Data processing method and device, computer equipment and storage medium

Abstract

The invention discloses a data processing method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining interest expression information representing the interest of obtaining respective corresponding promotion objects expressed by different promoted objects; obtaining first object feature information of a promoted object and second object feature information of a promoted object; according to the first object feature information and the second object feature information, predicting a first probability of selecting each promotion object by each promoted object through a first probability prediction model; according to the first probability that each promoted object is selected by each promoted object and the number of the promoted objects, predicting the number of expected obtaining objects; predicting the total acquisition cost of the interest performance information corresponding to the target promotion object according to the first probability of selecting the target promotion object in the promotion objects by each promoted object and the acquisition cost of the target promotion object; and performing quality evaluation on the interest performance information according to the expected acquisition object number and the total acquisition cost to obtain a quality evaluation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a data processing method and device, computer equipment and computer readable storage medium. BACKGROUND

[0002] Interest performance information represents the potential promoted object's interest in obtaining the promoted object. This usually occurs before the promoted object begins to consider their acquisition of the promoted object, meaning that the promoted object has already generated a possible intention to obtain the promoted object. Interest performance information can be information from various channels, such as web search, social media interaction, industry exhibition, etc. The quality of interest performance information reflects the possibility of the corresponding promoted object obtaining the corresponding promoted object to some extent, the higher the quality of interest performance information, the higher the possibility of the corresponding promoted object obtaining the corresponding promoted object; the higher the possibility of the corresponding promoted object obtaining the corresponding promoted object, the higher the success rate of promoting the corresponding promoted object to it, and the promotion degree of promoting the corresponding promoted object to it can also be set higher to improve the conversion rate of interest performance information to actual promoted objects that obtain the corresponding promoted object, therefore, in order to improve the conversion rate of interest performance information, it is necessary to evaluate the quality of interest performance information, obtain the quality evaluation result indicating the promotion degree of promoting the corresponding promoted object to the to-be-promoted object, and promote the corresponding promoted object to the to-be-promoted object according to the promotion degree. SUMMARY

[0003] The embodiments of the present application provide a data processing method and device, computer equipment and storage medium, which can realize quality evaluation of interest performance information.

[0004] The embodiments of the present application provide a data processing method, comprising:

[0005] Obtain interest performance information of different promoted objects for their respective corresponding promoted objects, the interest performance information representing the interest of different promoted objects in obtaining their respective corresponding promoted objects, and the types of the corresponding promoted objects of different promoted objects being the same;

[0006] Obtain the first object feature information of the promoted object and the second object feature information of the promoted object;

[0007] According to the first object feature information and the second object feature information, a first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object;

[0008] Obtain the number of promoted objects, and according to the first probability of each promoted object selecting each promoted object and the number of promoted objects, predict the expected number of objects.

[0009] obtain an acquisition cost of the target promoted object, and predict a total acquisition cost of the target promoted object corresponding to the interest performance information according to the first probability of each of the promoted objects selecting the target promoted object and the acquisition cost;

[0010] perform quality evaluation on the interest performance information according to the expected acquisition object quantity and the total acquisition cost, and obtain a quality evaluation result, the quality evaluation result being used to indicate a promotion degree of promoting the target promoted object to the promoted objects.

[0011] Correspondingly, an embodiment of the present application further provides a data processing apparatus, comprising:

[0012] a first information obtaining module, configured to obtain interest performance information of different promoted objects for respective corresponding promoted objects, the interest performance information representing acquisition interests of different promoted objects in acquiring respective corresponding promoted objects, and the types of the respective corresponding promoted objects being the same;

[0013] a second information obtaining module, configured to obtain first object feature information of the promoted objects and second object feature information of the promoted objects;

[0014] a probability predicting module, configured to predict a first probability of each of the promoted objects selecting each of the promoted objects according to the first object feature information and the second object feature information through a first probability predicting model;

[0015] a quantity predicting module, configured to obtain a quantity of the promoted objects, and predict an expected acquisition object quantity according to the first probability of each of the promoted objects selecting each of the promoted objects and the quantity of the promoted objects;

[0016] a cost predicting module, configured to obtain an acquisition cost of a target promoted object in the promoted objects, and predict a total acquisition cost of the target promoted object corresponding to the interest performance information according to the first probability of each of the promoted objects selecting the target promoted object and the acquisition cost;

[0017] a quality evaluating module, configured to perform quality evaluation on the interest performance information according to the expected acquisition object quantity and the total acquisition cost, and obtain a quality evaluation result, the quality evaluation result being used to indicate a promotion degree of promoting the target promoted object to the promoted objects.

[0018] Optionally, the target promoted object provides services in multiple gears, the acquisition costs of the target promoted object corresponding to services in different gears are different, and the probability predicting module is further configured to perform:

[0019] predict, by a second probability prediction model, a second probability that each of the promoted objects selects the service of each of the gears according to the first object feature information and the object feature information of the target promoted object corresponding to the service of each of the gears;

[0020] The cost prediction module is specifically configured to perform:

[0021] obtain the acquisition cost of the target promoted object corresponding to the service of each of the gears;

[0022] predict the total acquisition cost of the target promoted object corresponding to the interest performance information according to the first probability that each of the promoted objects selects the target promoted object, the second probability that each of the promoted objects selects the service of each of the gears, and the acquisition cost of the target promoted object corresponding to the service of each of the gears.

[0023] Optionally, the cost prediction module is specifically configured to perform:

[0024] weight the acquisition cost of the target promoted object corresponding to the service of each of the gears by the first probability that each of the promoted objects selects the target promoted object and the second probability that each of the promoted objects selects the service of each of the gears, and perform a weighted sum operation to obtain the total acquisition cost of the target promoted object corresponding to the interest performance information.

