Advertisement attribution method and device, computer device and readable storage medium

Abstract

The application discloses an advertisement attribution method and device, computer equipment and a readable storage medium, and relates to the field of attribution. The method comprises the following steps: obtaining user data and advertisement path data; encoding the identification data of each touch point according to the encoding module of the attribution model to obtain the identification data representation of each touch point; generating the weight of each touch point by using the attention module of the attribution model; obtaining the path representation of the advertisement path data based on the identification data representation and the weight of each touch point; inputting the user data and the channel data of multiple touch points into the second acquisition module of the attribution model to obtain the data feature representation; and inputting the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability. Thus, the prediction of the conversion probability considers the influence of different user trigger times on the contribution degree of the touch point, the contribution degree of the touch point is associated with the time factor of the user accessing the advertisement, and the effectiveness of the conversion probability is ensured.

Description

Technical Field

[0001] This invention relates to the field of attribution, and more particularly to an attribution method, apparatus, computer device, and readable storage medium for advertising. Background Technology

[0002] Common attribution methods include first-touch attribution, last-interaction attribution, and linear attribution. First-touch attribution and last-interaction attribution attribute the entire contribution of the conversion to the first or last touchpoint in the user's ad campaign path. Linear attribution, on the other hand, distributes the contribution of the conversion evenly to each touchpoint in the ad campaign path.

[0003] It is clear that initial contact attribution ignores the display / exposure effect of each touchpoint before conversion, final interaction attribution overestimates the contribution of the last touchpoint, and linear attribution ignores the differences in contribution between different touchpoints. In other words, existing advertising attribution methods fail to reasonably / accurately reflect the relationship between conversion and touchpoints, resulting in poor advertising attribution effectiveness / accuracy. Summary of the Invention

[0004] In view of this, the present invention provides an advertising attribution method, apparatus, computer device and readable storage medium to improve the current situation where existing advertising attribution methods are unable to reasonably / accurately reflect the relationship between conversion and touchpoint, resulting in poor advertising attribution effect / accuracy.

[0005] In a first aspect, embodiments of the present invention provide an attribution method for advertising, comprising:

[0006] Acquire user data and advertising path data, wherein the advertising path data includes touchpoint data of multiple touchpoints, the touchpoint data includes identification data and channel data, and the touchpoint data of multiple touchpoints is sorted according to the user trigger time;

[0007] The encoding module based on the pre-trained attribution model encodes the identification data of each touch point to obtain the identification data representation of each touch point;

[0008] Based on the attention module of the attribution model, a weight is generated for each touchpoint, wherein the greater the difference between the user trigger time of the touchpoint and the user trigger time of the last touchpoint, the lower the weight of the touchpoint.

[0009] Based on the identifier data representation and weight of each touchpoint, the path representation of the advertising path data is obtained;

[0010] The user data and the channel data of the multiple touchpoints are input into the second acquisition module of the attribution model to obtain a data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output the data feature representation.

[0011] The path representation and the data feature representation are input into the prediction module of the attribution model to obtain the conversion probability.

[0012] Optionally, in one feasible embodiment of the present invention, the attribution model further includes a splicing module, a flattening module, and a fully connected module;

[0013] The path representation of the advertising path data, obtained based on the identifier data representation and weight of each touchpoint, includes:

[0014] Calculate the product of the identifier data representation and the weight of the contact point to obtain the weighted data representation of the contact point;

[0015] The weighted data representation of each contact point is input into the splicing module to obtain the first splicing result;

[0016] The first stitching result is input into the flattening module to obtain the flattening result;

[0017] The flattening result is input into the fully connected module to obtain the path representation of the advertising path data.

[0018] Optionally, in one feasible embodiment of the present invention, the second acquisition module includes an embedding layer, a first splicing layer, and a first fully connected layer;

[0019] The step of inputting the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain data feature representation includes:

[0020] The user data and the channel data of the multiple touchpoints are respectively input into the embedding layer to obtain the embedding vector corresponding to the user data and the embedding vector corresponding to the channel data of each touchpoint.

