of the target ad in the third period, the total PCVR of the target ad in the third period, the total number of conversions of the target ad in the fourth period, the duration of the target ad in the third period, and the ad spend value. The fourth period is the period preceding the third period.

[0154] In one or more embodiments, after obtaining the target parameter set, PCVR calibration can be performed on target advertisements in a certain domain (e.g., direct-to-consumer e-commerce domain).

[0155] Specifically, after training, the target parameter set is assumed to include N=4, X1=3.75, X2=10.21, X3=21.13, X4=29.71, P1=8.7, P2=6.65, P3=11.77, P4=10.95, and default_smooth=15.3. Based on this, these optimized parameters are substituted into Equation 7 and Table 1 above. Then, the total conversions of the target ad in the third period (e.g., the current total conversions of the day), the total PCVR of the target ad in the third period (e.g., the current total PCVR of the day), the total conversions of the target ad in the fourth period (e.g., the total conversions of yesterday), the duration of the target ad in the third period, and the cost of the target ad are substituted into Equations 6, 7, and 8, respectively, to obtain the calibration parameters. Finally, the PCVR to be adjusted is calculated using the calibration parameters to obtain the calibrated PCVR of the target ad.

[0156] It is understandable that the fourth cycle is the cycle preceding the third cycle. Taking the third cycle as the current day (for example, May 22, 2021), the fourth cycle is yesterday, that is, May 21, 2021.

[0157] This application provides a method for determining conversion rate prediction. In this method, when calibrating the PCVR of a target advertisement, an optimized set of target parameters is used for calculation. Each optimized parameter in the target parameter set is trained based on sample data from multiple advertisements. Therefore, not only is more data introduced for parameter optimization, but it also has higher reliability compared to human experience values, better describing the actual situation and thus improving the accuracy of PCVR.

[0158] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, the sum of the PCVR of the effective clicks for each ad is determined, which may specifically include:

[0159] Based on the consumption segmentation parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined. The consumption segmentation parameters included in the consumption segmentation parameter set are pre-set.

[0160] Based on the set of associated parameters, the smoothing parameters for each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period.

[0161] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0162] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0163] The goal is to minimize the average difference. The set of parameters to be optimized is trained to obtain the target set of parameters, which may include:

[0164] The set of associated parameters is trained to obtain the target parameter set by minimizing the average difference.

[0165] In one or more embodiments, a method for training a set of parameters to be optimized is described. It is assumed that a set of consumption partitioning parameters has been pre-set, for example, the consumption partitioning parameter set includes N=4, X1=4, X2=10, X3=20, X4=40.

[0166] For specific details, please refer to Table 2, which is a schematic diagram of the consumption interval division method.

[0167] Table 2

[0168] Class Consumption range The conversion numbers and PCVR obtained from statistics 1 cost≤4*target_cpa Conv_category_1PCVR_category_1 2 4*target_cpa<cost≤10*target_cpa Conv_category_2PCVR_category_2 3 10*target_cpa<cost≤20*target_cpa Conv_category_3PCVR_category_3 4 20*target_cpa<cost≤40*target_cpa Conv_category_4PCVR_category_4 5 cost > 40 * target_cpa Conv_category_5PCVR_category_5

[0169] Here, "category" can be advertiser information, product information, or brand information. "target_cpa" represents the target CPA. As mentioned above, "Conv_valid_p" and "PCVR_valid_p" are related to the way the spend interval is divided. Based on the spend interval corresponding to the spend value of each advertisement, and combined with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated.

[0170] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period corresponding to each advertisement and the total number of conversions (sum_conversion_yesterday) in the second period. It should be noted that P1, P2, P3, P4 and the default associated parameter (default_smooth) in Equation 7 are all parameters in the associated parameter set.

[0171] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0172] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0173] Based on this, with the goal of minimizing the average difference (E), the set of association parameters is trained to obtain the target parameter set, which includes the optimized P1, P2, P3, P4 and the default association parameter (default_smooth).

[0174] It should be noted that this application incorporates factor parameters into the calculation of calibration parameters. In addition, other methods using similar advertising information, such as weighted summation, can also be applied to the calculation of calibration parameters, which is not limited here.

[0175] Secondly, in this embodiment of the application, a method for training a set of parameters to be optimized is provided. In this method, the parameter set is divided and the associated parameter set is optimized with a fixed amount of consumption. Thus, while saving the amount of parameter training, the parameters in the associated parameter set can be optimized, thereby improving the training efficiency and the accuracy of the calibration parameters.

[0176] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the set of associated parameters includes a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter;

[0177] Based on the set of associated parameters, and according to the duration of each ad in the first period and the total number of conversions in the second period, the smoothing parameters for each ad are determined. Specifically, these parameters may include:

[0178] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are not zero, then based on the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter, the smoothing parameter corresponding to any one ad is determined according to the duration in the first period and the total conversions in the second period corresponding to any one ad.

[0179] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter is determined to be the smoothing parameter corresponding to any one ad, where the default association parameter is preset.

[0180] The goal is to minimize the average difference. The set of associated parameters is trained to obtain the target parameter set, which may specifically include:

[0181] With the goal of minimizing the average difference, the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter are trained to obtain the target parameter set, which includes the optimized first correlation parameter, the optimized second correlation parameter, the optimized third correlation parameter, and the optimized fourth correlation parameter.

[0182] In one or more embodiments, another method for training the set of parameters to be optimized is described. It is assumed that a set of consumption partitioning parameters and a default association parameter (default_smooth) have been pre-set. For example, the consumption partitioning parameter set includes N=4, X1=4, X2=10, X3=20, X4=40, and the default association parameter is default_smooth=10.

[0183] Specifically, for clarity, please refer to Table 2 again. As mentioned above, Conv_valid_p and PCVR_valid_p are related to the way the spend interval is divided. Based on the spend interval where the spend value corresponding to each ad falls, and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each ad is calculated.

[0184] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period and the total conversions (sum_conversion_yesterday) of the second period. It should be noted that the first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), fourth correlation parameter (P4), and default correlation parameter (default_smooth) in Equation 7 all belong to the parameter set of correlation parameters. If the total conversions (sum_conversion_yesterday) of the advertisement in the second period are not 0, the smoothing parameter (smooth_base) is calculated based on the first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), fourth correlation parameter (P4), duration (H) of the first period, and total conversions (sum_conversion_yesterday) of the second period. If the total number of conversions (sum_conversion_yesterday) for the ad in the second period is 0, then the default associated parameter (default_smooth) will be directly set as the smoothing parameter (smooth_base) for the ad.

[0185] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0186] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0187] Based on this, with the goal of minimizing the average difference (E), the set of correlation parameters is trained to obtain the target parameter set, which includes the optimized first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), and fourth correlation parameter (P4).

[0188] Furthermore, this application provides another method for training the set of parameters to be optimized. By fixing the default associated parameters and optimizing other associated parameters, the training workload of parameters can be reduced while optimizing other associated parameters, thereby improving training efficiency and accuracy of calibration parameters.

[0189] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the set of associated parameters includes default associated parameters;

[0190] Based on the set of associated parameters, and according to the duration of each ad in the first period and the total number of conversions in the second period, the smoothing parameters for each ad are determined. Specifically, these parameters may include:

[0191] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are not zero, then based on the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter, the smoothing parameter corresponding to any one ad is determined according to the duration in the first period corresponding to any one ad and the total conversions in the second period. The first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter are preset.

[0192] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter will be set to the smoothing parameter corresponding to any one ad.

[0193] The goal is to minimize the average difference. The set of associated parameters is trained to obtain the target parameter set, which may specifically include:

[0194] With the goal of minimizing the average difference, the default association parameters are trained to obtain the target parameter set, which includes the optimized default association parameters.

[0195] In one or more embodiments, another method for training the set of parameters to be optimized is described. Assume that a consumption partitioning parameter set, a first correlation parameter (P1), a second correlation parameter (P2), a third correlation parameter (P3), and a fourth correlation parameter (P4) have been pre-set. For example, the consumption partitioning parameter set includes N=4, X1=4, X2=10, X3=20, X4=40, and the first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), and fourth correlation parameter (P4) are P1=10, P2=5, P3=10, and P4=10, respectively.

[0196] Specifically, for clarity, please refer to Table 2 again. As mentioned above, Conv_valid_p and PCVR_valid_p are related to the way the spend interval is divided. Based on the spend interval where the spend value corresponding to each ad falls, and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each ad is calculated.

[0197] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period and the total conversions (sum_conversion_yesterday) of the second period. It should be noted that the first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), fourth correlation parameter (P4), and default correlation parameter (default_smooth) in Equation 7 all belong to the parameter set of correlation parameters. If the total conversions (sum_conversion_yesterday) of the advertisement in the second period are not 0, the smoothing parameter (smooth_base) is calculated based on the first correlation parameter (P1), second correlation parameter (P2), third correlation parameter (P3), fourth correlation parameter (P4), duration (H) of the first period, and total conversions (sum_conversion_yesterday) of the second period. If the total number of conversions (sum_conversion_yesterday) for the ad in the second period is 0, then the default associated parameter (default_smooth) will be directly set as the smoothing parameter (smooth_base) for the ad.