[0025] Optionally, the quality evaluation module is specifically configured to perform:

[0026] construct a plane rectangular coordinate system, wherein a first direction axis of the plane rectangular coordinate system represents a quantity, and a second direction axis of the plane rectangular coordinate system represents an acquisition cost;

[0027] map the expected acquisition object quantity and the total acquisition cost to the plane rectangular coordinate system to obtain a coordinate point corresponding to the expected acquisition object quantity and the total acquisition cost;

[0028] perform quality evaluation on the interest performance information according to the position of the coordinate point in the plane rectangular coordinate system to obtain a quality evaluation result.

[0029] Optionally, the quantity prediction module is specifically configured to perform:

[0030] perform an addition operation on the first probability that each of the promoted objects selects each of the promoted objects to obtain a probability sum.

[0031] perform a division operation on the probability sum and the quantity of the promoted objects to obtain the expected acquisition object quantity.

[0032] Optionally, the data processing apparatus further comprises an object promotion module, and the object promotion module is configured to perform:

[0033] promote the target promoted object to the promoted object according to the promotion degree.

[0034] Optionally, the object promotion module is specifically configured to perform:

[0035] obtain a promotion strategy corresponding to the promotion degree;

[0036] promote the target promoted object to the promoted object according to the promotion strategy.

[0037] Correspondingly, an embodiment of the present application further provides a computer device, comprising a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to execute any data processing method provided by the embodiments of the present application.

[0038] Correspondingly, an embodiment of the present application further provides a computer readable storage medium, which is used to store a computer program, and the computer program is loaded by a processor to execute any data processing method provided by the embodiments of the present application.

[0039] Correspondingly, an embodiment of the present application further provides a computer program content, comprising computer programs / instructions, wherein the computer programs / instructions are executed by a processor to implement the steps of the data processing method shown in the embodiments of the present application.

[0040] In the embodiments of the present application, by obtaining interest performance information of different promoted objects for respective corresponding promoted objects, the interest performance information represents the interest of different promoted objects in obtaining respective corresponding promoted objects, wherein the types of the corresponding promoted objects of different promoted objects are the same; first object feature information of the promoted objects and second object feature information of the promoted objects are obtained; according to the first object feature information and the second object feature information, a first probability of each promoted object selecting each promoted object is predicted by a first probability prediction model; the number of the promoted objects is obtained, and according to the first probability of each promoted object selecting each promoted object and the number of the promoted objects, an expected number of objects is predicted; the cost of obtaining a target promoted object in the promoted objects is obtained, and according to the first probability of each promoted object selecting the target promoted object and the cost, a total cost of the target promoted object corresponding to the interest performance information is predicted; according to the expected number of objects and the total cost, the quality of the interest performance information is evaluated to obtain a quality evaluation result, and the quality evaluation result is used to indicate the promotion degree of promoting the target promoted object to the promoted objects, so that the quality of the interest performance information can be evaluated to obtain the quality evaluation result indicating the promotion degree of promoting the target promoted object to the promoted objects, and then the target promoted object can be promoted to the promoted objects according to the promotion degree, which can improve the conversion rate of the interest performance information compared with the scheme of fixedly promoting the target promoted object to the promoted objects according to a lower promotion degree. BRIEF DESCRIPTION OF DRAWINGS

[0041] The technical scheme and advantages of the present application will be apparent from the following detailed description of the embodiments of the present application, combined with the accompanying drawings.

[0042] Figure 1 FIG. 1 is a schematic diagram of a data processing system provided by an embodiment of the present application.

[0043] Figure 2 FIG. 2 is a first flowchart of a data processing method provided by an embodiment of the present application.

[0044] Figure 3 FIG. 3 is a first scenario diagram of a data processing method provided by an embodiment of the present application.

[0045] Figure 4 FIG. 4 is a second flowchart of a data processing method provided by an embodiment of the present application.

[0046] Figure 5 FIG. 5 is a second scenario diagram of a data processing method provided by an embodiment of the present application.

[0047] Figure 6is a structural schematic diagram of a data processing apparatus provided by an embodiment of the present application.

[0048] Figure 7 is a structural schematic diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0050] The embodiments of the present application provide a data processing method and related devices, which can include a data processing apparatus, a computer device, a computer readable storage medium and computer program content. The data processing apparatus can be integrated in the computer device, which can be a terminal or a server.

[0051] It can be understood that the data processing method of the present embodiment can be executed on a terminal, on a server, or by a terminal and a server together. The above examples should not be understood as limiting the present application.

[0052] Please refer to Figure 1 , Figure 1 is a data processing system schematic diagram provided by an embodiment of the present application. Taking the terminal and the server jointly executing the data processing method provided by the present embodiment as an example, the data processing system provided by the present embodiment includes a plurality of terminals 10 and a server 11, etc. The terminal 10 and the server 11 are connected through a network, such as a wired or wireless network connection, etc. The data processing apparatus can be integrated in the server.

[0053] The terminal 10 can be configured to acquire interest performance information of a promoted object corresponding to the terminal 10 for its corresponding promotion object, and send the interest performance information to the server 11. The interest performance information is used to represent the acquisition interest of the promoted object in acquiring its corresponding promotion object, and the types of the corresponding promotion objects of different promoted objects are the same.

[0054] The server 11 can be configured to receive the interest expression information of each promoted object corresponding to each terminal 10, so as to obtain the interest expression information of different promoted objects corresponding to the corresponding promoted objects; obtain the first object characteristic information of the promoted object and the second object characteristic information of the promoted object; according to the first object characteristic information and the second object characteristic information, the first probability of each promoted object selecting each promoted object is predicted by the first probability prediction model; the number of promoted objects is obtained, and the expected number of objects is predicted according to the first probability of each promoted object selecting each promoted object and the number of promoted objects; the acquisition cost of the target promoted object in the promoted object is obtained, and the total acquisition cost of the target promoted object corresponding to the interest expression information is predicted according to the first probability of each promoted object selecting the target promoted object and the acquisition cost; the quality of the interest expression information is evaluated according to the expected number of objects and the total acquisition cost, and the quality evaluation result is obtained, which is used to indicate the promotion degree of the target promoted object to the promoted object.