[0021] All the embedding vectors are input into the first concatenation layer to obtain the second concatenation result;

[0022] All the second concatenation results are input into the first fully connected layer to obtain the data feature representation.

[0023] Optionally, in one feasible embodiment of the present invention, the prediction module includes a second splicing layer, a second fully connected layer, and a probability calculation layer;

[0024] The step of inputting the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability includes:

[0025] The path representation and the data feature representation are input into the second splicing layer to obtain the third splicing result;

[0026] The third splicing result is input into the second fully connected layer to obtain the corresponding output data;

[0027] The output data is input into the probability calculation layer to obtain the conversion probability.

[0028] Optionally, in one feasible embodiment of the present invention, obtaining user data and advertising path data includes:

[0029] Obtain raw user data and raw ad path data;

[0030] The original user data and the original advertising path data are cleaned to obtain user data and advertising path data.

[0031] Optionally, in one feasible embodiment of the present invention, the encoding module is implemented based on a bidirectional long short-term memory network.

[0032] Secondly, embodiments of the present invention provide an attribution device for advertising, comprising:

[0033] The data acquisition module is used to acquire user data and advertising path data. The advertising path data includes touchpoint data of multiple touchpoints. The touchpoint data includes identification data and channel data. The touchpoint data of multiple touchpoints is sorted according to the user trigger time.

[0034] The encoding module is used to encode the identification data of each of the touch points based on the pre-trained attribution model to obtain the identification data representation of each of the touch points;

[0035] The generation module is used to generate the weight of each touch point based on the attention module of the attribution model, wherein the greater the difference between the user trigger time of the touch point and the user trigger time of the last touch point, the lower the weight of the touch point.

[0036] The path representation acquisition module is used to obtain the path representation of the advertising path data based on the identifier data representation and weight of each touchpoint;

[0037] The feature representation acquisition module is used to input the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain the data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output the data feature representation.

[0038] The prediction module is used to input the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability.

[0039] Optionally, in one feasible embodiment of the present invention, the attribution model further includes a splicing module, a flattening module, and a fully connected module;

[0040] The path represents the acquisition module, including:

[0041] The product calculation submodule is used to calculate the product of the identification data representation and the weight of the touch point to obtain the weighted data representation of the touch point;

[0042] The first splicing submodule is used to input the weighted data representation of each contact point into the splicing module to obtain the first splicing result;

[0043] The first flattening processing submodule is used to input the first splicing result into the flattening module to obtain the flattening result;

[0044] The fully connected processing submodule is used to input the flattening result into the fully connected module to obtain the path representation of the advertising path data.

[0045] Thirdly, embodiments of the present invention provide a computer device including a memory and a processor, the memory storing a computer program, the computer program executing an advertising attribution method as disclosed in the first aspect when running on the processor.

[0046] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when run on a processor, executes an advertising attribution method as disclosed in the first aspect.

[0047] In the advertising attribution method provided in this embodiment of the invention, the following steps are taken: First, data representing user personal characteristics information, such as records of ads clicked / browsed / accessed by the user, are acquired, i.e., user data and ad path data are acquired. Next, according to the encoding module of the pre-trained attribution model, the identifier data of each touchpoint in the ad path data is encoded to obtain the identifier data representation of each touchpoint. Then, based on the rule that the greater the difference between the user trigger time of a touchpoint and the user trigger time of the last / end touchpoint, the smaller the contribution of the touchpoint, the attention module of the attribution model is used to generate the weight of each touchpoint. Subsequently, based on the identifier data representation and weight of each touchpoint, the path representation of the ad path data is obtained. Afterward, based on the correlation between user data, channel data, and conversion, the user data and channel data of multiple touchpoints are input into the second acquisition module of the attribution model to obtain the data feature representation. Finally, the path representation and data feature representation are input into the prediction module of the attribution model to predict the probability that the user will perform a conversion behavior after the user trigger time of the last / end touchpoint, i.e., the conversion probability. Based on this, the embodiments of the present invention enable the prediction of conversion probability to take into account the contribution of different user trigger times to touchpoints, so that the contribution of each touchpoint can be correctly correlated with the time factor of user access to the advertisement, thereby ensuring the effectiveness of conversion probability. Moreover, in addition to considering the differences in the contribution of different user trigger times to touchpoints, the embodiments of the present invention also take into account the characteristics of user information and the characteristics of the advertising delivery channels corresponding to the touchpoints, so that the prediction of conversion probability can be based on multi-dimensional information features, thereby further improving the prediction accuracy of conversion probability. Attached Figure Description

