Material data processing method and device based on multi-target value learning model, electronic equipment and storage medium

Abstract

The application provides a material data processing method and device based on a multi-target value learning model, electronic equipment and a storage medium. The material data processing method based on the multi-target value learning model comprises the following steps: calculating a first revenue score of each preselected material and a second revenue score of each preselected material based on a multi-target learning model; calculating a recommendation score of each preselected material based on the multi-target learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, a first value improvement probability of each preselected material, a second value improvement probability of each preselected material and a click rate of each preselected material; and determining an optimal recommended material based on the recommendation score of each preselected material. The application comprehensively considers the revenue of preselected materials to multiple business products, so that the optimal material can be determined more accurately.

Description

Technical Field

[0001] This application relates to the field of computer science, and more specifically, to a method, apparatus, electronic device, and storage medium for processing material data based on a multi-objective value learning model. Background Technology

[0002] Currently, when recommending content to user terminals, the optimal content is usually determined based on the benefits that the content brings to a specific business product. For example, for content A and content B, if content A brings a benefit of 2 to business product A1 and content B brings a benefit of 4 to business product A1, then content B is the optimal content.

[0003] However, the above method has the following drawbacks: Materials can be applied to multiple business products, meaning the revenue of a single business product cannot accurately reflect whether the material is optimal. For example, materials A and B can both be applied to business products A2 and A1. In this case, if material A brings a revenue of 2 to business product A1 and a revenue of 4 to business product A2, while material B brings a revenue of 4 to business product A1 and a revenue of 1 to business product A2, then material B is not the optimal material. Therefore, the existing technology suffers from the defect of inaccurate calculation of the optimal material. Summary of the Invention

[0004] The purpose of this application is to provide a method, apparatus, electronic device, and storage medium for processing material data based on a multi-objective value learning model. This method calculates the recommendation score for each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material, thereby determining the optimal recommended material. Compared with existing technologies, this application comprehensively considers the revenue of pre-selected materials for multiple business products, thus enabling more accurate determination of the optimal material.

[0005] In a first aspect, the present invention provides a method for processing material data based on a multi-objective value learning model, wherein the method includes:

[0006] Obtain user data and attribute data for several pre-selected materials;

[0007] The user data and attribute data of several of the pre-selected materials are input into a multi-objective learning model;

[0008] The click-through rate of each of the pre-selected materials is calculated based on the multi-objective learning model.

[0009] calculate a first value promotion probability of each of the preselected materials and a second value promotion probability of each of the preselected materials based on the multi-objective learning model, wherein the first value promotion probability represents a probability that a first business product value can be improved after a user clicks the preselected material, and the second value promotion probability represents a probability that a second business product value can be improved after the user clicks the preselected material;

[0010] calculate a first revenue score of each of the preselected materials and a second revenue score of each of the preselected materials based on the multi-objective learning model;

[0011] calculate a recommendation score of each of the preselected materials based on the multi-objective learning model, the first revenue score of each of the preselected materials, the second revenue score of each of the preselected materials, the first value promotion probability of each of the preselected materials, the second value promotion probability of each of the preselected materials, and a click rate of each of the preselected materials;

[0012] determine an optimal recommended material based on the recommendation score of each of the preselected materials.

[0013] In the first aspect of the present application, by obtaining user data and attribute data of a plurality of preselected materials, the user data and the attribute data of the plurality of preselected materials can be input into a multi-objective learning model, and then the click rate of each of the plurality of preselected materials can be calculated based on the multi-objective learning model, and then the first value promotion probability of each of the preselected materials and the second value promotion probability of each of the preselected materials can be calculated based on the multi-objective learning model, wherein the first value promotion probability represents a probability that a first business product value can be improved after a user clicks the preselected material, and the second value promotion probability represents a probability that a second business product value can be improved after the user clicks the preselected material, and then the first revenue score of each of the preselected materials and the second revenue score of each of the preselected materials can be calculated based on the multi-objective learning model, and then the recommendation score of each of the preselected materials can be calculated based on the multi-objective learning model, the first revenue score of each of the preselected materials, the second revenue score of each of the preselected materials, the first value promotion probability of each of the preselected materials, the second value promotion probability of each of the preselected materials, and the click rate of each of the preselected materials, and then the optimal recommended material can be determined based on the recommendation score of each of the preselected materials.