[0198] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0199] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0200] Based on this, with the goal of minimizing the average difference (E), the set of association parameters is trained to obtain the target parameter set, which includes the optimized default association parameters (default_smooth).

[0201] Furthermore, in this embodiment of the application, another method is provided for training the set of parameters to be optimized. By fixing other related parameters and optimizing the default related parameters, the amount of parameter training can be saved while optimizing other related parameters, thereby improving training efficiency and accuracy of calibration parameters.

[0202] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, the sum of the PCVR of the effective clicks for each ad is determined, which may specifically include:

[0203] Based on the consumption volume, the parameter set and target CPA are divided, and the factor parameters corresponding to each advertisement are determined according to the consumption value corresponding to each advertisement.

[0204] Based on the set of associated parameters, the smoothing parameters corresponding to each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The associated parameters included in the set of associated parameters are pre-set.

[0205] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0206] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0207] The goal is to minimize the average difference. The set of parameters to be optimized is trained to obtain the target set of parameters, which may include:

[0208] With the goal of minimizing the average difference, the parameter set for consumption is divided and trained to obtain the target parameter set.

[0209] In one or more embodiments, another method for training the set of parameters to be optimized is described. It is assumed that a pre-set set of association parameters has been configured, including a first association parameter (P1), a second association parameter (P2), a third association parameter (P3), a fourth association parameter (P4), and a default association parameter, for example, P1=10, P2=5, P3=10, P4=10, and default_smooth=10.

[0210] Specifically, for clarity, please refer to Table 1 again. As mentioned above, Conv_valid_p and PCVR_valid_p are related to the method of dividing the consumption interval. Based on the consumption interval where the consumption value corresponding to each advertisement falls, and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated. It should be noted that the N value and X1, X2, ..., XN in Table 1 are all parameters in the consumption division parameter set.

[0211] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period corresponding to each advertisement and the total number of conversions (sum_conversion_yesterday) in the second period.

[0212] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0213] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0214] Based on this, with the goal of minimizing the average difference (E), the set of associated parameters is trained to obtain the target parameter set, which includes the optimized N value and N consumption partitioning parameters (i.e., X1, X2, ..., XN).

[0215] Secondly, this application provides another way to train the set of parameters to be optimized. By fixing the set of associated parameters and optimizing the consumption of the parameter set, the training amount of parameters can be saved while optimizing the parameters in the set of associated parameters, thereby improving the training efficiency and the accuracy of the calibration parameters.

[0216] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the consumption division parameter set includes N consumption division parameters, where N is an integer greater than or equal to 1;

[0217] Based on the consumption volume, a parameter set and target CPA are defined. Then, based on the consumption value corresponding to each ad, the factor parameters for each ad are determined. Specifically, these may include:

[0218] Based on the N value, N consumption division parameters, and target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined, where the N value is preset;

[0219] The goal is to minimize the average difference. The set of associated parameters is trained to obtain the target parameter set, which may specifically include:

[0220] With the goal of minimizing the average difference, each of the N consumption partitioning parameters is trained to obtain the target parameter set, which includes the optimized N consumption partitioning parameters.

[0221] In one or more embodiments, another method for training the set of parameters to be optimized is described. It is assumed that a set of associated parameters and a value of N have been pre-set. The set of associated parameters includes a first associated parameter (P1), a second associated parameter (P2), a third associated parameter (P3), a fourth associated parameter (P4), and a default associated parameter, for example, P1=10, P2=5, P3=10, P4=10, default_smooth=10. The value of N can be set to N=4.

[0222] Specifically, for clarity, please refer to Table 1 again. As mentioned above, Conv_valid_p and PCVR_valid_p are related to the method of dividing the consumption interval. Based on the consumption interval where the consumption value corresponding to each advertisement falls, and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated. It should be noted that the N value and X1, X2, ..., XN in Table 1 are all parameters in the consumption division parameter set.

[0223] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period corresponding to each advertisement and the total number of conversions (sum_conversion_yesterday) in the second period.

[0224] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0225] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0226] Based on this, with the goal of minimizing the average difference (E), the set of associated parameters is trained to obtain the target parameter set, which includes the optimized N consumption partitioning parameters (i.e., X1, X2, ..., XN).

[0227] Furthermore, in this embodiment of the application, another method for training the set of parameters to be optimized is provided. By fixing the value of N and optimizing the consumption of the parameter division, the parameters in the associated parameter set can be optimized while saving the amount of parameter training, thereby improving the training efficiency and the accuracy of the calibration parameters.

[0228] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, the sum of the PCVR of the effective clicks for each ad is determined, which may specifically include:

[0229] Based on the set of associated parameters, the smoothing parameters corresponding to each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The set of associated parameters includes the default associated parameter, the first associated parameter, the second associated parameter, the third associated parameter, and the fourth associated parameter.

[0230] Based on the consumption-based parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined. The consumption-based parameter set includes an N value and N consumption-based parameters, where N is an integer greater than or equal to 1.

[0231] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0232] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0233] The goal is to minimize the average difference. The set of parameters to be optimized is trained to obtain the target set of parameters, which may include:

[0234] With the goal of minimizing the average difference, the set of association parameters is trained to obtain the target parameter set, which includes the optimized first association parameter, optimized second association parameter, optimized third association parameter, optimized fourth association parameter, optimized default association parameter, optimized N value, and optimized N consumption partitioning parameters.

[0235] In one or more embodiments, another method for training the set of parameters to be optimized is described. The set of parameters to be optimized includes a set of correlation parameters and a set of consumption partitioning parameters. The set of correlation parameters includes a first correlation parameter (P1), a second correlation parameter (P2), a third correlation parameter (P3), a fourth correlation parameter (P4), and a default correlation parameter. The set of consumption partitioning parameters includes a value of N and N consumption partitioning parameters (X1, X2, ..., XN).

[0236] Specifically, for clarity, please refer to Table 1 again. As mentioned above, Conv_valid_p and PCVR_valid_p are related to the way the spend interval is divided. Based on the spend interval where the spend value corresponding to each ad falls, and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each ad is calculated.

[0237] Combining Equation 7 above, the smoothing parameter (smooth_base) for each advertisement is determined based on the duration (H) of the first period corresponding to each advertisement and the total number of conversions (sum_conversion_yesterday) in the second period.

[0238] Combining Equation 6 above, the calibration parameter (calibration_rate) for each advertisement is determined based on the total number of conversions (conversion_num), the total PCVR (sum_PCVR) in the first period, the smoothing parameter (smooth_base), and the factor parameter (history_PCVR_bias_factor) for each advertisement.

[0239] Combining Equation 5 above, based on the PCVR before each click for each advertisement (e.g., the PCVR before the j-th click is PCVRj) and the calibration parameter (calibration_rate), determine the sum of the PCVR of the effective clicks for each advertisement (e.g., the sum of the PCVR of the effective clicks for the i-th advertisement is sumPCVRi).

[0240] Based on this, with the goal of minimizing the average difference (E), the set of association parameters is trained to obtain the target parameter set, which includes the optimized first association parameter, the optimized second association parameter, the optimized third association parameter, the optimized fourth association parameter, the optimized default association parameter, the optimized N value, and the optimized N consumption partitioning parameters.

[0241] Secondly, this application provides another method for training the set of parameters to be optimized. Through the above method, the fixed correlation parameter set and the consumption partitioning parameter set can be optimized, which helps to improve the accuracy of the calibration parameters, thereby better calibrating PCVR, which is beneficial to controlling advertising costs and improving the reliability and usability of OCPA products.

[0242] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the sample data for each advertisement further includes advertiser information, the total number of conversions counted per unit time, and the total PCVR counted per unit time;

[0243] Based on the consumption volume, a parameter set and target CPA are defined. Then, based on the consumption value corresponding to each ad, the factor parameters for each ad are determined. Specifically, these may include:

[0244] Based on the advertiser information corresponding to each advertisement, the K advertisements are divided into T1 advertisement sets, where each advertisement set includes at least one advertisement, and T1 is an integer greater than or equal to 1;

[0245] If the total spending value corresponding to each ad set in the T1 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set. The (N+1) spending intervals are determined based on the spending parameter set and the target CPA. Each ad subset includes at least one ad, and the same ad subset corresponds to the same spending interval. N is an integer greater than or equal to 1.

[0246] For each of at least one subset of advertisements, the first conversion and the first PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in each subset within a unit of time, wherein the current period includes at least one unit of time.

[0247] For each of the T1 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the second number of conversions and the second PCVR of each ad set in the current period.