[0055] In an optional embodiment, the data processing steps in the server 11 described above can also be performed by any terminal 10, that is, the interest expression information corresponding to each terminal 10 is sent to a terminal 10 performing the data processing method provided in the embodiment of the present application, and the terminal 10 performs the data processing method provided in the embodiment of the present application according to the interest expression information corresponding to the terminal 10 and the received interest expression information.

[0056] In a possible implementation, the data processing steps in the server 11 described above can also be performed by a terminal for performing the data processing method provided in the embodiment of the present application, that is, the interest expression information corresponding to each terminal 10 is sent to the terminal for performing the data processing method provided in the embodiment of the present application, and the terminal performs the data processing method provided in the embodiment of the present application according to the received interest expression information.

[0057] The following will be described in detail. It should be noted that the order of the following embodiments is not limited as the preferred order of the embodiments.

[0058] The embodiment will be described from the perspective of a data processing device, which can be integrated in a computer device, which can be a server or a terminal, and the following will be described in detail with the computer device as a server.

[0059] It can be understood that in the specific embodiments of the present application, data related to the promoted object, such as interest performance information and first object feature information, needs to obtain the user's permission or consent when the above embodiments of the present application are applied to specific content or technology, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0060] For example, when the embodiments of the present application need to obtain interest performance information, a separate permission or separate consent for the interest performance information can be obtained through a pop-up window or by jumping to a confirmation page, and after determining the separate permission or separate consent, the interest performance information is obtained.

[0061] Please refer to Figure 2 , Figure 2 is the first flowchart of the data processing method provided by the embodiments of the present application, and the flowchart can include:

[0062] In 101, the interest performance information of different promoted objects for their corresponding promotion objects is obtained, and the interest performance information represents the interest of different promoted objects in obtaining their corresponding promotion objects. The types of the promotion objects corresponding to different promoted objects are the same.

[0063] Among them, the promoted object can be an object with biological feature information that can show interest in obtaining the promotion object, such as a person. When the promoted object has strong interest in obtaining the promotion object, it usually obtains, such as pays a certain acquisition cost to obtain the promotion object with strong interest. The acquisition cost is the cost required by the promoted object to obtain the promotion object.

[0064] The promotion object can be an object that the promoted object can obtain, such as an object that can be obtained by paying a certain acquisition cost, such as a product or a service, etc. Among them, the product usually refers to the entity product produced or provided by the enterprise. These products can be physical, such as cars, electronic products, etc., or virtual, such as software or digital content. The purpose of the product is to meet the needs of the promoted object or solve specific problems. Services refer to a series of non-physical help or benefits provided by enterprises. This usually involves some form of professional skills or labor, such as hotel services, medical services, etc. Services are usually provided to provide convenience, solve problems or meet other needs of the promoted object.

[0065] The interest expression information represents the acquisition interest of the promoted object that the promoted object shows in acquiring the corresponding promoted object, that is, the promoted object is interested in acquiring the corresponding promoted object. This usually occurs before the promoted object starts to consider the acquisition idea of the promoted object, which means that the promoted object has already generated an intention to possibly acquire the corresponding promoted object. If the corresponding promoted object or the promoted object of the same type as the corresponding promoted object is promoted to the promoted object, it is possible to prompt the promoted object to acquire the corresponding promoted object or the promoted object of the same type as the corresponding promoted object. The interest expression information can be information from various channels, such as network search, network browsing, social media interaction, industry exhibition, etc. The types of corresponding promoted objects of different promoted objects are the same, for example, all are security management software such as computer manager, mobile phone manager, etc., or all are cloud storage service software such as network disk, etc.

[0066] For example, if the promoted object A1 searches the promoted object C1 through the search software on the terminal B1, the terminal B1 obtains the interest expression information D1 representing the acquisition interest of the promoted object A1 in acquiring the corresponding promoted object C1.

[0067] If the promoted object A2 browses the promoted object C2 through the software on the terminal B2 for displaying the promoted object C2 of the same type as the promoted object C1, the terminal B2 obtains the interest expression information D2 representing the acquisition interest of the promoted object A2 in acquiring the corresponding promoted object C2.

[0068] If the promoted object A3 browses the promoted object C1 through the software on the terminal B3 for displaying the promoted object C1, the terminal B3 obtains the interest expression information D3 representing the acquisition interest of the promoted object A3 in acquiring the corresponding promoted object C1.

[0069] When different terminals such as the terminal B1, the terminal B2, and the terminal B3 obtain the interest expression information, the different terminals can respectively send the obtained interest expression information to the computer device integrated with the data processing apparatus, so that the computer device obtains the interest expression information from different terminals, such as the interest expression information D1, the interest expression information D2, and the interest expression information D3.

[0070] In 102, the first object feature information of the promoted object and the second object feature information of the promoted object are obtained.

[0071] It can be understood that the computer device integrated with the data processing apparatus can generally obtain a large amount of interest performance information. If the quality evaluation is directly performed on the large amount of interest performance information, the quality evaluation result may be relatively inaccurate. Therefore, in the embodiment, the large amount of interest performance information can be divided into multiple batches of interest performance information according to corresponding division rules. Each batch of interest performance information can include multiple interest performance information.

[0072] For example, the obtained interest performance information can be divided into multiple batches of interest performance information according to the use scene of the promoted object by the promoted object using the obtained promoted object. The use scene can include a work scene, a learning scene, and the like.

[0073] For example, the interest performance information corresponding to the use scene of the promoted object by the promoted object using the obtained promoted object can be divided into a batch of interest performance information, and the interest performance information corresponding to the use scene of the promoted object by the promoted object using the obtained promoted object can be divided into another batch of interest performance information.

[0074] For another example, the obtained interest performance information can be divided into multiple batches of interest performance information according to the acquisition cost of the obtained promoted object paid by the promoted object.