[0014] Compared with the prior art, the present application comprehensively considers the revenue of preselected materials on multiple business products, so that the optimal material can be more accurately determined.

[0015] In an optional implementation, the method further comprises:

[0016] The first revenue amount and the second revenue amount of each preselected material are calculated based on the multi-objective learning model, wherein the first revenue amount represents actual income brought by the preselected material to the first business product, and the second revenue amount represents actual income brought by the preselected material to the second business product.

[0017] The first revenue score of each preselected material is obtained by normalizing the first revenue amount of each preselected material based on the multi-objective learning model.

[0018] The second revenue score of each preselected material is obtained by normalizing the second revenue amount of each preselected material based on the multi-objective learning model.

[0019] In the above optional implementation, the first revenue amount and the second revenue amount of each preselected material can be calculated based on the multi-objective learning model, and then the first revenue score of each preselected material can be obtained by normalizing the first revenue amount of each preselected material based on the multi-objective learning model, and the second revenue score of each preselected material can be obtained by normalizing the second revenue amount of each preselected material based on the multi-objective learning model.

[0020] In an optional implementation, before the recommendation score of each preselected material is calculated based on the multi-objective learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, and the click rate of each preselected material, the method further comprises:

[0021] It is determined whether the weight of the first business product and the weight of the second business product exist.

[0022] When the weight of the first business product and the weight of the second business product both exist, the recommendation score of each preselected material is calculated based on the multi-objective learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, the weight of the first business product, the weight of the second business product, and the click rate of each preselected material.

[0023] In the optional implementation, when the weight of the first business product and the weight of the second business product both exist, the recommendation score of each preselected material can be calculated based on the multi-objective learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, the weight of the first business product, the weight of the second business product, and the click rate of each preselected material.

[0024] In the optional implementation, the recommendation score of each preselected material is calculated based on the multi-objective learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, the weight of the first business product, the weight of the second business product, and the click rate of each preselected material, and the corresponding calculation formula is:

[0025] S=A*(P1*A1*B1+P2*A2*B2);

[0026] wherein S represents the recommendation score of the preselected material, A represents the click rate of the preselected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the preselected material, B1 represents the first value improvement probability of the preselected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the preselected material, and B2 represents the second value improvement probability of the preselected material.

[0027] In the optional implementation, the score of each preselected material can be accurately calculated by the calculation formula.

[0028] In a second aspect, the application provides a material data processing device based on a multi-objective value learning model, wherein the device comprises:

[0029] an acquisition module configured to acquire user data and attribute data of a plurality of preselected materials;

[0030] an input module configured to input the user data and the attribute data of the plurality of preselected materials into a multi-objective learning model;

[0031] a first calculation module configured to calculate a click rate of each preselected material in the plurality of preselected materials based on the multi-objective learning model;

[0032] a second computing module configured to calculate, based on the multi-objective learning model, a first value improvement probability of each of the preselected materials and a second value improvement probability of each of the preselected materials, wherein the first value improvement probability represents a probability that a first business product value can be improved after a user clicks the preselected material, and the second value improvement probability represents a probability that a second business product value can be improved after the user clicks the preselected material;

[0033] a third computing module configured to calculate, based on the multi-objective learning model, a first revenue score of each of the preselected materials and a second revenue score of each of the preselected materials;

[0034] a fourth computing module configured to calculate, based on the multi-objective learning model, the first revenue score of each of the preselected materials, the second revenue score of each of the preselected materials, the first value improvement probability of each of the preselected materials, the second value improvement probability of each of the preselected materials, and a click rate of each of the preselected materials, a recommendation score of each of the preselected materials;

[0035] a determining module configured to determine an optimal recommended material based on the recommendation score of each of the preselected materials.