[0248] Based on the first conversion count, the first PCVR, the second conversion count, and the second PCVR, determine the factor parameters corresponding to each advertisement.

[0249] In one or more embodiments, a method for determining factor parameters based on advertiser information is described. As described in the foregoing embodiments, based on the target CPA, the N value, and N consumption segmentation parameters (X1, X2, ..., XN), (N+1) consumption intervals can be defined. Before training begins, the consumption segmentation parameter set can be set based on human experience, for example, N=4, X1=4, X2=10, X3=20, X4=40. It is understood that during training, if the consumption segmentation parameter set needs to be optimized, the N value and the N consumption segmentation parameters (X1, X2, ..., XN) may change. The following description will use the parameters before optimization of the consumption segmentation parameter set as an example.

[0250] For clarity, please refer to Table 3, which is a schematic diagram of the consumption interval division method based on advertiser information.

[0251] Table 3

[0252] Class Consumption range The conversion numbers and PCVR obtained from statistics 1 cost≤4*target_cpa Conv_advertiser_1PCVR_advertiser_1 2 4*target_cpa<cost≤10*target_cpa Conv_advertiser_2PCVR_advertiser_2 3 10*target_cpa<cost≤20*target_cpa Conv_advertiser_3PCVR_advertiser_3 4 20*target_cpa<cost≤40*target_cpa Conv_advertiser_4PCVR_advertiser_4 5 cost > 40 * target_cpa Conv_advertiser_5PCVR_advertiser_5

[0253] Here, "advertiser" refers to the advertiser's information. "target_cpa" represents the target CPA. As mentioned above, based on the spending range corresponding to each advertisement's spend value, and combined with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated.

[0254] Specifically, firstly, based on the advertiser information corresponding to each ad, the K ads are divided into T1 ad sets. That is, data is collected under different dimensions for different advertiser information to obtain T1 ad sets. Assuming there are 50 ads from 5 advertisers, these 50 ads are first divided, grouping ads with the same advertiser information into one category. Based on this, each group of ads is further divided. For example, each group of ads has site sets 27 and 28, so data is collected under different spending levels for the four combined dimensions: [27, new ad], [27, old ad], [28, new ad], and [28, old ad]. At this point, 20 ad sets are obtained.

[0255] Next, for each ad set, the total cost of each ad set is calculated. If the total cost of each ad set is greater than or equal to the cost threshold (e.g., 4 * target_cpa), then at least one ad subset is generated based on (N+1) cost intervals and the cost of each ad in each ad set. That is, the cost interval for each ad is determined based on its cost, and the ad subsets included in each ad set are calculated, thus obtaining at least one ad subset, with each ad subset corresponding to the same cost interval.

[0256] For example, in one implementation, for each ad set, the first conversion number and the first PCVR for each ad subset in the current period are calculated as follows:

[0257] Formula 9

[0258] Formula 10

[0259] For a single advertisement, I satisfies `Conv_advertiser_p` represents the first conversion of the ad subset within the p-th spending tier in the current period (e.g., today). `PCVR_advertiser_p` represents the first PCVR of the ad subset within the p-th spending tier in the current period (e.g., today). `I` indicates that, up to the current hour, the cost of a single ad is within the corresponding spending tier (e.g., cost ≤ 4 * target_cpa). `lambda` represents the time decay factor (e.g., 0.05). `Conv_advertiser_hour` i This represents the total number of conversions counted within a unit of time (e.g., the i-th hour). PCVR_advertiser_hour i This represents the total PCVR counted per unit time (e.g., the i-th hour).

[0260] For example, in another implementation, the total number of ad conversions for the day and the most recent hour can be further calculated, as well as the sum of PCVR corresponding to the total number of valid clicks. For instance, the data for different spending tiers under the four combined dimensions [27, new ads], [27, old ads], [28, new ads], and [28, old ads] can be calculated for the entire day and the most recent hour. When calculating the data for the entire day, a relatively intuitive time decay strategy is used, namely:

[0261] Formula 11

[0262] Formula 12

[0263] Wherein, Conv_advertiser_day represents the number of conversions obtained on that day, and PCVR_advertiser_day represents the sum of the total PCVR obtained on that day.

[0264] If sufficient data is available for the current hour, then the data for the current hour will be used; otherwise, the data for the entire day will be used.

[0265] Formula 13

[0266] Formula 14

[0267] It should be noted that Conv_advertiser_1 is the sum of conversions for all ads under the same advertiser when the cost of a single ad is ≤ 4 * target_cpa. PCVR_advertiser_1 is the sum of PCVRs for valid clicks of all ads under the same advertiser when the cost of a single ad is ≤ 4 * target_cpa. The statistical methods for other tiers are similar and will not be elaborated here.

[0268] For each ad set, based on the total conversions and total PCVR counts for each ad within a unit of time, determine the second conversion count and the second PCVR for the current period. Specifically, calculate the total conversion count for the ads in the current period (e.g., today) under the dimension of [site set, whether new or old ads], which is the second conversion count (Conv_advertiser), and calculate the sum of the PCVRs corresponding to the total valid clicks in the current period (e.g., today) under the dimension of [site set, whether new or old ads], which is the second PCVR (PCVR_advertiser).

[0269] Finally, the first conversion number (Conv_advertiser_p) corresponding to each advertisement is taken as Conv_valid_p, the first PCVR (PCVR_advertiser_p) is taken as PCVR_valid_p, the second conversion number (Conv_advertiser) is taken as Conv_valid, the second PCVR (PCVR_advertiser) is taken as PCVR_valid, and combined with Equation 8 above, the factor parameters corresponding to each advertisement are calculated.

[0270] It is understandable that the dimensions used in this application can also include attribution, audience, and content, and these dimensions can be easily expanded, for example, to include the same group, the same region, and the same targeting. This application considers the site set and the dimensions of whether an ad is new or old. The strategy for defining new and old ads is that only ads exposed on the current day that have not been exposed before are considered new ads. This definition of new ads can be expanded. For example, only ads exposed in the current two days are considered new ads. A time decay coefficient is considered when calculating and statistically analyzing the data for the current day. This application uses a linear time decay strategy; in addition, other time decay strategies (such as Gaussian decay) can also be used.

[0271] Furthermore, in this application embodiment, a method for determining factor parameters based on advertiser information is provided. Through the above method, advertisements with the same advertiser information are classified, and division and calculation are performed based on similar advertisements, thereby introducing more data for optimizing parameters.

[0272] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the sample data of each advertisement also includes brand information;

[0273] It may also include:

[0274] If any one of the T1 ad sets has a total spending value less than the spending threshold, then based on the brand information corresponding to each ad, the K ads are divided into T2 ad sets, where each ad set includes at least one ad and T2 is an integer greater than or equal to 1.

[0275] If the total spending value corresponding to each ad set in the T2 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0276] For each of at least one subset of advertisements, the third conversion and the third PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in the current period.

[0277] For each of the T2 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the fourth number of conversions and the fourth PCVR of each ad set in the current period.

[0278] Based on the third conversion count, third PCVR, fourth conversion count, and fourth PCVR, determine the factor parameters corresponding to each advertisement.

[0279] In one or more embodiments, a method for determining factor parameters based on brand information is introduced. As described in the foregoing embodiments, if the division is based on advertiser information, and there exists an ad set whose total ad spend is greater than or equal to the ad spend threshold, then K ads are divided into T2 ad sets based on brand information. Similarly, based on the target CPA, the N value, and N spending parameters (X1, X2, ..., XN), (N+1) spending intervals can be defined.

[0280] For clarity, please refer to Table 4, which is a schematic diagram of the consumption range division method based on brand information.

[0281] Table 4

[0282] Class Consumption range The conversion numbers and PCVR obtained from statistics 1 cost≤4*target_cpa Conv_brand_1PCVR_brand_1 2 4*target_cpa<cost≤10*target_cpa Conv_brand_2PCVR_brand_2 3 10*target_cpa<cost≤20*target_cpa Conv_brand_3PCVR_brand_3 4 20*target_cpa<cost≤40*target_cpa Conv_brand_4PCVR_brand_4 5 cost > 40 * target_cpa Conv_brand_5PCVR_brand_5

[0283] Here, "brand" represents brand information. "target_cpa" indicates the target CPA. As mentioned above, based on the spending range corresponding to the spending value of each advertisement and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated.

[0284] Specifically, firstly, based on the brand information corresponding to each advertisement, the K advertisements are divided into T2 ad sets. That is, data is collected under different dimensions for different brand information to obtain T2 ad sets. Assuming there are 50 advertisements from 5 brands, these 50 advertisements are first divided, grouping those with the same brand information into one category. Based on this, each group of advertisements is further divided. For example, each group of advertisements has site sets 27 and 28, so data is collected under different spending levels for the four combined dimensions: [27, new ad], [27, old ad], [28, new ad], and [28, old ad]. At this point, 20 ad sets are obtained.