[0075] For example, multiple acquisition cost intervals can be set, and the obtained interest performance information can be divided into multiple batches of interest performance information according to the acquisition cost interval in which the acquisition cost of the obtained promoted object paid by the promoted object is located. The interest performance information corresponding to the same acquisition cost interval in which the acquisition cost is located can be divided into the same batch of interest performance information.

[0076] It should be noted that in the embodiment, if the interest performance information represents the acquisition interest of a certain promoted object in acquiring a corresponding promoted object, the promoted object and the promoted object are the promoted object and the promoted object corresponding to the interest performance information, respectively.

[0077] For example, when multiple batches of interest performance information are obtained, the first object feature information of the promoted object corresponding to each interest performance information in each batch of interest performance information and the second object feature information of the promoted object corresponding to each interest performance information in each batch of interest performance information can be obtained.

[0078] Optionally, since the promoted objects corresponding to different promoted objects can be the same, after the second object feature information of a certain promoted object is obtained, the second object feature information of the promoted object can no longer be repeatedly obtained.

[0079] In a possible implementation, after obtaining a plurality of batches of interest expression information, for each batch of interest expression information, each interest expression information corresponding to a promoted object in each batch of interest expression information can be integrated first, the same promoted objects are integrated into one promoted object, and then the second object feature information of the integrated promoted object is obtained.

[0080] It should be noted that, since the processing procedure of each batch of interest expression information is the same, the processing procedure of one batch of interest expression information will be taken as an example to describe the data processing method provided in this embodiment.

[0081] For example, it is assumed that the obtained batch of interest expression information includes interest expression information D1, D2, D3, and D4, the interest expression information D1 represents that the promoted object A1 expresses the acquisition interest in the corresponding promoted object C1, the interest expression information D2 represents that the promoted object A2 expresses the acquisition interest in the corresponding promoted object C2, the interest expression information D3 represents that the promoted object A3 expresses the acquisition interest in the corresponding promoted object C3, and the interest expression information D4 represents that the promoted object A4 expresses the acquisition interest in the corresponding promoted object C3. Then, the first object feature information of the promoted object A1, the first object feature information of the promoted object A2, the first object feature information of the promoted object A3, the first object feature information of the promoted object A4, the second object feature information of the promoted object C1, the second object feature information of the promoted object C2, and the second object feature information of the promoted object C3 can be obtained.

[0082] The first object feature information of the promoted object refers to a series of features and data related to the promoted object. The first object feature information of the promoted object can include at least one of the following: basic information such as gender, a use scenario of the promoted object, an estimated acquisition cost (estimated acquisition cost) that the promoted object can pay for acquiring the promoted object, an acquisition cost (historical acquisition cost) that the promoted object has paid, a promoted object that the promoted object expresses the acquisition interest in (corresponding promoted object of the promoted object), a service provided by a target promoted object that the promoted object expresses the acquisition interest in (corresponding service of the promoted object), a promoted object selected and acquired by the promoted object, a use degree of the promoted object (related to the use frequency, use function, use duration, etc.), and a service provided by a target promoted object selected and acquired by the promoted object.

[0083] For example, taking two promoted objects with object id 1001 and 1002 as an example, the promoted object with object id 1001 corresponds to the promotion object of product "computer manager", and the promoted object with object id 1002 corresponds to the promotion object of product "mobile phone manager". The first object feature information of the promoted object can be:

[0084]

[0085]

[0086] The second object feature information of the promotion object refers to a series of characteristics and performance indicators that the promotion object itself has. These information are key factors considered by the promoted object when selecting the promotion object. The second object feature information of the promotion object can include at least one of the following: the name of the promotion object, the memory occupation of the promotion object, the acquisition cost of the promotion object, the scale of the promoted object using the promotion object, the installation mode of the promotion object, etc.

[0087] For example, taking two products of computer manager and mobile phone manager as an example, the second object feature information of the promotion object can be:

[0088] Product name Product memory usage Promoted object size Product installation method Product acquisition cost Computer housekeeper 398 6 Self-installation 100 Mobile phone housekeeper 250 8 Mobile phone built-in 0

[0089] In 103, according to the first object feature information and the second object feature information, the first probability of each promoted object selecting each promotion object is predicted by a first probability prediction model.

[0090] The first probability prediction model can be a discrete choice model (DCM). The DCM model is a econometric model used to analyze single selection from a set of mutually exclusive and preferred options, which is used to analyze and predict the selection behavior of individuals when facing multiple discrete options. The DCM model is widely used in marketing, traffic engineering, economics, sociology and other fields, helping researchers and decision makers to understand how people make choices between different options, and to predict the influence of different factors on the selection results. Its main tasks include:

[0091] Predicting the decision-making behavior of a group of decision makers.

[0092] Determining the influence of different option feature information on the selection decision of the decision maker.

[0093] Understanding how different groups evaluate different characteristics of an alternative, so as to modify the characteristics of the option that has a significant impact on individual decision makers through carefully designed strategies, and to change behavior in an active way.

[0094] The DCM model is designed based on probabilistic choice theory, and its principle is random utility theory. The model describes the preference of decision-maker t for option i, which can be expressed using the utility value U. it To represent. U it It is an unknown function, and its expression is:

[0095] U it =V it +ε it

[0096] Among them, V it For observable utility, ε it This is the difference between the total real utility and the certainty of utility, i.e., the residual. Therefore, when a decision-maker pursues the utility-maximizing option, i.e., when decision-maker t chooses option i, we have:

[0097] U it >U jt (j≠i)

[0098] The DCM model calculates the probability of choice, not the prediction of which choice a person will make. These probabilities reflect the probability P(i) of a decision-maker with given feature information and facing the same set of options choosing each option. t :

[0099] P(i) t =Prob(V it +ε it >V jt +ε jt ,j≠i)

[0100] =Prob(V it -V jt >ε jt -ε it ,j≠i)

[0101] =∫I(V it -V jt >ε jt -ε it ,j≠i)f(ε t )dε t

[0102] Where, f(ε) t f(ε) is the joint density function of the residuals, and I is a judgment function; if the statement within the parentheses is true, the function returns 1, otherwise 0. Different DCMs have different forms of f(ε). t), commonly used are the logit model and the probit model. Since the logit model has more computational advantages, the multinomial logit model (MNL) of the DCM model is used in this embodiment to predict the first probability of each promoted object selecting each promotion object.