[0048] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection of the present invention. In the various drawings, similar components are numbered similarly.

[0049] Figure 1 A flowchart illustrating the first advertising attribution method provided in an embodiment of the present invention is shown;

[0050] Figure 2 A flowchart illustrating the second advertising attribution method provided in an embodiment of the present invention is shown;

[0051] Figure 3 A flowchart illustrating the third advertising attribution method provided in an embodiment of the present invention is shown;

[0052] Figure 4 A schematic diagram of the advertising attribution device provided in an embodiment of the present invention is shown. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0054] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0055] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0056] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0057] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0058] To better illustrate the various embodiments provided by the present invention, the attribution of advertising is briefly described as follows:

[0059] Suppose a user, after browsing and clicking on various ads, ultimately performs a pre-defined action on the product corresponding to the ad, such as purchasing, adding to favorites / saving, or subscribing. The user's behavior / action of browsing or clicking on ads at different times is considered a "touchpoint," the user's behavior / action of performing the pre-defined action is considered a "conversion," and the series of processes by which a user, based on touchpoints at different times, ultimately generates or does not generate a "conversion" within a certain period of time is called the advertising path.

[0060] Furthermore, due to the different channels through which ads are delivered, each touchpoint can be further categorized according to the channel, such as display ad touchpoints, video ad touchpoints, and search ad touchpoints, etc.

[0061] Example 1

[0062] Reference Figure 1 The diagram illustrates a flowchart of a first advertising attribution method provided by an embodiment of the present invention. The advertising attribution method provided by the embodiment of the present invention includes:

[0063] Step S110: Obtain user data and advertising path data, wherein the advertising path data includes touchpoint data of multiple touchpoints, the touchpoint data includes identification data and channel data, and the multiple touchpoint data are sorted according to the user trigger time.

[0064] It is understood that the user data in the embodiments of the present invention includes, but is not limited to, the user's network nickname, location, beacon information, IP (Internet Protocol) address, gender and age, etc., which can be set according to the actual situation.

[0065] It is also understood that the advertising path data in the embodiments of the present invention refers to the relevant data of the advertisements that the user browses and / or clicks within a period of time, that is, the touch point data of each touch point sorted by time.

[0066] Furthermore, the identifier data in the touchpoint data represents information that identifies the advertisement viewed or clicked by the user, such as the advertisement's unique serial number, name, and hash value. Channel data indicates the channel corresponding to the advertisement viewed or clicked by the user, such as identifying whether the advertisement is a video ad, a click ad, or a display ad.

[0067] Furthermore, it is understood that the attribution method provided in this embodiment of the invention is used to predict the probability of conversion based on touchpoint data and user data, and the conversion will occur after the user's last browsing / clicking of the advertisement, that is, the conversion will occur at the end / last touchpoint after the user's trigger time.

[0068] It is easy to understand that, since the contact data of multiple contacts in this embodiment of the invention are sorted according to the user trigger time, the user trigger time of the last contact is the latest, and the user trigger time of the first contact is the earliest. If a transformation occurs after the user trigger time of the last contact, the transformation time is closest to the user trigger time of the last contact and furthest from the user trigger time of the first contact.