[0285] Next, for each ad set, the total cost of each ad set is calculated. If the total cost of each ad set is greater than or equal to the cost threshold (e.g., 4 * target_cpa), then at least one ad subset is generated based on (N+1) cost intervals and the cost of each ad in each ad set. That is, the cost interval for each ad is determined based on its cost, and the ad subsets included in each ad set are calculated, thus obtaining at least one ad subset, with each ad subset corresponding to the same cost interval.

[0286] For example, in one implementation, for each ad set, the third conversion number and the third PCVR for each ad subset in the current period are calculated as follows:

[0287] Formula 15

[0288] Formula 16

[0289] For a single advertisement, I satisfies `Conv_brand_p` represents the third conversion of the ad subset within the p-th spend range in the current period (e.g., today). `PCVR_brand_p` represents the third PCVR of the ad subset within the p-th spend range in the current period (e.g., today). `I` indicates that, up to the current hour, the cost of a single ad is within the corresponding spend range (e.g., cost ≤ 4 * target_cpa). `lambda` represents the time decay factor (e.g., 0.05). `Conv_brand_hour` i This represents the total number of conversions counted within a unit of time (e.g., the i-th hour). PCVR_brand_hour i This represents the total PCVR counted per unit time (e.g., the i-th hour).

[0290] For example, in another implementation, the total number of ad conversions for the day and the most recent hour can be further calculated, as well as the sum of PCVR corresponding to the total number of valid clicks. For instance, the data for different spending tiers under the four combined dimensions [27, new ads], [27, old ads], [28, new ads], and [28, old ads] can be calculated for the entire day and the most recent hour. When calculating the data for the entire day, a relatively intuitive time decay strategy is used, namely:

[0291] Formula 17

[0292] Formula 18

[0293] Wherein, Conv_brand_day represents the number of conversions obtained on that day, and PCVR_brand_day represents the total PCVR obtained on that day.

[0294] If sufficient data is available for the current hour, then the data for the current hour will be used; otherwise, the data for the entire day will be used.

[0295] Formula 19

[0296] Formula 20

[0297] It should be noted that Conv_brand_1 is the sum of conversions for all ads under the same brand information when the cost of a single ad is ≤ 4 * target_cpa. PCVR_brand_1 is the sum of PCVR corresponding to valid clicks for all ads under the same brand information when the cost of a single ad is ≤ 4 * target_cpa. The statistical methods for other tiers are similar and will not be elaborated here.

[0298] For each ad set, based on the total conversions and total PCVR counts for each ad within a unit of time, determine the second conversion count and the fourth PCVR for each ad set in the current period. Specifically, calculate the total conversion count for the ads in the current period (e.g., today) under the dimension of [site set, whether new or old ad], which is the second conversion count (Conv_brand), and calculate the sum of the PCVRs corresponding to the total valid clicks in the current period (e.g., today) under the dimension of [site set, whether new or old ad], which is the fourth PCVR (PCVR_brand).

[0299] Finally, the third conversion number (Conv_brand_p) corresponding to each advertisement is taken as Conv_valid_p, the third PCVR (PCVR_brand_p) is taken as PCVR_valid_p, the fourth conversion number (Conv_brand) is taken as Conv_valid, the fourth PCVR (PCVR_brand) is taken as PCVR_valid, and combined with Equation 8 above, the factor parameters corresponding to each advertisement are calculated.

[0300] Furthermore, in this embodiment of the application, a method for determining factor parameters based on brand information is provided. Through the above method, advertisements with the same brand information are classified, and division and calculation are performed based on similar advertisements, thereby introducing more data for optimizing parameters.

[0301] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the sample data for each advertisement also includes product information;

[0302] It may also include:

[0303] If any one of the T2 ad sets has a total spending value less than the spending threshold, then based on the product information corresponding to each ad, the K ads are divided into T3 ad sets, where each ad set includes at least one ad and T3 is an integer greater than or equal to 1.

[0304] If the total spending value corresponding to each of the T3 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0305] For each ad subset in at least one ad subset, determine the fifth conversion number and the fifth PCVR of each ad subset in the current period based on the total conversion number and the total PCVR count of each ad in each ad subset within the unit time.

[0306] For each of the T3 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the sixth conversion and the sixth PCVR of each ad set in the current period.

[0307] Based on the fifth conversion count, fifth PCVR, sixth conversion count, and sixth PCVR, determine the factor parameters corresponding to each advertisement.

[0308] In one or more embodiments, a method for determining factor parameters based on product information is introduced. As described in the foregoing embodiments, if the division is based on brand information, and there exists a total ad spend value corresponding to a brand information that is greater than or equal to the ad spend threshold, then K ads are divided into T3 ad sets based on product information. Similarly, based on the target CPA, the N value, and N spending parameters (X1, X2, ..., XN), (N+1) spending intervals can be defined.

[0309] For clarity, please refer to Table 5, which is a schematic diagram of the consumption interval division method based on product information.

[0310] Table 5

[0311] Class Consumption range The conversion numbers and PCVR obtained from statistics 1 cost≤4*target_cpa Conv_product_1PCVR_product_1 2 4*target_cpa<cost≤10*target_cpa Conv_product_2PCVR_product_2 3 10*target_cpa<cost≤20*target_cpa Conv_product_3PCVR_product_3 4 20*target_cpa<cost≤40*target_cpa Conv_product_4PCVR_product_4 5 cost > 40 * target_cpa Conv_product_5PCVR_product_5

[0312] Here, "product" represents product information. "target_cpa" indicates the target CPA. As mentioned above, based on the spending range corresponding to the spending value of each advertisement and in conjunction with Equation 8 above, the factor parameter (history_PCVR_bias_factor) corresponding to each advertisement is calculated.

[0313] Specifically, firstly, based on the product information corresponding to each advertisement, the K advertisements are divided into T3 ad sets. That is, data is collected under different dimensions for different product information to obtain T3 ad sets. Assuming there are 50 advertisements and 5 product categories, these 50 advertisements are first divided, grouping those with the same product information into one category. Based on this, each group of advertisements is further divided. For example, each group of advertisements has site sets 27 and 28, so data is collected under different consumption levels for the four combined dimensions: [27, new advertisement], [27, old advertisement], [28, new advertisement], and [28, old advertisement]. At this point, 20 ad sets are obtained.

[0314] Next, for each ad set, the total cost of each ad set is calculated. If the total cost of each ad set is greater than or equal to the cost threshold (e.g., 4 * target_cpa), then at least one ad subset is generated based on (N+1) cost intervals and the cost of each ad in each ad set. That is, the cost interval for each ad is determined based on its cost, and the ad subsets included in each ad set are calculated, thus obtaining at least one ad subset, with each ad subset corresponding to the same cost interval.

[0315] For example, in one implementation, for each ad set, the fifth conversion number and the fifth PCVR for each ad subset in the current period are calculated as follows:

[0316] Formula 21

[0317] Formula 22

[0318] For a single advertisement, I satisfies `Conv_product_p` represents the fifth conversion of the ad subset within the p-th spending tier in the current period (e.g., today). `PCVR_product_p` represents the fifth PCVR of the ad subset within the current period (e.g., today). `I` indicates that, up to the current hour, the cost of a single ad is within the corresponding spending tier (e.g., cost ≤ 4 * target_cpa). `lambda` represents the time decay factor (e.g., 0.05). `Conv_product_hour` i This represents the total number of conversions counted within a unit of time (e.g., the i-th hour). PCVR_product_hour i This represents the total PCVR counted per unit time (e.g., the i-th hour).

[0319] For example, in another implementation, the total number of ad conversions for the day and the most recent hour can be further calculated, as well as the sum of PCVR corresponding to the total number of valid clicks. For instance, the data for different spending tiers under the four combined dimensions [27, new ads], [27, old ads], [28, new ads], and [28, old ads] can be calculated for the entire day and the most recent hour. When calculating the data for the entire day, a relatively intuitive time decay strategy is used, namely:

[0320] Formula 23

[0321] Formula 24

[0322] Wherein, Conv_product_day represents the number of conversions obtained on that day, and PCVR_product_day represents the total PCVR obtained on that day.

[0323] If sufficient data is available for the current hour, then the data for the current hour will be used; otherwise, the data for the entire day will be used.

[0324] Formula 25

[0325] Formula 26

[0326] It should be noted that Conv_product_1 is the sum of conversions for all ads under the same product information when the cost of a single ad is ≤ 4 * target_cpa. PCVR_product_1 is the sum of PCVR corresponding to valid clicks for all ads under the same product information when the cost of a single ad is ≤ 4 * target_cpa. The statistical methods for other tiers are similar and will not be elaborated here.