[0103] The MNL model is a statistical model used to handle discrete choice data, especially when there are more than two choices. The MNL model is an extension of the logit model to predict the probability of selecting a particular option among multiple options.

[0104] The utility function of the MNL model is as follows:

[0105] V it =X t β i +ε it (i=1,2,……,n)

[0106] where V it is the utility of the promoted object t selecting the promotion object i, X t is the set of first object feature information, β i is the estimated parameter of the first object feature information in the selection of the i-th promotion object, and ε it is the error term. When V it >V jt , the promoted object selects the promotion object i, and the selection probability is:

[0107]

[0108] where J is the set of promotion objects, including multiple promotion objects of the same type. In the above formula, β i is calculated by maximum likelihood estimation, and the likelihood function is defined as:

[0109]

[0110] where δ jt is the selection result, and if the promoted object t selects the promotion object j, δ jt =1, otherwise δ jt =0. P it represents the probability of the promoted object t selecting the promotion object i. The first derivative of the likelihood function is obtained by taking the first derivative of the likelihood function and equating it to 0, to obtain the maximum value of the likelihood function.

[0111] For example, assuming that a first probability prediction model for predicting a first probability of each promoted object selecting the same type of promoted object C1, C2 and C3 is needed to be trained, an MNL model can be created through a relevant model package of python, and the dataset data, different selection items alt_id_col (in this embodiment, referring to promoted objects C1, C2 and C3), different observation values or individuals obs_id_col (in this embodiment, referring to different promoted objects), selection result column choice_col (in this embodiment, referring to which promoted object selects which promoted object), result storage form specification (in this embodiment, specified as a list), model type model_type, and model name names need to be specified.

[0112]

[0113] Then, the object feature information of each historical promoted object in the plurality of historical promoted objects and the second object feature information of the promoted objects C1, C2 and C3 are obtained, and the object feature information of each historical promoted object and the second object feature information of the promoted objects C1, C2 and C3 are integrated as first sample data, so as to obtain a plurality of first sample data. Wherein, if the historical interest performance information represents that a promoted object shows an interest in obtaining its corresponding promoted object, such as promoted object C1, C2 or C3, and then the promoted object selects and obtains the promoted object C1, C2 or C3, the promoted object is a historical promoted object.

[0114] For example, taking the object feature information of the historical promoted object A11 and A12 and the second object feature information of the promoted objects C1, C2 and C3 as the first sample data, assuming that the second object feature information of the promoted object C1 includes that the memory occupation of the promoted object C1 is 398, the installation mode of the promoted object C1 is self-installation, the acquisition cost of the promoted object C1 is 100, the second object feature information of the promoted object C2 includes that the memory occupation of the promoted object C2 is 250, the installation mode of the promoted object C2 is device built-in, the acquisition cost of the promoted object C2 is 0, the second object feature information of the promoted object C3 includes that the memory occupation of the promoted object C3 is 200, the installation mode of the promoted object C3 is self-installation, the acquisition cost of the promoted object C3 is 50, the object feature information of the historical promoted object A11 includes that the id of the historical promoted object A11 is 2001, the gender of the historical promoted object A11 is female, the promoted object that the historical promoted object A11 shows the acquisition interest of acquiring is the promoted object C1, the promoted object selected by the historical promoted object A11 is the promoted object C2, and the object feature information of the historical promoted object A12 includes that the id of the historical promoted object A12 is 2002, the gender of the historical promoted object A12 is male, the promoted object that the historical promoted object A12 shows the acquisition interest of acquiring is the promoted object C1, and the promoted object selected by the historical promoted object A12 is the promoted object C1, the first sample data obtained by integrating the object feature information of the historical promoted object A11 and the second object feature information of the promoted objects C1, C2 and C3 is [id of the promoted object: 2001, gender of the promoted object: female, promoted object that the promoted object shows the acquisition interest of acquiring: promoted object C1, promoted object selected by the promoted object: promoted object C2, memory occupation of the promoted object C1: 398, installation mode of the promoted object C1: self-installation, acquisition cost of the promoted object C1: 100, memory occupation of the promoted object C2: 250, installation mode of the promoted object C2: device built-in, acquisition cost of the promoted object C2: 0, memory occupation of the promoted object C3: 200, installation mode of the promoted object C3: self-installation, acquisition cost of the promoted object C3: 50], and the first sample data obtained by integrating the object feature information of the historical promoted object A12 and the second object feature information of the promoted objects C1, C2 and C3 is [id of the promoted object: 2002, gender of the promoted object: male, promoted object that the promoted object shows the acquisition interest of acquiring: promoted object C1, promoted object selected by the promoted object: promoted object C1, memory occupation of the promoted object C1: 398, installation mode of the promoted object C1: self-installation, acquisition cost of the promoted object C1: 100, memory occupation of the promoted object C2: 250, installation mode of the promoted object C2: device built-in, acquisition cost of the promoted object C2: 0, memory occupation of the promoted object C3: 200, installation mode of the promoted object C3: self-installation, acquisition cost of the promoted object C3: 50].The installation mode of the promotion object C3 is self-installation, and the acquisition cost of the promotion object C3 is 50.

[0115] It should be noted that the above is only two examples of obtaining the first sample data provided by the embodiments of the present application, and is not used to limit the present application. In actual application, the content included by the object feature information of the historical promoted object and the second object feature information of the promotion object can be determined according to actual needs, which is not specifically limited here.