[0069] Another point to consider is that the user trigger time of the end touchpoint in the embodiments of the present invention can be regarded as the end time of ad access, that is, the time when the user stops viewing / clicking / accessing the ad.

[0070] Furthermore, it is understood that if the quality of the contact data is low, it will lead to a decrease in prediction accuracy. In response to this, step S110, in one feasible embodiment of the present invention, specifically includes:

[0071] Obtain raw user data and raw ad path data;

[0072] The original user data and the original advertising path data are cleaned to obtain user data and advertising path data.

[0073] In other words, after obtaining user data and advertising path data, this embodiment of the invention will perform data cleaning on the user data and advertising path data to eliminate erroneous data / invalid data / blank data in the user data and advertising path data, thereby improving the data quality of user data and advertising path data, and ensuring that subsequent method steps can be executed based on effective data support.

[0074] Step S120: The encoding module based on the pre-trained attribution model encodes the identification data of each touch point to obtain the identification data representation of each touch point.

[0075] That is, in this embodiment of the invention, the encoding module in the pre-trained attribution model will be used to encode the identification data of each touch point to obtain the identification data representation of each touch point.

[0076] It is understood that the encoding module in the embodiments of the present invention can be implemented by selecting any model / network / tool ​​according to the actual situation.

[0077] As an example, in one feasible embodiment of the present invention, the encoding module is implemented based on a recurrent neural network (RNN).

[0078] In another feasible approach, since the advertisement that a user browses / clicks at the current moment may be continuous with the advertisement at the previous moment and / or the advertisement at the next moment, that is, the advertisements corresponding to the previous moment, the current moment and the next moment may have temporal correlation and / or content correlation, the encoding module in the embodiment of the present invention is implemented based on a bidirectional long short-term memory network.

[0079] Therefore, when the encoding module receives the identification data of multiple touch points sorted by time, it will encode the identification data of each touch point according to the identification data of the previous touch point and the identification data of the next touch point. This will make the identification data representation of any touch point contain contextual features. As a result, compared with the identification data representation based on recurrent neural networks, the identification data representation based on bidirectional long short-term memory networks will contain richer information, allowing the attribution model to obtain more data / information support when performing subsequent steps.

[0080] S130, Based on the attention module of the attribution model, generate the weight of each touch point, wherein the greater the difference between the user trigger time of the touch point and the user trigger time of the last touch point, the lower the weight of the touch point.

[0081] Understandably, for conversion behavior, the contribution of ads viewed / clicked by users at different times to conversion behavior varies. Therefore, existing technologies include first-touch attribution, last-interaction attribution, and linear attribution, etc., to determine the contribution of touchpoints at different times.

[0082] In this embodiment of the invention, the following assumptions are made regarding the contribution: if a conversion occurs, the closer the conversion occurs to the user trigger time of the touchpoint, the higher the contribution of the touchpoint, and vice versa.

[0083] Therefore, since the conversion occurs after the user trigger time of the last touchpoint and is closest to the user trigger time of the last touchpoint, the closer the user trigger time of the touchpoint is to the user trigger time of the last touchpoint, the higher the contribution of the touchpoint, and vice versa.

[0084] Based on this, embodiments of the present invention provide a method for determining the contribution of each touchpoint based on an attention mechanism. That is, an attention layer / module is set in the attribution model, and during the inference process after training, based on the attention layer / module, the larger the difference between the user trigger time of a touchpoint and the user trigger time of the last / final touchpoint, the smaller the weight is assigned.

[0085] Therefore, in subsequent steps, the present invention will also determine the contribution of each touch point according to the weight of each touch point. That is, the contribution will be considered according to the user trigger time of the touch point, thereby improving the phenomenon of unreasonable allocation / consideration / setting of the contribution of different touch points at different time nodes.

[0086] Furthermore, it is understood that the specific settings of the attention module / layer can be configured according to the actual situation, and the embodiments of the present invention do not limit this.

[0087] Step S140: Based on the identifier data representation and weight of each touchpoint, obtain the path representation of the advertising path data.