[0327] For each ad set, based on the total conversions and total PCVR counts for each ad within a unit of time, determine the second conversion count and the sixth PCVR for each ad set in the current period. Specifically, calculate the total conversion count for the ads in the current period (e.g., today) under the dimension of [site set, whether new or old ads], i.e., the second conversion count (Conv_product), and calculate the sum of the PCVRs corresponding to the total valid clicks in the current period (e.g., today) under the dimension of [site set, whether new or old ads], i.e., the sixth PCVR (PCVR_product).

[0328] Finally, the fifth conversion number (Conv_product_p) corresponding to each advertisement is taken as Conv_valid_p, the fifth PCVR (PCVR_product_p) is taken as PCVR_valid_p, the sixth conversion number (Conv_product) is taken as Conv_valid, the sixth PCVR (PCVR_product) is taken as PCVR_valid, and combined with Equation 8 above, the factor parameters corresponding to each advertisement are calculated.

[0329] Furthermore, in this application embodiment, a method for determining factor parameters based on product information is provided. Through the above method, advertisements with the same product information are classified, and division and calculation are performed based on similar advertisements, thereby introducing more data for optimizing parameters.

[0330] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment provided by this application, the calibration parameters corresponding to each advertisement are determined according to the total conversions in the first period, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Specifically, this may include:

[0331] For any given ad, determine the target CPA range for the ad's spend value;

[0332] Based on the target CPA range, the target expansion coefficient is determined according to the ad spend value and the total conversions in the first period.

[0333] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters of the advertisement in the first period, determine the calibration parameters to be adjusted for the advertisement.

[0334] The calibration parameters corresponding to the advertisement are determined based on the calibration parameters to be adjusted and the target magnification factor.

[0335] In one or more embodiments, a method for adjusting calibration parameters using an amplification factor during the ad maturation stage is described. For ease of explanation, the following description focuses on any single ad; it is understood that the processing method for other ads is similar and will not be elaborated upon here.

[0336] Specifically, the ad spend is pre-divided into several CPA ranges. This application uses three CPA ranges as an example: one range is below 8 times the target_cpa; another range is between 8 and 25 times the target_cpa; and the third range is above 25 times the target_cpa. Thus, the target CPA range is determined based on the ad's cost.

[0337] For example, if the target CPA range is less than 8 times target_cpa, then PCVR is multiplied by a factor of 1.0, which is equivalent to not multiplying by a factor. This is because advertising spending is lower and the overall volume is smaller at this stage, making it more susceptible to interference and deviations, ultimately leading to cost overruns. Therefore, the factor does not need to be increased.

[0338] For example, if the target CPA range is between 8x target_cpa and 25x target_cpa, the coefficient multiplied by PCVR can be obtained by the following formula:

[0339] Formula 27

[0340] Where coef represents the target scaling factor, cost represents the ad spend, target_cpa represents the target CPA, and current_cpa_bias represents the current CPA bias of the ad. The current CPA bias of the ad can be calculated as follows:

[0341] current_cpa_bias=cost / (conv_num*target_cpa)–1; Formula 28

[0342] Where current_cpa_bias represents the current CPA bias of the ad, cost represents the ad spend (i.e., the current spend of the ad), and conv_num represents the total number of conversions of the ad in the first period (i.e., the current number of conversions of the ad).

[0343] For example, if the target CPA range is a CPA range greater than 25 times target_cpa, the coefficient multiplied by PCVR can be obtained by the following formula:

[0344] Formula 29

[0345] Wherein, coef represents the target scaling factor, cost represents the ad spend, target_cpa represents the target CPA, and current_cpa_bias represents the current CPA bias of the ad. The calculation method for the current CPA bias of the ad can be found in Equation 28, and will not be elaborated here.

[0346] Based on this, the calibration parameters to be adjusted are calculated as follows:

[0347] Formula 30

[0348] in, The parameters to be adjusted for the advertisement are: history_PCVR_bias_factor, conversion_num, smooth_base, and sum_PCVR.

[0349] The calibration parameters corresponding to the advertisement are calculated using the following method:

[0350] calibration_rate=calibration_rate'*coef; Formula 31

[0351] Where calibration_rate represents the calibration parameters of the advertisement, calibration_rate' represents the calibration parameters of the advertisement to be adjusted, and coef represents the target amplification factor.

[0352] Furthermore, in this embodiment of the application, a method is provided to adjust the calibration parameters by using an expansion factor during the advertising maturity stage. Through the above method, the calibration parameters can also be adjusted in conjunction with the advertising stage, so that the calibration parameters are closer to the actual situation, thereby improving the rationality of parameter optimization.

[0353] Optionally, in the above Figure 3Based on the corresponding embodiments, in another optional embodiment provided by this application, the sample data for each advertisement also includes the total number of conversions counted per unit time and the total PCVR counted per unit time;

[0354] It may also include:

[0355] For any ad, if the total number of conversions in the first period corresponding to the ad is less than the conversion threshold and the ad spend is less than the target CPA threshold, then the seventh conversion and the seventh PCVR of the ad in the current period are determined based on the total number of conversions and the total PCVR of the ad in the unit time. The current period includes at least one unit time.

[0356] The calibration parameters corresponding to the advertisement are determined based on the seventh conversion number of the advertisement in the current period, the seventh PCVR in the current period, and the preset expansion coefficient.

[0357] Based on the PCVR before calibration for each click corresponding to the advertisement and the calibration parameters, determine the sum of the PCVR of the effective clicks for the advertisement.

[0358] In one or more embodiments, a method for adjusting calibration parameters using an amplification factor in the initial stage of an advertisement is described. For ease of explanation, the following description focuses on any single advertisement; it is understood that the processing method for other advertisements is similar and will not be elaborated upon here.

[0359] Specifically, if the total number of conversions (conversion_num) in the first period corresponding to the advertisement is less than the conversion threshold (e.g., 2), and the cost is less than the target CPA threshold (e.g., 2*target_cpa), then the calibration parameters can be calculated as follows:

[0360] calibration_rate'=Conv_valid / PCVR_valid;

[0361] Here, 'calibration_rate' represents the calibration parameters to be adjusted for the advertisement, 'Conv_valid' represents the seventh conversion number of the advertisement in the current period, and 'PCVR_valid' represents the seventh PCVR of the advertisement in the current period. The seventh conversion number is determined based on the total number of conversions counted for the advertisement within a unit of time. For example, if the unit of time is 1 hour, the current period is the day, and assuming the current time is 8 o'clock, then the seventh conversion number of the current period can be the sum of the total conversions counted over 8 hours. Similarly, the seventh PCVR is determined based on the total PCVR counted for the advertisement within a unit of time.

[0362] The calibration parameters corresponding to the advertisement are calculated using the following method:

[0363] calibration_rate=calibration_rate'*coef; Formula 32

[0364] Wherein, calibration_rate represents the calibration parameters of the advertisement, calibration_rate' represents the calibration parameters of the advertisement to be adjusted, and coef represents the preset amplification factor, which is set to 1.

[0365] Secondly, this application provides a method for adjusting calibration parameters using an expansion factor in the initial stage of advertising. Through the above method, the calibration parameters can also be adjusted in conjunction with the advertising stage, thereby making the calibration parameters closer to the actual situation and improving the rationality of parameter optimization.

[0366] In conjunction with the above embodiments, this application applies the optimized target parameter set to real data. Please refer to Table 6, which shows the average PCVR_bias for the advertising dimension across four time periods throughout the day.

[0367] Table 6

[0368] Time period abs(bias-1) Consumption-weighted 0～6 0.300296 6～12 0.170387 12～18 0.138765 18～24 0.101298

[0369] Please refer to Table 7, which shows the PCVR_bias distribution for the advertising dimension across the four time periods throughout the day.

[0370] Table 7

[0371] Time period ＞1.5 ≤1.5&&＞1.2 ≤1.2&&＞1.1 ≤1.1&&＞1 ≤1&&＞0.9 ≤0.9&&＞0.8 ≤0.8&&＞0.5 ≤0.5 0 to 6 0.01039241 0.0301424 0.13843152 0.15750941 0.16365921 0.09027276 0.26865092 0.14253319 6 to 12 0.00807043 0.03623721 0.14829387 0.18378691 0.14713943 0.10209177 0.27434482 0.11713534 12 to 18 0.00710738 0.05163021 0.16543901 0.12271714 0.13515805 0.19374584 0.23198627 0.09828426 18 to 24 0.00656818 0.05831331 0.19511632 0.1481637 0.14153723 0.16157365 0.19718917 0.08941736

[0372] Please refer to Table 8, which shows the average PCVR_bias for the four time periods throughout the day for the advertiser dimension.

[0373] Table 8

[0374] Time period abs(bias-1) Consumption-weighted 0～6 0.200161 6～12 0.127125 12～18 0.101429 18～24 0.078564

[0375] Please refer to Table 9, which shows the PCVR_bias distribution for the four time periods throughout the day from the advertiser's perspective.