[0036] In the second aspect, by obtaining user data and attribute data of a plurality of preselected materials, the user data and the attribute data of the plurality of preselected materials can be input into a multi-objective learning model, and then the click rate of each of the plurality of preselected materials can be calculated based on the multi-objective learning model, and then the first value improvement probability of each of the preselected materials and the second value improvement probability of each of the preselected materials can be calculated based on the multi-objective learning model, wherein the first value improvement probability represents a probability that a first business product value can be improved after a user clicks the preselected material, and the second value improvement probability represents a probability that a second business product value can be improved after the user clicks the preselected material, and then the first revenue score of each of the preselected materials and the second revenue score of each of the preselected materials can be calculated based on the multi-objective learning model, and then the recommendation score of each of the preselected materials can be calculated based on the multi-objective learning model, the first revenue score of each of the preselected materials, the second revenue score of each of the preselected materials, the first value improvement probability of each of the preselected materials, the second value improvement probability of each of the preselected materials, and the click rate of each of the preselected materials, and then the optimal recommended material can be determined based on the recommendation score of each of the preselected materials.

[0037] Compared with the prior art, the application comprehensively considers the revenue of the preselected materials on multiple business products, so that the optimal material can be more accurately determined.

[0038] In an optional embodiment, the third calculation module comprises:

[0039] The calculation sub-module is configured to calculate a first revenue amount and a second revenue amount of each of the preselected materials based on the multi-objective learning model, wherein the first revenue amount represents actual income brought by the preselected material to the first business product, and the second revenue amount represents actual income brought by the preselected material to the second business product.

[0040] The normalization processing module is configured to perform normalization processing on the first revenue amount of each of the preselected materials based on the multi-objective learning model to obtain a first revenue score of each of the preselected materials.

[0041] The normalization processing module is further configured to perform normalization processing on the second revenue amount of each of the preselected materials based on the multi-objective learning model to obtain a second revenue score of each of the preselected materials.

[0042] In the above optional embodiment, based on the multi-objective learning model, the first revenue amount and the second revenue amount of each of the preselected materials can be calculated, and then the first revenue amount of each of the preselected materials can be normalized based on the multi-objective learning model to obtain the first revenue score of each of the preselected materials, and the second revenue amount of each of the preselected materials can be normalized based on the multi-objective learning model to obtain the second revenue score of each of the preselected materials.

[0043] In an optional embodiment, the device further comprises:

[0044] The judgment module is configured to judge whether the weight of the first business product and the weight of the second business product exist.

[0045] The fourth calculation module is further configured to, when the weight of the first business product and the weight of the second business product both exist, calculate a recommendation score of each of the preselected materials based on the multi-objective learning model, the first revenue score of each of the preselected materials, the second revenue score of each of the preselected materials, the first value improvement probability of each of the preselected materials, the second value improvement probability of each of the preselected materials, the weight of the first business product, the weight of the second business product, and the click rate of each of the preselected materials.

[0046] In the optional implementation, when the weight of the first business product and the weight of the second business product both exist, the recommendation score of each preselected material can be calculated based on the multi-target learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, the weight of the first business product, the weight of the second business product, and the click rate of each preselected material.

[0047] In the optional implementation, the calculation formula used by the fourth calculation module is:

[0048] S=A*(P1*A1*B1+P2*A2*B2)

[0049] wherein S represents the recommendation score of the preselected material, A represents the click rate of the preselected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the preselected material, B1 represents the first value improvement probability of the preselected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the preselected material, and B2 represents the second value improvement probability of the preselected material.

[0050] In the optional implementation, the score of each preselected material can be accurately calculated by using the above calculation formula.

[0051] In a third aspect, the present application provides an electronic device, comprising:

[0052] a processor; and

[0053] a memory configured to store machine-readable instructions, which, when executed by the processor, perform the material data processing method based on the multi-target value learning model as described in any one of the preceding embodiments.

[0054] The electronic device of the third aspect of the present application can obtain user data and attribute data of a plurality of preselected materials by executing the material data processing method based on the multi-target value learning model, and input the user data and the attribute data of the plurality of preselected materials into the multi-target learning model. Then, the click rate of each preselected material in the plurality of preselected materials can be calculated based on the multi-target learning model. Then, the first value improvement probability of each preselected material and the second value improvement probability of each preselected material can be calculated based on the multi-target learning model. The first value improvement probability represents the probability that the first business product value can be improved after the user clicks on the preselected material. The second value improvement probability represents the probability that the second business product value can be improved after the user clicks on the preselected material. Then, the first revenue score of each preselected material and the second revenue score of each preselected material can be calculated based on the multi-target learning model. Then, the recommendation score of each preselected material can be calculated based on the multi-target learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, and the click rate of each preselected material. Then, the optimal recommended material can be determined based on the recommendation score of each preselected material.