[0088] That is, the embodiments of the present invention will utilize the identification data representation and weight of each to make the identification data representation of each touchpoint in the advertising path data include the characteristics of the user trigger time of the touchpoint, and then make predictions / estimates based on the touchpoints at different times when making conversion predictions.

[0089] It is understood that the method for obtaining the path representation of the advertising path data in this embodiment of the invention is a content that can be set according to the actual situation.

[0090] As an example, in one feasible embodiment of the present invention, the computer device performs a weighted operation based on the identifier data representation and weight of each touchpoint to obtain a weighted identifier data representation of each touchpoint, and then merges the weighted identifier data representations of each touchpoint to obtain a path representation of the advertising path data.

[0091] In another feasible embodiment provided by the present invention, please refer to the following: Figure 2 The diagram shows a flowchart of a second advertising attribution method provided by an embodiment of the present invention. In this feasible mode, the attribution model further includes a splicing module, a flattening module, and a fully connected module.

[0092] Furthermore, step S140, as described above, specifically includes the following feasible method provided in this embodiment of the invention:

[0093] S141, Calculate the product of the identification data representation of the contact point and the weight to obtain the weighted data representation of the contact point;

[0094] S142, input the weighted data representation of each contact point into the splicing module to obtain the first splicing result;

[0095] S143, input the first splicing result into the flattening module to obtain the flattening result;

[0096] S144, The flattening result is input into the fully connected module to obtain the path representation of the advertising path data.

[0097] That is, in the embodiments of the present invention, after obtaining the weighted data representation / weighted identifier data representation of each touch point using the identifier data representation and weight of each touch point, the weighted data representation of all touch points will be flattened. After obtaining the flattened result, the flattened result will be integrated into the corresponding value, i.e., the path representation, using the fully connected layer / module. Then, the path representation will be used to complete the corresponding prediction in the subsequent process.

[0098] Furthermore, it is understood that the embodiments of the present invention do not limit the dimensions of the input data and the output data of the fully connected module, and can be set according to the actual situation.

[0099] Optionally, to avoid the data dimensions of the weighted data representation or the identifier data representation being difficult to adapt to the fully connected layer / module, in one feasible embodiment of the present invention, the attribution model further includes a RepeatVector module. The RepeatVector module is used to adjust the dimensions of the weighted data representation or the identifier data representation of each touch point before executing step S152, so that the adjusted weighted data representation or the identifier data representation can match the fully connected layer / module.

[0100] Step S150: Input the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain a data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output a data representation.

[0101] Understandably, the conversion is related to users' personal factors. For example, older users typically don't buy or pay attention to extreme sports-related products, and unmarried adults typically don't buy infant products.

[0102] It's also understandable that the channels through which ads are placed will influence conversion behavior. For example, for users of social media platforms / software, social ads for products related to the type / trend of the social media platform are more likely to encourage users to make purchases.

[0103] Based on this, the embodiments of the present invention, while considering the contribution of touchpoints at different time points, also use various static data for conversion prediction, that is, use user data and channel data of each touchpoint for conversion prediction.

[0104] Specifically, in this embodiment of the invention, user data and channel data for each touchpoint are embedded to obtain embedding vectors corresponding to user data and channel data respectively. After concatenating all the embedding vectors, conversion prediction is performed based on the concatenated vector.

[0105] Based on this, the embodiments of the present invention will take into account the characteristic information of the advertising channel and the characteristic information of the user when making conversion predictions, so that multi-dimensional information can be referenced during the prediction. Therefore, the embodiments of the present invention can make conversion predictions based on multi-dimensional information features, so that the credibility of the prediction results can be guaranteed.

[0106] It is understood that the specific process for obtaining data feature representations can be set according to actual conditions. For example, in one feasible method provided in this embodiment of the invention, please refer to [the relevant documentation]. Figure 3 The diagram illustrates a flowchart of a third advertising attribution method provided by an embodiment of the present invention. In this feasible method, the second acquisition module includes an embedding layer, a first splicing layer, and a first fully connected layer.