[0116] Then, the created MNL model is iteratively trained through the obtained first sample data, and whether to stop the model training is determined through the model training result obtained each time. When it is determined to stop the model training, the MNL model obtained by the last training is taken as the first probability prediction model, and the first probability prediction model is used to predict the first probability of each promoted object selecting the promotion object C1, the first probability of each promoted object selecting the promotion object C2 and the first probability of each promoted object selecting the promotion object C3 according to the first object feature information of each promoted object obtained afterwards and the second object feature information of the promotion objects C1, C2 and C3.

[0117] In an optional embodiment, the model training result obtained each time can be output, and whether to stop the model training is determined by the technical personnel through the output model training result.

[0118] For example, part of the output model training result is as follows:

[0119]

[0120] The above model training result is introduced as follows:

[0121] Pseudo R-squ.: 0.965- pseudo R-square, an index to measure the goodness of fit of the model, close to 1 indicates that the model fits well.

[0122] Pseudo R-bar-squ.: 0.857- adjusted pseudo R-square, also measures the goodness of fit of the model, but considers the degree of freedom of the model.

[0123] AIC: 31.813- Akaike Information Criterion, used for model selection, and a smaller AIC value usually indicates a better model fit.

[0124] BIC: 60.397- Bayesian Information Criterion, also used for model selection, similar to AIC, but considering the sample size.

[0125] The output model training results can also include the coefficients of each independent variable, namely the coefficients corresponding to the first and second object features of the promoted object, such as the coefficient corresponding to the gender of the promoted object, the coefficient corresponding to the memory usage of promoted object C1, the coefficient corresponding to the installation method of promoted object C2, the coefficient corresponding to the acquisition cost of promoted object C3, the standard error of each coefficient estimate, the p-value of each coefficient estimate, and the 95% confidence interval of each coefficient. The coefficients of the independent variables are the β values ​​in the utility function. i It represents the expected change in the dependent variable, i.e., the selection outcome of the promoted object, when the independent variable changes by one unit. The standard error of the coefficient estimate is used to measure the precision of the coefficient, and the p-value of the coefficient estimate is used to test whether the coefficient is significantly non-zero. The 95% confidence interval of the coefficient refers to the interval in which the estimated value of the coefficient falls around its true value at a certain confidence level (here, 95%).

[0126] Technicians can determine whether to stop model training based on the model training results shown above. If technicians determine to stop model training based on the model training results shown above, then the MNL model obtained from the last training can be used as the first probability prediction model.

[0127] Taking the above-mentioned indicators for measuring model fit as an example to determine whether to stop model training, a fit threshold can be set, and model training can be stopped when the above-mentioned indicators for measuring model fit are greater than or equal to the fit threshold.

[0128] In some embodiments, the obtained first sample data can be divided into a first training dataset and a first validation dataset. The MNL model is trained using the first training dataset and the trained MNL is validated using the first validation dataset to obtain the prediction accuracy of the trained MNL. When the prediction accuracy is greater than or equal to the accuracy threshold, the training of the MNL model can be stopped, and the MNL model obtained from the last training can be used as the first probabilistic prediction model.

[0129] Optionally, after obtaining the first probability prediction model through the above process, the first probability prediction model can be incrementally trained based on new data. There are no specific restrictions here, and the actual needs shall prevail.

[0130] After step 102 is completed, the first object feature information of each promoted object and the second object feature information of promoted objects C1, C2 and C3 can be integrated as the first data for prediction, resulting in multiple first data for prediction.

[0131] It should be noted that the second object feature information of the promotion objects C1, C2 and C3 obtained in step 102 must be the same as the second object feature information of the promotion objects C1, C2 and C3 obtained in training the first probability prediction model, and the first object feature information of the promoted objects obtained in step 102 must correspond to the object feature information of the historical promoted objects obtained in training the first probability prediction model.

[0132] For example, taking the integration of the first object feature information of the promoted object A1 and the second object feature information of promoted objects C1, C2 and C3 as the first data for prediction, assuming that the first object feature information of the promoted object A1 includes: the id of the promoted object A1 is 1001, the gender of the promoted object A1 is female, and the promoted object that the promoted object A1 shows interest in acquiring is promoted object C1, then the first data for prediction corresponding to the promoted object A1 is [id of the promoted object: 1001, gender of the promoted object: female, promoted object that the promoted object shows interest in acquiring: promoted object C1, memory usage of promoted object C1: 398, installation method of promoted object C1: self-installed, acquisition cost of promoted object C1: 100, memory usage of promoted object C2: 250, installation method of promoted object C2: device-provided, acquisition cost of promoted object C2: 0, memory usage of promoted object C3: 200, installation method of promoted object C3: self-installed, acquisition cost of promoted object C3: 50].

[0133] After obtaining multiple first data points for prediction in the above manner, each first data point can be input into the obtained first probability prediction model to obtain the first probability of each promoted object selecting promoted object C1, the first probability of selecting promoted object C2, and the first probability of selecting promoted object C3.

[0134] For example, using the first probability prediction model, the following is an example of the predicted first probabilities for promoted objects with IDs 1001 and 1002 to choose promoted object C1, C2, and C3, respectively. Here, OBS_ID represents the ID of the promoted object, PRODUCT_ALT_ID represents the ID of the promoted object, where "0" represents the ID of promoted object C1, "1" represents the ID of promoted object C2, "2" represents the ID of promoted object C3, and PRODUCT_PROB represents the first probability of each promoted object being chosen.