[0055] Compared with the prior art, the present application comprehensively considers the revenue of the preselected material to multiple business products, so that the optimal material can be more accurately determined.

[0056] In a fourth aspect, the present application provides a storage medium storing a computer program, wherein the computer program is executed by a processor to perform the material data processing method based on the multi-target value learning model according to any one of the preceding embodiments.

[0057] The storage medium of the fourth aspect of the present application can obtain user data and attribute data of a plurality of preselected materials by executing the material data processing method based on the multi-target value learning model, input the user data and the attribute data of the plurality of preselected materials into the multi-target learning model, and then calculate the click rate of each preselected material in the plurality of preselected materials based on the multi-target learning model. Then, the first value improvement probability of each preselected material and the second value improvement probability of each preselected material can be calculated based on the multi-target learning model, wherein the first value improvement probability represents the probability that the first business product value can be improved after the user clicks the preselected material, and the second value improvement probability represents the probability that the second business product value can be improved after the user clicks the preselected material. Then, the first revenue score of each preselected material and the second revenue score of each preselected material can be calculated based on the multi-target learning model. Then, the recommendation score of each preselected material can be calculated based on the multi-target learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, and the click rate of each preselected material. Then, the optimal recommended material can be determined based on the recommendation score of each preselected material.

[0058] Compared with the prior art, the present application comprehensively considers the revenue of the preselected material to multiple business products, so that the optimal material can be more accurately determined. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0060] Figure 1 is a flow diagram of a material data processing method based on a multi-target value learning model disclosed by an embodiment of the present application;

[0061] Figure 2 is a structural diagram of a material data processing device based on a multi-target value learning model disclosed by an embodiment of the present application;

[0062] Figure 3 is a structural diagram of an electronic device disclosed by an embodiment of the present application. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

[0064] Embodiment one

[0065] Please refer to Figure 1 , Figure 1 is a flowchart of a material data processing method based on a multi-target value learning model disclosed in the embodiments of the present application, as Figure 1 shown, the method of the present application comprises the following steps:

[0066] 101, obtaining user data and attribute data of a plurality of preselected materials;

[0067] 102, inputting the user data and the attribute data of the plurality of preselected materials into a multi-target learning model;

[0068] 103, calculating the click rate of each preselected material in the plurality of preselected materials based on the multi-target learning model;

[0069] 104, calculating the first value improvement probability of each preselected material and the second value improvement probability of each preselected material based on the multi-target learning model, wherein the first value improvement probability represents the probability that the user can improve the value of the first business product after clicking the preselected material, and the second value improvement probability represents the probability that the user can improve the value of the second business product after clicking the preselected material;

[0070] 105, calculating the first revenue score of each preselected material and the second revenue score of each preselected material based on the multi-target learning model;

[0071] 106, calculating the recommendation score of each preselected material based on the multi-target learning model, the first revenue score of each preselected material, the second revenue score of each preselected material, the first value improvement probability of each preselected material, the second value improvement probability of each preselected material, and the click rate of each preselected material;

[0072] 107, determining the optimal recommended material based on the recommendation score of each preselected material.

[0073] In this embodiment, by acquiring user data and attribute data of several pre-selected materials, the user data and attribute data of several pre-selected materials can be input into a multi-objective learning model. The click-through rate (CTR) of each pre-selected material can be calculated based on the multi-objective learning model. Furthermore, a first value enhancement probability and a second value enhancement probability of each pre-selected material can be calculated based on the multi-objective learning model. The first value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a first business product, and the second value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a second business product. A first revenue score and a second revenue score of each pre-selected material can be calculated based on the multi-objective learning model. Finally, a recommendation score for each pre-selected material can be calculated based on the multi-objective learning model, the first revenue score, the second revenue score, the first value enhancement probability, the second value enhancement probability, and the CTR of each pre-selected material. Finally, the optimal recommended material can be determined based on the recommendation score of each pre-selected material.

[0074] Compared with existing technologies, this application comprehensively considers the benefits of pre-selected materials to multiple business products, thereby enabling more accurate determination of the optimal materials.