[0107] Furthermore, in this feasible embodiment of the present invention, step S150 specifically includes:

[0108] S151, the user data and the channel data of the multiple touchpoints are respectively input into the embedding layer to obtain the embedding vector corresponding to the user data and the embedding vector corresponding to the channel data of each touchpoint;

[0109] S152, input all the embedded vectors into the first splicing layer to obtain the second splicing result;

[0110] S153, input all the second splicing results into the first fully connected layer to obtain the data feature representation.

[0111] That is, after concatenating all the embedded vectors in the embodiments of the present invention, a fully connected layer will be used to integrate all the embedded vectors into a corresponding numerical representation, so as to make predictions based on the numerical representation corresponding to the second concatenation result in subsequent processes.

[0112] Furthermore, it is understood that the embodiments of the present invention do not limit the dimensions of the input data and the output data of the first fully connected layer, and can be set according to the actual situation.

[0113] Step S160: Input the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability.

[0114] In other words, the embodiments of the present invention will predict the conversion probability based on the relationship between contribution and time, user attributes / information, and the channel characteristics / information of the advertisement corresponding to the touchpoint. That is, the probability of a user performing a conversion behavior after the user trigger time of the last / final touchpoint is calculated.

[0115] It is understandable that the specific settings of the prediction module can be configured according to the actual situation. For example, in one feasible approach, the prediction module includes a first fully connected layer, a second fully connected layer, a third fully connected layer, and a classification function layer connected in sequence.

[0116] In another feasible embodiment of the present invention, the prediction module includes a second splicing layer, a second fully connected layer, and a probability calculation layer;

[0117] Furthermore, step S160, as described above, specifically includes the following feasible method provided in this embodiment of the invention:

[0118] The path representation and the data feature representation are input into the second splicing layer to obtain the third splicing result;

[0119] The third splicing result is input into the second fully connected layer to obtain the corresponding output data;

[0120] The output data is input into the probability calculation layer to obtain the conversion probability.

[0121] That is, in this embodiment of the invention, the path representation and data feature representation are first integrated into a vector / third concatenation result, and then the integrated vector is converted into one-dimensional data / output data using the second fully connected layer. Finally, the probability calculation layer is used to perform corresponding probability calculations on this one-dimensional data to obtain the conversion probability.

[0122] It is understood that the probability calculation layer can be set according to the actual situation. For example, in one feasible way provided in the embodiment of the present invention, the probability calculation layer is implemented based on the SIGMOD function.

[0123] The advertising attribution method provided in this invention allows the prediction of conversion probability to consider the impact of different user trigger times on the contribution of touchpoints, ensuring that the contribution of each touchpoint is correctly correlated with the time factor of user access to the advertisement, thereby ensuring the effectiveness of the conversion probability. Furthermore, in addition to considering the differences in the contribution of different user trigger times to touchpoints, this invention also incorporates the characteristics of user information and the characteristics of the advertising delivery channels corresponding to the touchpoints, enabling the prediction of conversion probability to be based on multi-dimensional information features, thereby further improving the accuracy of conversion probability prediction.

[0124] Example 2

[0125] Corresponding to the advertising attribution method provided in Embodiment 1 of the present invention, Embodiment 2 of the present invention also provides an advertising attribution device, referring to... Figure 4 The diagram shows a schematic of the structure of an advertising attribution device provided in an embodiment of the present invention. The advertising attribution device 200 provided in an embodiment of the present invention includes:

[0126] The data acquisition module 210 is used to acquire user data and advertising path data, wherein the advertising path data includes touch point data of multiple touch points, the touch point data includes identification data and channel data, and the touch point data of multiple touch points is sorted according to the user trigger time;

[0127] Encoding module 220 is used to encode the identification data of each of the touch points based on the pre-trained attribution model to obtain the identification data representation of each of the touch points;

[0128] The generation module 230 is used to generate the weight of each touch point based on the attention module of the attribution model, wherein the greater the difference between the user trigger time of the touch point and the user trigger time of the last touch point, the lower the weight of the touch point.