[0376] Table 9

[0377] Time period ＞1.5 ≤1.5&&＞1.2 ≤1.2&&＞1.1 ≤1.1&&＞1 ≤1&&＞0.9 ≤0.9&&＞0.8 ≤0.8&&＞0.5 ≤0.5 0 to 6 0.00401623 0.01992328 0.0241241 0.14316418 0.16215476 0.17659573 0.30333079 0.16161749 6 to 12 0.00114191 0.02714417 0.14326144 0.12897183 0.11029272 0.12219456 0.31177432 0.08223347 12 to 18 0.00109321 0.05148427 0.15763002 0.13948304 0.1740271 0.14675583 0.2502171 0.0783976 18 to 24 0.00028772 0.02682952 0.18373983 0.18301896 0.21367493 0.1468374 0.1912829 0.06518341

[0378] Therefore, it is evident that, from the mean perspective, the optimized parameters have brought a significant positive effect, achieving substantial improvements in PCVR prediction for ads across all time periods. Looking at the PCVR bias distribution, the proportion of underestimation and overestimation has been significantly reduced in each time period. Clear optimizations have been made in PCVR prediction across all time periods, demonstrating that optimization effects have been achieved for all time segments of the ad campaign.

[0379] The conversion rate prediction and determination apparatus in this application is described in detail below. Please refer to [link / reference]. Figure 4 , Figure 4 This is a schematic diagram of one embodiment of the conversion rate prediction and determination device in this application. The conversion rate prediction and determination device 20 includes:

[0380] The acquisition module 201 is used to acquire the set of parameters to be optimized, the target cost per action (CPA) value, and the sample data set of K advertisements. The sample data set includes sample data for each advertisement. The sample data for each advertisement includes the actual total number of conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period, and K is an integer greater than 1.

[0381] The determination module 202 is used to determine the sum of the PCVR of the effective clicks for each advertisement based on the set of parameters to be optimized and the target CPA, and according to the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration.

[0382] The determination module 202 is also used to determine the average difference for K ads based on the sum of PCVR of the effective clicks of each ad and the actual total conversions;

[0383] Training module 203 is used to train the set of parameters to be optimized with the goal of minimizing the average difference, so as to obtain the target parameter set, wherein the target parameter set includes at least one optimized parameter;

[0384] The determination module 202 is also used to determine the calibrated PCVR of the target advertisement based on the target parameter set, the associated data of the target advertisement, and the PCVR to be adjusted. The associated data of the target advertisement includes the total number of conversions of the target advertisement in the third period, the total PCVR of the target advertisement in the third period, the total number of conversions of the target advertisement in the fourth period, the duration of the target advertisement in the third period, and the consumption value of the target advertisement. The fourth period is the period preceding the third period.

[0385] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application,

[0386] The determination module 202 is specifically used to divide the parameter set based on the consumption and the target CPA, and to determine the factor parameters corresponding to each advertisement according to the consumption value corresponding to each advertisement. The consumption division parameters included in the consumption division parameter set are preset.

[0387] Based on the set of associated parameters, the smoothing parameters for each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period.

[0388] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0389] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0390] Training module 203 is specifically used to train the set of associated parameters with the goal of minimizing the average difference, so as to obtain the target parameter set.

[0391] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the set of associated parameters includes a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter;

[0392] The determination module 202 is specifically used for any one of the K ads. If the total number of conversions in the second period corresponding to any one ad is not zero, then based on the first association parameter, the second association parameter, the third association parameter, and the fourth association parameter, and according to the duration in the first period corresponding to any one ad and the total number of conversions in the second period, the smoothing parameter corresponding to any one ad is determined.

[0393] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter is determined to be the smoothing parameter corresponding to any one ad, where the default association parameter is preset.

[0394] The training module 203 is specifically used to train the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter with the goal of minimizing the average difference, so as to obtain the target parameter set, wherein the target parameter set includes the optimized first correlation parameter, the optimized second correlation parameter, the optimized third correlation parameter, and the optimized fourth correlation parameter.

[0395] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the set of associated parameters includes default associated parameters;

[0396] The determination module 202 is specifically used for any one of the K advertisements. If the total conversion number of any one advertisement in the second period is not zero, then based on the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter, according to the duration of any one advertisement in the first period and the total conversion number in the second period, the smoothing parameter corresponding to any one advertisement is determined. The first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter are preset.

[0397] For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter will be set to the smoothing parameter corresponding to any one ad.

[0398] The training module 203 is specifically used to train the default association parameters with the goal of minimizing the average difference, so as to obtain the target parameter set, wherein the target parameter set includes the optimized default association parameters.

[0399] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application,

[0400] The determination module 202 is specifically used to divide the parameter set and target CPA based on the consumption, and to determine the factor parameters corresponding to each advertisement based on the consumption value corresponding to each advertisement;

[0401] Based on the set of associated parameters, the smoothing parameters corresponding to each advertisement are determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The associated parameters included in the set of associated parameters are pre-set.

[0402] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0403] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0404] Training module 203 is specifically used to train the parameter set of consumption with the goal of minimizing the average difference, so as to obtain the target parameter set.

[0405] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the consumption division parameter set includes N consumption division parameters, where N is an integer greater than or equal to 1;

[0406] The determination module 202 is specifically used to determine the factor parameters corresponding to each advertisement based on the N value, N consumption division parameters and target CPA, and according to the consumption value corresponding to each advertisement, where the N value is preset;

[0407] The training module 203 is specifically used to train each of the N consumption partitioning parameters with the goal of minimizing the average difference, so as to obtain a target parameter set, wherein the target parameter set includes the optimized N consumption partitioning parameters.

[0408] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application,

[0409] The determination module 202 is specifically used to determine the smoothing parameter corresponding to each advertisement based on the set of associated parameters, according to the duration of each advertisement in the first period and the total number of conversions in the second period. The set of associated parameters includes the default associated parameter, the first associated parameter, the second associated parameter, the third associated parameter, and the fourth associated parameter.

[0410] Based on the consumption-based parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined. The consumption-based parameter set includes an N value and N consumption-based parameters, where N is an integer greater than or equal to 1.

[0411] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters for each advertisement in the first period, determine the calibration parameters for each advertisement.

[0412] Based on the PCVR before calibration for each click and the calibration parameters for each ad, determine the sum of the PCVR for each ad's effective click count;

[0413] The training module 203 is specifically used to train the set of association parameters with the goal of minimizing the average difference, so as to obtain the target parameter set. The target parameter set includes the optimized first association parameter, the optimized second association parameter, the optimized third association parameter, the optimized fourth association parameter, the optimized default association parameter, the optimized N value, and the optimized N consumption partitioning parameters.

[0414] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the sample data for each advertisement also includes advertiser information, the total number of conversions counted per unit time, and the total PCVR counted per unit time;

[0415] The determination module 202 is specifically used to divide K advertisements into T1 advertisement sets according to the advertiser information corresponding to each advertisement, wherein each advertisement set includes at least one advertisement, and T1 is an integer greater than or equal to 1;

[0416] If the total spending value corresponding to each ad set in the T1 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set. The (N+1) spending intervals are determined based on the spending parameter set and the target CPA. Each ad subset includes at least one ad, and the same ad subset corresponds to the same spending interval. N is an integer greater than or equal to 1.

[0417] For each of at least one subset of advertisements, the first conversion and the first PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in each subset within a unit of time, wherein the current period includes at least one unit of time.

[0418] For each of the T1 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the second number of conversions and the second PCVR of each ad set in the current period.

[0419] Based on the first conversion count, the first PCVR, the second conversion count, and the second PCVR, determine the factor parameters corresponding to each advertisement.

[0420] Optionally, in the above Figure 4Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the sample data for each advertisement also includes brand information;

[0421] The determination module 202 is specifically used to divide the K advertisements into T2 advertisement sets according to the brand information corresponding to each advertisement if the total consumption value corresponding to any one of the T1 advertisement sets is less than the consumption threshold. Each advertisement set includes at least one advertisement, and T2 is an integer greater than or equal to 1.

[0422] If the total spending value corresponding to each ad set in the T2 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0423] For each of at least one subset of advertisements, the third conversion and the third PCVR of each advertisement subset in the current period are determined based on the total conversions and the total PCVR of each advertisement in the current period.

[0424] For each of the T2 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the fourth number of conversions and the fourth PCVR of each ad set in the current period.

[0425] Based on the third conversion count, third PCVR, fourth conversion count, and fourth PCVR, determine the factor parameters corresponding to each advertisement.

[0426] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the sample data for each advertisement also includes product information;

[0427] The determination module 202 is specifically used to divide the K advertisements into T3 advertisement sets according to the product information corresponding to each advertisement if the total consumption value corresponding to any one of the T2 advertisement sets is less than the consumption threshold. Each advertisement set includes at least one advertisement, and T3 is an integer greater than or equal to 1.