[0075] In this embodiment of the application, for step 101, the pre-selected material can be information prepared to be pushed to the user terminal. For example, the pre-selected material can be a fund purchase advertisement that is accurately pushed to the user terminal.

[0076] In this embodiment, for step 101, the user data includes data generated by the user when using the user terminal, such as the product browsing history reserved by the user on the user terminal, wherein the product browsing history includes information such as the type, quantity, and price of the product. Furthermore, the attribute data of the pre-selected material may include information such as the type and price of the pre-selected material.

[0077] In this embodiment of the application, for step 102, the multi-objective learning model can be the open-source multi-task learning model MMoE provided by Google.

[0078] In this embodiment, regarding step 103, the click-through rate (CTR) of the pre-selected material refers to the ratio of user clicks on the pre-selected material to the number of times the pre-selected material is delivered. For example, if the pre-selected material is delivered 100 times and the user clicks on it 20 times, the CTR is 20%. Furthermore, the multi-objective learning model can calculate the CTR of the pre-selected material through pre-training. For example, for pre-selected material A, the multi-objective learning model can predict the number of clicks based on user data, thereby calculating the CTR based on the simulated delivery count and the number of clicks.

[0079] In this embodiment of the application, regarding step 104, the multi-objective learning model can calculate the first value enhancement probability and the second value enhancement probability of each pre-selected material. For example, the multi-objective learning model can simulate the delivery of pre-selected material A, and then predict the user's order price based on user data. When the user's order price is higher than the preset price of the first business product, it is considered that the user can bring higher value to the first business product. The multi-objective learning model can predict multiple user order prices based on different conditions, such as obtaining user order price A, user order price B, user order price C, and user order price D. Furthermore, if both user order price A and user order price B are higher than the preset price of the first business product, the first value enhancement probability is 50%.

[0080] In an optional implementation, step 105: Calculate the first revenue score and the second revenue score for each pre-selected material based on a multi-objective learning model, including the above:

[0081] The first revenue amount and the second revenue amount of each pre-selected material are calculated based on a multi-objective learning model. The first revenue amount represents the actual revenue brought by the pre-selected material to the first business product, and the second revenue amount represents the actual revenue brought by the pre-selected material to the second business product.

[0082] The first revenue amount of each pre-selected material is normalized based on the multi-objective learning module to obtain the first revenue score of each pre-selected material.

[0083] The second revenue amount of each pre-selected material is normalized based on the multi-objective learning module to obtain the second revenue score of each pre-selected material.

[0084] In the above optional implementation, the multi-objective learning model can calculate the first revenue amount and the second revenue amount based on multiple user order prices. For example, for the first business product, the first revenue amount is calculated based on the value and probability corresponding to user order price A, user order price B, user order price C, and user order price D.

[0085] In the above optional implementation, normalizing the first revenue amount and the second revenue amount is to make the first revenue amount and the second revenue amount dimensionless.

[0086] In the above optional implementation, the first revenue amount and the second revenue amount of each pre-selected material can be calculated based on the multi-objective learning model. Then, the first revenue amount of each pre-selected material can be normalized based on the multi-objective learning module to obtain the first revenue score of each pre-selected material. And the second revenue amount of each pre-selected material can be normalized based on the multi-objective learning module to obtain the second revenue score of each pre-selected material.

[0087] In an optional implementation, before the step of calculating the recommendation score for each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material, the method of this application embodiment further includes the following:

[0088] Determine whether the weights of the first business product and the second business product exist;

[0089] When both the weights of the first business product and the weights of the second business product exist, the recommendation score of each pre-selected material is calculated based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material.

[0090] In the above optional implementation, when both the weight of the first business product and the weight of the second business product exist, the recommendation score of each pre-selected material can be calculated based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material.

[0091] In the above optional implementation, the weights of the first business product and the second business product can be adjusted according to the needs of the application scenario. For example, for application scenario A, the weight of the first business product is 0.5 and the weight of the second business product is 0.5. On the other hand, for application scenario B, the weight of the first business product is 0.6 and the weight of the second business product is 0.4.