[0135]

[0136]

[0137] Conventional practices in this field typically assess the quality of interest performance information solely based on the second object characteristic information of the promoted object. This approach only quantifies the second object characteristic information of the promoted object, without quantifying the difference in utility contributed by the second object characteristic information of the promoted object to different promoted objects. However, the utility of different promoted objects to different promoted objects is not the same. Therefore, in this embodiment, based on the first object characteristic information and the second object characteristic information of each promoted object, a first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object. This quantifies the interaction between the second object characteristic information of the promoted object and the first object characteristic information of the promoted object, and quantifies the difference in utility contributed by the second object characteristic information of the promoted object to different promoted objects. Then, based on the first probability of each promoted object selecting each promoted object, the expected number of objects to be acquired is predicted, and based on the first probability of each promoted object selecting each promoted object, the total acquisition cost of the interest performance information corresponding to the target promoted object is predicted. The quality of the interest performance information is then assessed based on the expected number of objects to be acquired and the total acquisition cost. Compared with the scheme that only assesses the quality of interest performance information based on the second object characteristic information of the promoted object, the accuracy of the quality assessment of interest performance information is higher.

[0138] In step 104, the number of promoted objects is obtained, and the expected number of objects to be obtained is predicted based on the first probability of each promoted object selecting each promoted object and the number of promoted objects.

[0139] In this embodiment, the number of promoted objects can be obtained, and after obtaining the first probability of each promoted object selecting each promoted object, the expected number of objects to be obtained can be predicted based on the first probability of each promoted object selecting each promoted object and the number of promoted objects.

[0140] Each piece of interest-based information corresponds to a target audience for promotion. Therefore, the number of interest-based information entries can be used as the number of targets to be promoted. For example, if there are 100 pieces of interest-based information, then the number of targets to be promoted is 100.

[0141] The expected number of targets to be acquired is the number of promoted targets that are projected to be converted into actual targets. For example, if the number of promoted targets is 100 and the projected number of promoted targets that are projected to be converted into actual targets is 70, then the expected number of targets to be acquired is 70.

[0142] In one possible implementation, predicting the expected number of targets to be acquired based on the first probability of each target selecting each target and the number of targets includes: performing an addition operation on the first probability of each target selecting each target to obtain a probability sum; and performing a division operation on the probability sum and the number of targets to obtain the expected number of targets to be acquired.

[0143] For example, assuming you have 500 first probabilities, you can add these 500 first probabilities together to get the sum of probabilities; then divide the sum of probabilities by the number of objects to be promoted, and round up or down to get the expected number of objects to be obtained.

[0144] For example, the desired number of targets can be calculated using a first formula based on the first probability of each target selecting each target and the number of targets.

[0145] The first formula is:

[0146] Where E(j) is the expected number of objects to be obtained, and P it Let c be the first probability of selecting the i-th promotion target for the t-th promotion target. j The number of targets being promoted.

[0147] It is understandable that calculating the expected number of targets in the above way can reflect the true intention of each target to acquire each target, taking into account the acquisition interest of each target. This can ensure the accuracy of predicting the expected number of targets to a certain extent. Furthermore, the calculation process of the expected number of targets is relatively simple, which can reduce the amount of computation required to calculate the expected number of targets and save processing resources.

[0148] In 105, the acquisition cost of the target promotion object among the promotion objects is obtained, and the total acquisition cost of the interest performance information corresponding to the target promotion object is predicted based on the first probability and acquisition cost of each promoted object selecting the target promotion object.

[0149] In this embodiment, the acquisition cost of the target promotion object among the promotion objects can be obtained, and after obtaining the first probability of each promoted object acquiring the target promotion object, the total acquisition cost of the interest performance information corresponding to the target promotion object can be predicted based on the first probability of each promoted object acquiring the target promotion object and the acquisition cost of the target promotion object.

[0150] The target promotion object can be at least one of the promotion objects. For example, the target promotion object can be a product produced by a certain company or some products of the same type.

[0151] Total acquisition cost is the sum of the acquisition costs paid by the target audience for a batch of interest performance information to acquire the target audience.

[0152] In some embodiments, the target promotion object provides services at multiple tiers, and the acquisition cost of the target promotion object for each tier of service is different. To more accurately predict the total acquisition cost of a batch of interest performance information corresponding to the target promotion object, and thus more accurately assess the quality of that batch of interest performance information, when predicting the total acquisition cost of a batch of interest performance information corresponding to the target promotion object, the second probability of each promotion object selecting each tier of the target promotion object's services, and the acquisition cost of each tier of the target promotion object's services, can be considered as factors in the prediction. That is, before predicting the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability of each promoted object selecting the target promotion object and the acquisition cost, the following factors are also considered: The method includes: based on the first object feature information and the object feature information of the services corresponding to each tier of the target promotion object, using a second probability prediction model, predicting the second probability of each promoted object selecting each tier of services; obtaining the acquisition cost of the target promotion object among the promotion objects, including: obtaining the acquisition cost of the services corresponding to each tier of the target promotion object; and based on the first probability and acquisition cost of each promoted object selecting the target promotion object, predicting the total acquisition cost of the interest performance information corresponding to the target promotion object, including: based on the first probability of each promoted object selecting the target promotion object, the second probability of each promoted object selecting each tier of services, and the acquisition cost of the services corresponding to each tier of the target promotion object, predicting the total acquisition cost of the interest performance information corresponding to the target promotion object.

[0153] Understandably, for certain promotional targets, such as certain products, multiple service tiers are often offered. For example, cloud storage service software may offer multiple tiers of speed-up services and multiple tiers of storage space services. The speed-up service primarily aims to improve the download speed of resources such as video and audio files downloaded through cloud storage service software. The storage space service primarily provides a corresponding amount of storage space for the promoted target to store the corresponding resources, such as video and audio files. Since different service tiers offer different experiences to the target audience, the acquisition costs for each tier also vary. To more accurately estimate the total acquisition cost of interest performance information for these target audiences, in addition to predicting the first probability of each target audience choosing the target audience, we can also obtain the object feature information of each tier of service for the target audience. Based on the first object feature information and the object feature information of each tier of service for the target audience, we use a second probability prediction model to predict the second probability of each target audience choosing each tier of service. Finally, based on the first probability of each target audience choosing the target audience, the second probability of each target audience choosing each tier of service, and the acquisition cost of each tier of service for the target audience, we can predict the total acquisition cost of interest performance information for the target audience.