[0129] The path representation acquisition module 240 is used to obtain the path representation of the advertising path data based on the identifier data representation and weight of each touchpoint;

[0130] The feature representation acquisition module 250 is used to input the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain the data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output the data feature representation.

[0131] The prediction module 260 is used to input the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability.

[0132] Optionally, in one feasible embodiment of the present invention, the attribution model further includes a splicing module, a flattening module, and a fully connected module;

[0133] The path represents the acquisition module, including:

[0134] The product calculation submodule is used to calculate the product of the identification data representation and the weight of the touch point to obtain the weighted data representation of the touch point;

[0135] The first splicing submodule is used to input the weighted data representation of each contact point into the splicing module to obtain the first splicing result;

[0136] The first flattening processing submodule is used to input the first splicing result into the flattening module to obtain the flattening result;

[0137] The fully connected processing submodule is used to input the flattening result into the fully connected module to obtain the path representation of the advertising path data.

[0138] Optionally, in one feasible embodiment of the present invention, the second acquisition module includes an embedding layer, a first splicing layer, and a first fully connected layer;

[0139] The feature representation acquisition module includes:

[0140] An embedding submodule is used to input the user data and the channel data of the multiple touchpoints into the embedding layer respectively, and obtain the embedding vector corresponding to the user data and the embedding vector corresponding to the channel data of each touchpoint.

[0141] The second splicing submodule is used to input all the embedding vectors into the first splicing layer to obtain the second splicing result;

[0142] The data feature representation acquisition submodule is used to input all the second concatenation results into the first fully connected layer to obtain the data feature representation.

[0143] Optionally, in one feasible embodiment of the present invention, the prediction module includes a second splicing layer, a second fully connected layer, and a probability calculation layer;

[0144] The prediction module includes:

[0145] The third splicing submodule is used to input the path representation and the data feature representation into the second splicing layer to obtain the third splicing result;

[0146] The output submodule is used to input the third splicing result into the second fully connected layer to obtain the corresponding output data;

[0147] The calculation submodule is used to input the output data into the probability calculation layer to obtain the conversion probability.

[0148] Optionally, in one feasible embodiment of the present invention, the data acquisition module includes:

[0149] The raw data acquisition submodule is used to acquire raw user data and raw advertising path data;

[0150] The cleaning submodule is used to clean the original user data and the original advertising path data to obtain user data and advertising path data.

[0151] Optionally, in one feasible embodiment of the present invention, the encoding module is implemented based on a bidirectional long short-term memory network.

[0152] The advertising attribution device 200 provided in this application embodiment can implement each process of the advertising attribution method corresponding to Embodiment 1 and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0153] This invention also provides a computer device, including a memory and a processor. The memory stores a computer program, which, when run on the processor, executes the advertising attribution method as described in Embodiment 1.

[0154] This invention also provides a computer-readable storage medium storing a computer program that, when run on a processor, executes the advertising attribution method as described in Embodiment 1.

[0155] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, as an alternative implementation, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0156] In addition, the functional modules or units in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0157] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion 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 smartphone, 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 invention. 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.

[0158] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An attribution method for advertising, characterized in that, include: Acquire user data and advertising path data, wherein the advertising path data includes touchpoint data of multiple touchpoints, the touchpoint data includes identification data and channel data, and the touchpoint data of multiple touchpoints is sorted according to the user trigger time; The encoding module based on the pre-trained attribution model encodes the identification data of each touch point to obtain the identification data representation of each touch point; Based on the attention module of the attribution model, a weight is generated for each touchpoint, wherein the greater the difference between the user trigger time of the touchpoint and the user trigger time of the last touchpoint, the lower the weight of the touchpoint. Based on the identifier data representation and weight of each touchpoint, the path representation of the advertising path data is obtained; The user data and the channel data of the multiple touchpoints are input into the second acquisition module of the attribution model to obtain a data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output the data feature representation. The path representation and the data feature representation are input into the prediction module of the attribution model to obtain the conversion probability.