[0092] In an optional implementation, based on a multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material, the recommendation score of each pre-selected material is calculated, and the corresponding calculation formula is:

[0093] S = A*(P1*A1*B1+P2*A2*B2);

[0094] Where S represents the recommendation score of the pre-selected material, A represents the click-through rate of the pre-selected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the pre-selected material, B1 represents the first value enhancement probability of the pre-selected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the pre-selected material, and B2 represents the second value enhancement probability of the pre-selected material.

[0095] In the above optional implementation, the score of each pre-selected material can be accurately calculated using the above calculation formula.

[0096] Example 2

[0097] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a material data processing device based on a multi-objective value learning model disclosed in an embodiment of this application, as shown below. Figure 2 As shown, the apparatus in this embodiment includes the following functional modules:

[0098] The acquisition module 201 is used to acquire user data and attribute data of several pre-selected materials;

[0099] Input module 202 is used to input user data and attribute data of several pre-selected materials into the multi-objective learning model;

[0100] The first calculation module 203 is used to calculate the click-through rate of each of several pre-selected materials based on a multi-objective learning model;

[0101] The second calculation module 204 is used to calculate the first value enhancement probability and the second value enhancement probability of each pre-selected material based on a multi-objective learning model. The first value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the first business product, and the second value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the second business product.

[0102] The third calculation module 205 is used to calculate the first revenue score and the second revenue score of each pre-selected material based on a multi-objective learning model.

[0103] The fourth calculation module 206 is used to calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material.

[0104] Module 207 is used to determine the optimal recommended material based on the recommendation score of each pre-selected material.

[0105] In this embodiment, by acquiring user data and attribute data of several pre-selected materials, the user data and attribute data of several pre-selected materials can be input into a multi-objective learning model. The click-through rate (CTR) of each pre-selected material can be calculated based on the multi-objective learning model. Furthermore, a first value enhancement probability and a second value enhancement probability of each pre-selected material can be calculated based on the multi-objective learning model. The first value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a first business product, and the second value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a second business product. A first revenue score and a second revenue score of each pre-selected material can be calculated based on the multi-objective learning model. Finally, a recommendation score for each pre-selected material can be calculated based on the multi-objective learning model, the first revenue score, the second revenue score, the first value enhancement probability, the second value enhancement probability, and the CTR of each pre-selected material. Finally, the optimal recommended material can be determined based on the recommendation score of each pre-selected material.

[0106] Compared with existing technologies, this application comprehensively considers the benefits of pre-selected materials to multiple business products, thereby enabling more accurate determination of the optimal materials.

[0107] In an optional implementation, the third calculation module includes the following sub-modules:

[0108] The calculation submodule is used to calculate the first revenue amount and the second revenue amount for each pre-selected material based on a multi-objective learning model. The first revenue amount represents the actual revenue brought by the pre-selected material to the first business product, and the second revenue amount represents the actual revenue brought by the pre-selected material to the second business product.

[0109] The normalization module is used to normalize the first revenue amount of each pre-selected material based on the multi-objective learning module, so as to obtain the first revenue score of each pre-selected material.

[0110] The normalization module is also used to normalize the second revenue amount of each pre-selected material based on the multi-objective learning module, so as to obtain the second revenue score of each pre-selected material.

[0111] In the above optional implementation, the first revenue amount and the second revenue amount of each pre-selected material can be calculated based on the multi-objective learning model. Then, the first revenue amount of each pre-selected material can be normalized based on the multi-objective learning module to obtain the first revenue score of each pre-selected material. And the second revenue amount of each pre-selected material can be normalized based on the multi-objective learning module to obtain the second revenue score of each pre-selected material.

[0112] In optional embodiments, the apparatus of this application further includes the following functional modules:

[0113] The judgment module is used to determine whether the weight of the first business product and the weight of the second business product exist.

[0114] Furthermore, the fourth calculation module is also used to calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material when both the weight of the first business product and the weight of the second business product exist.

[0115] In the above optional implementation, when both the weight of the first business product and the weight of the second business product exist, the recommendation score of each pre-selected material can be calculated based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material.

[0116] In an optional implementation, the fourth calculation module uses the following formula:

[0117] S = A*(P1*A1*B1+P2*A2*B2);

[0118] Where S represents the recommendation score of the pre-selected material, A represents the click-through rate of the pre-selected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the pre-selected material, B1 represents the first value enhancement probability of the pre-selected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the pre-selected material, and B2 represents the second value enhancement probability of the pre-selected material.