[0428] If the total spending value corresponding to each of the T3 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set.

[0429] For each ad subset in at least one ad subset, determine the fifth conversion number and the fifth PCVR of each ad subset in the current period based on the total conversion number and the total PCVR count of each ad in each ad subset within the unit time.

[0430] For each of the T3 ad sets, based on the total number of conversions and the total PCVR of each ad in each ad set within a unit of time, determine the sixth conversion and the sixth PCVR of each ad set in the current period.

[0431] Based on the fifth conversion count, fifth PCVR, sixth conversion count, and sixth PCVR, determine the factor parameters corresponding to each advertisement.

[0432] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application,

[0433] The determination module 202 is specifically used to determine the target CPA range for the ad's spend value for any given ad.

[0434] Based on the target CPA range, the target expansion coefficient is determined according to the ad spend value and the total conversions in the first period.

[0435] Based on the total conversions, total PCVR, smoothing parameters, and factor parameters of the advertisement in the first period, determine the calibration parameters to be adjusted for the advertisement.

[0436] The calibration parameters corresponding to the advertisement are determined based on the calibration parameters to be adjusted and the target magnification factor.

[0437] Optionally, in the above Figure 4 Based on the corresponding embodiments, in another embodiment of the conversion rate prediction and determination device 20 provided in this application, the sample data for each advertisement also includes the total number of conversions counted per unit time and the total PCVR counted per unit time;

[0438] The determination module 202 is also used to determine, for any advertisement, if the total number of conversions in the first period corresponding to the advertisement is less than the conversion threshold and the ad spend is less than the target CPA threshold, then based on the total number of conversions and the total PCVR counted by the advertisement in the unit time, the seventh conversion and the seventh PCVR of the advertisement in the current period are determined, wherein the current period includes at least one unit time.

[0439] The determination module 202 is also used to determine the calibration parameters corresponding to the advertisement based on the seventh conversion number of the advertisement in the current period, the seventh PCVR in the current period, and the preset expansion coefficient;

[0440] The determination module 202 is also used to determine the sum of the PCVR of the effective clicks of the advertisement based on the PCVR before each click calibration and the calibration parameters.

[0441] Figure 5 This is a schematic diagram of a computer device structure provided in an embodiment of this application. The computer device 300 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 322 (e.g., one or more processors) and a memory 332, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 342 or data 344. The memory 332 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the computer device. Furthermore, the CPU 322 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the computer device 300.

[0442] Computer device 300 may also include one or more power supplies 326, one or more wired or wireless network interfaces 350, one or more input / output interfaces 358, and / or one or more operating systems 341, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0443] The steps performed by the computer device in the above embodiments can be based on this Figure 5 The computer device structure shown.

[0444] This application also provides a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the methods described in the foregoing embodiments.

[0445] This application also provides a computer program product including a program, which, when run on a computer, causes the computer to perform the methods described in the foregoing embodiments.

[0446] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0447] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0448] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0449] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0450] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0451] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for determining conversion rate prediction, characterized in that, include: Obtain the set of parameters to be optimized, the target cost per action (CPA), and a sample data set of K ads. The sample data set includes sample data for each ad, and the sample data for each ad includes the actual total conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period. K is an integer greater than 1. The target CPA is the target cost per conversion. Based on the set of parameters to be optimized and the target CPA, and according to the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, the sum of the PCVR of the effective clicks for each ad is determined. The average difference for the K ads is determined based on the sum of the PCVR of the effective clicks for each ad and the actual total conversions. With the goal of minimizing the average difference, the set of parameters to be optimized is trained to obtain a target parameter set, wherein the target parameter set includes at least one optimized parameter; Based on the target parameter set, and according to the associated data of the target ad and the PCVR to be adjusted, the calibrated PCVR of the target ad is determined. The associated data of the target ad includes the total number of conversions of the target ad in the third period, the total PCVR of the target ad in the third period, the total number of conversions of the target ad in the fourth period, the duration of the target ad in the third period, and the ad spend value. The fourth period is the period preceding the third period.

2. The determination method according to claim 1, characterized in that, The step of determining the sum of PCVR of effective clicks for each ad based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, includes: Based on the consumption segmentation parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined, wherein the consumption segmentation parameters included in the consumption segmentation parameter set are preset; Based on the set of associated parameters, the smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The step of training the set of parameters to be optimized with the objective of minimizing the average difference to obtain the target parameter set includes: The set of associated parameters is trained with the goal of minimizing the average value of the difference to obtain the target parameter set.

3. The determination method according to claim 2, characterized in that, The set of associated parameters includes a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter; The step of determining the smoothing parameter corresponding to each advertisement based on the set of associated parameters, according to the duration of each advertisement in the first period and the total number of conversions in the second period, includes: For any one of the K ads, if the total conversions in the second period corresponding to any one ad are not zero, then based on the first association parameter, the second association parameter, the third association parameter, and the fourth association parameter, and according to the duration in the first period corresponding to any one ad and the total conversions in the second period, the smoothing parameter corresponding to any one ad is determined. For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter is determined to be the smoothing parameter corresponding to any one ad, wherein the default association parameter is preset. The step of training the set of related parameters to obtain the target parameter set with the objective of minimizing the average difference includes: With the goal of minimizing the average difference, the first association parameter, the second association parameter, the third association parameter, and the fourth association parameter are trained to obtain the target parameter set, wherein the target parameter set includes the optimized first association parameter, the optimized second association parameter, the optimized third association parameter, and the optimized fourth association parameter.

4. The determination method according to claim 2, characterized in that, The set of associated parameters includes default associated parameters; The step of determining the smoothing parameter corresponding to each advertisement based on the set of associated parameters, according to the duration of each advertisement in the first period and the total number of conversions in the second period, includes: For any one of the K ads, if the total conversions in the second period corresponding to any one ad are not zero, then based on the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter, and according to the duration in the first period corresponding to any one ad and the total conversions in the second period, a smoothing parameter corresponding to any one ad is determined, wherein the first correlation parameter, the second correlation parameter, the third correlation parameter, and the fourth correlation parameter are preset. For any one of the K ads, if the total conversions in the second period corresponding to any one ad are zero, then the default association parameter is determined as the smoothing parameter corresponding to any one ad; The step of training the set of related parameters to obtain the target parameter set with the objective of minimizing the average difference includes: With the goal of minimizing the average difference, the default association parameters are trained to obtain the target parameter set, wherein the target parameter set includes the optimized default association parameters.

5. The determination method according to claim 1, characterized in that, The step of determining the sum of PCVR of effective clicks for each ad based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, includes: Based on the consumption-based parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined; Based on the set of associated parameters, the smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The associated parameters included in the set of associated parameters are preset. The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The step of training the set of parameters to be optimized with the objective of minimizing the average difference to obtain the target parameter set includes: With the goal of minimizing the average difference, the parameter set for the consumption amount is trained to obtain the target parameter set.

6. The determination method according to claim 5, characterized in that, The consumption division parameter set includes N consumption division parameters, where N is an integer greater than or equal to 1; The parameter set based on consumption and the target CPA, and the factor parameters corresponding to each advertisement determined according to the ad spend value corresponding to each advertisement, include: Based on the N value, the N consumption division parameters, and the target CPA, and according to the ad spend value corresponding to each ad, the factor parameters corresponding to each ad are determined, wherein the N value is preset; The step of training the set of related parameters to obtain the target parameter set with the objective of minimizing the average difference includes: With the goal of minimizing the average difference, each of the N consumption partitioning parameters is trained to obtain the target parameter set, wherein the target parameter set includes the optimized N consumption partitioning parameters.

7. The determination method according to claim 1, characterized in that, The step of determining the sum of PCVR of effective clicks for each ad based on the set of parameters to be optimized and the target CPA, and according to the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration, includes: Based on the set of associated parameters, a smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The set of associated parameters includes a default associated parameter, a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter. Based on the consumption segmentation parameter set and the target CPA, and according to the ad spend value corresponding to each ad, the factor parameters corresponding to each ad are determined, wherein the consumption segmentation parameter set includes an N value and N consumption segmentation parameters, and N is an integer greater than or equal to 1; The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The step of training the set of parameters to be optimized with the objective of minimizing the average difference to obtain the target parameter set includes: With the goal of minimizing the average difference, the set of associated parameters is trained to obtain the target parameter set, wherein the target parameter set includes the optimized first associated parameter, the optimized second associated parameter, the optimized third associated parameter, the optimized fourth associated parameter, the optimized default associated parameter, the optimized N value, and the optimized N consumption partitioning parameters.