2. The attribution method for advertising according to claim 1, characterized in that, The attribution model also includes a splicing module, a flattening module, and a fully connected module; The path representation of the advertising path data, obtained based on the identifier data representation and weight of each touchpoint, includes: Calculate the product of the identifier data representation and the weight of the contact point to obtain the weighted data representation of the contact point; The weighted data representation of each contact point is input into the splicing module to obtain the first splicing result; The first stitching result is input into the flattening module to obtain the flattening result; The flattening result is input into the fully connected module to obtain the path representation of the advertising path data.

3. The attribution method for advertising according to claim 1, characterized in that, The second acquisition module includes an embedding layer, a first splicing layer, and a first fully connected layer; The step of inputting the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain data feature representation includes: The user data and the channel data of the multiple touchpoints are respectively input into the embedding layer to obtain the embedding vector corresponding to the user data and the embedding vector corresponding to the channel data of each touchpoint. All the embedding vectors are input into the first concatenation layer to obtain the second concatenation result; All the second concatenation results are input into the first fully connected layer to obtain the data feature representation.

4. The attribution method for advertising according to claim 1, characterized in that, The prediction module includes a second splicing layer, a second fully connected layer, and a probability calculation layer; The step of inputting the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability includes: The path representation and the data feature representation are input into the second splicing layer to obtain the third splicing result; The third splicing result is input into the second fully connected layer to obtain the corresponding output data; The output data is input into the probability calculation layer to obtain the conversion probability.

5. The attribution method for advertising according to claim 1, characterized in that, The acquisition of user data and advertising path data includes: Obtain raw user data and raw ad path data; The original user data and the original advertising path data are cleaned to obtain user data and advertising path data.

6. The attribution method for advertising according to claim 1, characterized in that, The encoding module is implemented based on a bidirectional long short-term memory network.

7. An attribution device for advertising, characterized in that, include: The data acquisition module is used to acquire user data and advertising path data. The advertising path data includes touchpoint data of multiple touchpoints. The touchpoint data includes identification data and channel data. The touchpoint data of multiple touchpoints is sorted according to the user trigger time. The encoding module is used to encode the identification data of each of the touch points based on the pre-trained attribution model to obtain the identification data representation of each of the touch points; The generation module is used to generate the weight of each touch point based on the attention module of the attribution model, wherein the greater the difference between the user trigger time of the touch point and the user trigger time of the last touch point, the lower the weight of the touch point. The path representation acquisition module is used to obtain the path representation of the advertising path data based on the identifier data representation and weight of each touchpoint; The feature representation acquisition module is used to input the user data and the channel data of the multiple touchpoints into the second acquisition module of the attribution model to obtain the data feature representation. The second acquisition module is used to embed the user data and the channel data of the multiple touchpoints into corresponding embedding vectors, and to concatenate all the embedding vectors to output the data feature representation. The prediction module is used to input the path representation and the data feature representation into the prediction module of the attribution model to obtain the conversion probability.

8. The advertising attribution device according to claim 7, characterized in that, The attribution model also includes a splicing module, a flattening module, and a fully connected module; The path represents the acquisition module, including: The product calculation submodule is used to calculate the product of the identification data representation and the weight of the touch point to obtain the weighted data representation of the touch point; The first splicing submodule is used to input the weighted data representation of each contact point into the splicing module to obtain the first splicing result; The first flattening processing submodule is used to input the first splicing result into the flattening module to obtain the flattening result; The fully connected processing submodule is used to input the flattening result into the fully connected module to obtain the path representation of the advertising path data.

9. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer program that, when executed on the processor, performs the attribution method for advertising as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed on a processor, performs the attribution method for advertising as described in any one of claims 1-6.

Images