[0119] In the above optional implementation, the score of each pre-selected material can be accurately calculated using the above calculation formula.

[0120] Example 3

[0121] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application, such as... Figure 3 As shown, the electronic device in this application embodiment includes:

[0122] Processor 301; and

[0123] The memory 302 is configured to store machine-readable instructions that, when executed by the processor 301, perform a material data processing method based on a multi-objective value learning model as described in any of the foregoing embodiments.

[0124] The electronic device in this embodiment of the application, by executing a material data processing method based on a multi-objective value learning model, can acquire user data and attribute data of several pre-selected materials, and input the user data and attribute data of several pre-selected materials into a multi-objective learning model. Then, it can calculate the click-through rate of each pre-selected material based on the multi-objective learning model, and further calculate the first value enhancement probability and the second value enhancement probability of each pre-selected material based on the multi-objective learning model. The first value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a first business product, and the second value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a second business product. Then, it can calculate the first revenue score and the second revenue score of each pre-selected material based on the multi-objective learning model, and further calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score, the second revenue score, the first value enhancement probability, the second value enhancement probability, and the click-through rate of each pre-selected material. Finally, it can determine the optimal recommended material based on the recommendation score of each pre-selected material.

[0125] Compared with existing technologies, this application comprehensively considers the benefits of pre-selected materials to multiple business products, thereby enabling more accurate determination of the optimal materials.

[0126] Example 4

[0127] This application provides a storage medium storing a computer program, which is executed by a processor as described in any of the foregoing embodiments, a material data processing method based on a multi-objective value learning model.

[0128] The storage medium in this embodiment of the application, by executing a material data processing method based on a multi-objective value learning model, can acquire user data and attribute data of several pre-selected materials, and input the user data and attribute data of several pre-selected materials into a multi-objective learning model. Then, it can calculate the click-through rate of each pre-selected material based on the multi-objective learning model, and further calculate the first value enhancement probability and the second value enhancement probability of each pre-selected material. The first value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a first business product, and the second value enhancement probability represents the probability that clicking on a pre-selected material will increase the value of a second business product. Then, it can calculate the first revenue score and the second revenue score of each pre-selected material based on the multi-objective learning model, and further calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score, the second revenue score, the first value enhancement probability, the second value enhancement probability, and the click-through rate of each pre-selected material. Finally, it can determine the optimal recommended material based on the recommendation score of each pre-selected material.

[0129] Compared with existing technologies, this application comprehensively considers the benefits of pre-selected materials to multiple business products, thereby enabling more accurate determination of the optimal materials.

[0130] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interface; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0131] Furthermore, 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.

[0132] Furthermore, the functional modules in the various embodiments of this application 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.

[0133] It should be noted that if a function is implemented as a software module 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 a 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 of 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.

[0134] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0135] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for processing source data based on a multi-objective value learning model, wherein, The method includes: Obtain user data and attribute data for several pre-selected materials; The user data and attribute data of several of the pre-selected materials are input into a multi-objective learning model; The click-through rate of each of the pre-selected materials is calculated based on the multi-objective learning model. The multi-objective learning model calculates a first value enhancement probability and a second value enhancement probability for each of the pre-selected materials. The first value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the first business product, and the second value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the second business product. The multi-objective learning model is used to simulate the delivery of the pre-selected materials, predict the user's order price based on the user data and different conditions, and determine the first value enhancement probability based on the user's order price. Based on the multi-objective learning model, calculate the first revenue score and the second revenue score for each of the pre-selected materials; Based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material, the recommendation score of each pre-selected material is calculated. The optimal recommended material is determined based on the recommendation score of each of the pre-selected materials; Before calculating the recommendation score for each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material, the method further includes: Determine whether the weights of the first business product and the second business product exist; When both the weight of the first business product and the weight of the second business product exist, the recommendation score of each pre-selected material is calculated based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material. The weight of the first business product and the weight of the second business product can be adjusted according to the needs of the application scenario.