8. The determining method according to any one of claims 2 to 7, characterized in that, The sample data for each advertisement also includes advertiser information, the total number of conversions counted per unit time, and the total PCVR counted per unit time; The parameter set based on consumption and the target CPA, and the factor parameters corresponding to each advertisement determined according to the ad spend value corresponding to each advertisement, include: Based on the advertiser information corresponding to each advertisement, the K advertisements are divided into T1 advertisement sets, wherein each advertisement set includes at least one advertisement, and T1 is an integer greater than or equal to 1; If the total spending value corresponding to each of the T1 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on (N+1) spending intervals and the spending value corresponding to each ad in each ad set. The (N+1) spending intervals are determined based on the spending amount division parameter set and the target CPA. Each ad subset includes at least one ad, and the same ad subset corresponds to the same spending interval. N is an integer greater than or equal to 1. For each of the at least one subset of advertisements, based on the total number of conversions counted per unit time and the total PCVR counted per unit time for each advertisement in each subset of advertisements, a first number of conversions and a first PCVR for each subset of advertisements in the current period are determined, wherein the current period includes at least one unit of time. For each of the T1 ad sets, based on the total conversion count and the total PCVR count for each ad in each ad set within a unit time, determine the second conversion count and the second PCVR for each ad set in the current period. Based on the first conversion count, the first PCVR, the second conversion count, and the second PCVR, determine the factor parameters corresponding to each advertisement.

9. The determining method according to claim 8, characterized in that, The sample data for each advertisement also includes brand information; The method further includes: If the total spending value of any one of the T1 ad sets is less than the spending threshold, the K ads are divided into T2 ad sets according to the brand information corresponding to each ad, wherein each ad set includes at least one ad, and T2 is an integer greater than or equal to 1. If the total spending value corresponding to each ad set in the T2 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on the (N+1) spending intervals and the spending value corresponding to each ad in each ad set. For each of the at least one subset of advertisements, the third conversion number and the third PCVR of each subset of advertisements in the current period are determined based on the total conversion number and the total PCVR of each advertisement in the current period corresponding to each advertisement in the current subset of advertisements. For each of the T2 ad sets, based on the total number of conversions counted per unit time and the total PCVR counted per unit time for each ad in each ad set, determine the fourth number of conversions and the fourth PCVR for each ad set in the current period. Based on the third conversion count, the third PCVR, the fourth conversion count, and the fourth PCVR, determine the factor parameters corresponding to each advertisement.

10. The determination method according to claim 9, characterized in that, The sample data for each advertisement also includes product information; The method further includes: If the total spending value corresponding to any one of the T2 ad sets is less than the spending threshold, the K ads are divided into T3 ad sets according to the product information corresponding to each ad, wherein each ad set includes at least one ad, and T3 is an integer greater than or equal to 1. If the total spending value corresponding to each of the T3 ad sets is greater than or equal to the spending threshold, then at least one ad subset is generated based on the (N+1) spending intervals and the spending value corresponding to each ad in each ad set. For each of the at least one subset of advertisements, the fifth conversion number and the fifth PCVR of each subset of advertisements in the current period are determined based on the total conversion number and the total PCVR of each advertisement in the current period corresponding to each advertisement in the current subset. For each of the T3 ad sets, based on the total number of conversions counted per unit time and the total PCVR counted per unit time for each ad in each ad set, determine the sixth conversion number and the sixth PCVR for each ad set in the current period. Based on the fifth conversion number, the fifth PCVR, the sixth conversion number, and the sixth PCVR, determine the factor parameters corresponding to each advertisement.

11. The determining method according to any one of claims 2 to 7, characterized in that, The step of determining the calibration parameters corresponding to each advertisement based on the total conversions within the first period corresponding to each advertisement, the total PCVR within the first period, the smoothing parameter, and the factor parameter includes: For any given advertisement, determine the target CPA range within which the ad spend value corresponds to the advertisement; Based on the target CPA range, the target expansion coefficient is determined according to the ad spend value and the total conversions in the first period. Based on the total conversions within the first period corresponding to the advertisement, the total PCVR within the first period, the smoothing parameter, and the factor parameter, determine the calibration parameters to be adjusted for the advertisement. The calibration parameters corresponding to the advertisement are determined based on the calibration parameters to be adjusted corresponding to the advertisement and the target magnification coefficient.

12. The determination method according to claim 1, characterized in that, The sample data for each advertisement also includes the total number of conversions counted per unit time and the total PCVR counted per unit time; The method further includes: For any given ad, if the total number of conversions in the first period corresponding to the ad is less than the conversion threshold and the ad spend is less than the target CPA threshold, then the seventh conversion and the seventh PCVR of the ad in the current period are determined based on the total number of conversions counted per unit time and the total PCVR counted per unit time, wherein the current period includes at least one unit time. The calibration parameters corresponding to the advertisement are determined based on the seventh conversion number of the advertisement in the current period, the seventh PCVR in the current period, and the preset expansion coefficient. Based on the PCVR before each click calibration corresponding to the advertisement and the calibration parameters, the sum of the PCVR of the effective clicks of the advertisement is determined.

13. A conversion rate prediction and determination device, characterized in that, include: The acquisition module is used to acquire a set of parameters to be optimized, a target cost-per-action (CPA) value, and a sample data set of K advertisements. The sample data set includes sample data for each advertisement, and the sample data for each advertisement includes the actual total conversions, the estimated conversion rate (PCVR) before each effective click calibration, the total conversions in the first period, the total PCVR in the first period, the total conversions in the second period, the duration in the first period, and the campaign cost. The second period is the period preceding the first period, and K is an integer greater than 1. The target CPA is the target cost per conversion. The determination module is used to determine the sum of the PCVR of the effective clicks of each advertisement based on the set of parameters to be optimized and the target CPA, and according to the total number of conversions in the first period, the total PCVR in the first period, the total number of conversions in the second period, the duration in the first period, the campaign cost, and the PCVR before each click calibration. The determining module is further configured to determine the average difference for the K ads based on the sum of the PCVR of the effective clicks of each ad and the actual total conversions; The training module is used to train the set of parameters to be optimized with the goal of minimizing the average difference, so as to obtain a target parameter set, wherein the target parameter set includes at least one optimized parameter; The determining module is further configured to determine the calibrated PCVR of the target advertisement based on the target parameter set, according to the associated data of the target advertisement and the PCVR to be adjusted, wherein the associated data of the target advertisement includes the total number of conversions of the target advertisement in the third period, the total PCVR of the target advertisement in the third period, the total number of conversions of the target advertisement in the fourth period, the duration of the target advertisement in the third period, and the consumption value of the target advertisement, wherein the fourth period is the period preceding the third period.

14. The determining device according to claim 13, characterized in that, The determining module is specifically used for: Based on the consumption segmentation parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined, wherein the consumption segmentation parameters included in the consumption segmentation parameter set are preset; Based on the set of associated parameters, the smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The training module is specifically used for: The set of associated parameters is trained with the goal of minimizing the average value of the difference to obtain the target parameter set.

15. The determining device according to claim 13, characterized in that, The determining module is specifically used for: Based on the consumption-based parameter set and the target CPA, and according to the consumption value corresponding to each advertisement, the factor parameters corresponding to each advertisement are determined; Based on the set of associated parameters, the smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The associated parameters included in the set of associated parameters are preset. The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The training module is specifically used for: With the goal of minimizing the average difference, the parameter set for the consumption amount is trained to obtain the target parameter set.

16. The determining device according to claim 13, characterized in that, The determining module is specifically used for: Based on the set of associated parameters, a smoothing parameter corresponding to each advertisement is determined according to the duration of each advertisement in the first period and the total number of conversions in the second period. The set of associated parameters includes a default associated parameter, a first associated parameter, a second associated parameter, a third associated parameter, and a fourth associated parameter. Based on the consumption segmentation parameter set and the target CPA, and according to the ad spend value corresponding to each ad, the factor parameters corresponding to each ad are determined, wherein the consumption segmentation parameter set includes an N value and N consumption segmentation parameters, and N is an integer greater than or equal to 1; The calibration parameters corresponding to each advertisement are determined based on the total number of conversions in the first period corresponding to each advertisement, the total PCVR in the first period, the smoothing parameter, and the factor parameter. Based on the PCVR before each click calibration for each advertisement and the calibration parameters, determine the sum of the PCVR of the effective click count for each advertisement; The training module is specifically used for: With the goal of minimizing the average difference, the set of associated parameters is trained to obtain the target parameter set, wherein the target parameter set includes the optimized first associated parameter, the optimized second associated parameter, the optimized third associated parameter, the optimized fourth associated parameter, the optimized default associated parameter, the optimized N value, and the optimized N consumption partitioning parameters.

17. A computer device, characterized in that, include: Memory, processor, and bus system; The memory is used to store programs; The processor is configured to execute a program in the memory, and the processor is configured to execute the determination method according to any one of claims 1 to 12 according to the instructions in the program code; The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.

18. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the determining method as claimed in any one of claims 1 to 12.

19. A computer program product, characterized in that, The method includes computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions to cause the computer device to perform the determination method as described in any one of claims 1 to 12.

Images