2. The method as described in claim 1, characterized in that, The calculation of the first revenue score and the second revenue score for each of the pre-selected materials based on the multi-objective learning model includes: Based on the multi-objective learning model, a first revenue amount and a second revenue amount are calculated for each of the pre-selected materials, wherein the first revenue amount represents the actual revenue brought by the pre-selected materials to the first business product, and the second revenue amount represents the actual revenue brought by the pre-selected materials to the second business product; Based on the multi-objective learning model, the first revenue amount of each of the pre-selected materials is normalized to obtain the first revenue score of each of the pre-selected materials. The second revenue amount of each of the pre-selected materials is normalized based on the multi-objective learning model to obtain the second revenue score of each of the pre-selected materials.

3. The method as described in claim 1, characterized in that, The recommendation score for each pre-selected material is calculated based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material. The corresponding calculation formula is as follows: S=A*(P1*A1*B1+ P2*A2*B2); Wherein, S represents the recommendation score of the pre-selected material, A represents the click-through rate of the pre-selected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the pre-selected material, B1 represents the first value enhancement probability of the pre-selected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the pre-selected material, and B2 represents the second value enhancement probability of the pre-selected material.

4. A material data processing device based on a multi-objective value learning model, wherein, The device includes: The acquisition module is used to acquire user data and attribute data of several pre-selected materials; The input module is used to input the user data and attribute data of several of the pre-selected materials into the multi-objective learning model; The first calculation module is used to calculate the click-through rate of each of the pre-selected materials in a plurality of pre-selected materials based on the multi-objective learning model; The second calculation module is used to calculate a first value enhancement probability and a second value enhancement probability for each of the pre-selected materials based on the multi-objective learning model. The first value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the first business product, and the second value enhancement probability represents the probability that clicking on the pre-selected material will increase the value of the second business product. The multi-objective learning model is used to simulate the delivery of the pre-selected materials, predict the user's order price based on the user data and different conditions, and determine the first value enhancement probability based on the user's order price. The third calculation module is used to calculate the first revenue score and the second revenue score of each of the pre-selected materials based on the multi-objective learning model. The fourth calculation module is used to calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, and the click-through rate of each pre-selected material. The determination module is used to determine the optimal recommended material based on the recommendation score of each of the pre-selected materials; The device further includes: The judgment module is used to determine whether the weights of the first business product and the second business product exist. Furthermore, the fourth calculation module is also used to calculate the recommendation score of each pre-selected material based on the multi-objective learning model, the first revenue score of each pre-selected material, the second revenue score of each pre-selected material, the first value enhancement probability of each pre-selected material, the second value enhancement probability of each pre-selected material, the weight of the first business product, the weight of the second business product, and the click-through rate of each pre-selected material when both the weight of the first business product and the weight of the second business product exist. The weight of the first business product and the weight of the second business product can be adjusted according to the needs of the application scenario.

5. The apparatus as described in claim 4, characterized in that, The third computing module includes: The calculation submodule is used to calculate the first revenue amount and the second revenue amount for each of the pre-selected materials based on the multi-objective learning model, wherein the first revenue amount represents the actual revenue brought by the pre-selected materials to the first business product, and the second revenue amount represents the actual revenue brought by the pre-selected materials to the second business product; The normalization processing module is used to normalize the first revenue amount of each of the pre-selected materials based on the multi-objective learning model to obtain the first revenue score of each of the pre-selected materials. The normalization processing module is further used to normalize the second revenue amount of each of the pre-selected materials based on the multi-objective learning model, so as to obtain the second revenue score of each of the pre-selected materials.

6. The apparatus as claimed in claim 5, characterized in that, The calculation formula used by the fourth calculation module is: S=A*(P1*A1*B1+ P2*A2*B2); Wherein, S represents the recommendation score of the pre-selected material, A represents the click-through rate of the pre-selected material, P1 represents the weight of the first business product, A1 represents the first revenue score of the pre-selected material, B1 represents the first value enhancement probability of the pre-selected material, P2 represents the weight of the second business product, A2 represents the second revenue score of the pre-selected material, and B2 represents the second value enhancement probability of the pre-selected material.

7. An electronic device, characterized in that, include: processor; as well as The memory is configured to store machine-readable instructions that, when executed by the processor, perform the material data processing method based on a multi-objective value learning model as described in any one of claims 1-3.

8. A storage medium, characterized in that, The storage medium stores a computer program, which is executed by a processor as described in any one of claims 1-3, for processing material data based on a multi-objective value learning model.

Images