[0154] The object characteristic information for each tier of services of the target promotion object refers to a series of characteristics and performance indicators of each tier of services of the target promotion object. This information is a key factor that the promoted object considers when selecting each tier of services of the target promotion object. The second object characteristic information of the promotion object may include at least one of the following: the name of each tier of services of the target promotion object, the memory usage of each tier of services of the target promotion object, the acquisition cost of each tier of services of the target promotion object, and the scale of the promoted object using each tier of services of the target promotion object, etc.

[0155] For example, the second probability prediction model can also be a DCM model. Since the logarithmic model has more computational advantages, this embodiment will use the MNL model of the DCM model to predict the second probability of each promoted object selecting each tier of the target promoted object's service.

[0156] For example, suppose we need to train a second probability prediction model to predict whether each promoted object will choose the first, second, or third tier of the target promoted object's services. We can first create an MNL model using the relevant Python model package. We need to specify the dataset (data), different choices (in this example, the first, second, and third tiers of the target promoted object's services), different observations or individuals (in this example, different promoted objects), selection result column (in this example, which promoted object chose which tier of the target promoted object's services), result storage format (in this example, a list), model type (model_type), and model name (names).

[0157] Then, the object feature information of each historical promoted object, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object are obtained from multiple historical promoted objects. These object feature information, along with the object feature information of the first-tier service corresponding to the target promoted object, the second-tier service corresponding to the target promoted object, and the third-tier service corresponding to the target promoted object, are integrated as second sample data to obtain multiple second sample data sets. Specifically, if a certain historical interest expression information represents a promoted object's interest in acquiring the first-tier, second-tier, or third-tier service of the target promoted object, and the promoted object subsequently selects and acquires the first-tier, second-tier, or third-tier service of the target promoted object, then that promoted object is a historical promoted object.

[0158] For example, taking the integrated object feature information of historical promoted objects A13 and A14, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object as the second sample data, assuming that the object feature information of the first-tier service corresponding to the target promoted object includes: the memory usage of the first-tier service corresponding to the target promoted object is 400, and the acquisition cost of the first-tier service corresponding to the target promoted object is 120, and the object feature information of the second-tier service corresponding to the target promoted object includes: the memory usage of the second-tier service corresponding to the target promoted object is... 398. The acquisition cost of the second-tier service corresponding to the target promotion object is 100. The object characteristic information of the third-tier service corresponding to the target promotion object includes: the memory usage of the third-tier service corresponding to the target promotion object is 396, the acquisition cost of the third-tier service corresponding to the target promotion object is 80, and the object characteristic information of the historical promoted object A13 includes: the ID of the historical promoted object A13 is 2003, the gender of the historical promoted object A13 is female, the service that the historical promoted object A13 showed interest in acquiring is the first-tier service of the target promotion object, and the service selected by the historical promoted object A13 is the second-tier service of the target promotion object. The target characteristics of the promoted object A14 include: the historical promoted object A14's ID is 2004, the historical promoted object A14's gender is male, the services that the historical promoted object A14 showed interest in acquiring are the first-tier services of the target promoted object, and the services that the historical promoted object A14 selected are the first-tier services of the target promoted object. The second sample data obtained by integrating the target characteristics of the historical promoted object A13, the target promoted object's first-tier service, the target promoted object's second-tier service, and the target promoted object's third-tier service is [promoted object]. ID: 2003, Gender of the target audience: Female, Services the target audience shows interest in: First-tier services of the target audience, Services selected by the target audience: Second-tier services of the target audience, Memory usage of the first-tier services for the target audience: 400, Acquisition cost of the first-tier services for the target audience: 120, Memory usage of the second-tier services for the target audience: 398, Acquisition cost of the second-tier services for the target audience: 100, Memory usage of the third-tier services for the target audience: 396, Acquisition cost of the third-tier services for the target audience: 80.The second sample data, obtained by integrating the object feature information of historical promoted object A14, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object, is as follows: [Promoted object id: 2004, promoted object gender: male, service the promoted object showed interest in: the first-tier service of the target promoted object, service selected by the promoted object: the first-tier service of the target promoted object, memory usage of the first-tier service corresponding to the target promoted object: 400, acquisition cost of the first-tier service corresponding to the target promoted object: 120, memory usage of the second-tier service corresponding to the target promoted object: 398, acquisition cost of the second-tier service corresponding to the target promoted object: 100, memory usage of the third-tier service corresponding to the target promoted object: 396, acquisition cost of the third-tier service corresponding to the target promoted object: 80].

[0159] It should be noted that the above are merely two examples of obtaining the second sample data provided in the embodiments of this application, and are not intended to limit this application. In practical applications, the content included in the object feature information of the historical promoted objects and the object feature information of the target promoted objects corresponding to each level of service can be determined according to actual needs, and no specific limitations are made here.

[0160] Next, the created MNL model is trained iteratively multiple times using the obtained second sample data, and the model training results at each stage are used to determine whether to stop training. When it is determined to stop training, the MNL model obtained from the last training is used as the second probability prediction model. This second probability prediction model is then used to predict the second probability of each promoted object choosing the first tier service, the second tier service, and the third tier service of the target promoted object, based on the first object feature information, the object feature information of the first tier service, the second tier service, and the third tier service of the target promoted object obtained subsequently.

[0161] It should be noted that the specific method for determining whether to stop model training based on the model training results obtained each time can be found in the previous examples, and will not be repeated here.

[0162] In some embodiments, the obtained second sample data can be divided into a second training dataset and a second validation dataset. The MNL model is trained using the second training dataset and the trained MNL is validated using the second validation dataset to obtain the prediction accuracy of the trained MNL. When the prediction accuracy is greater than or equal to the accuracy threshold, the training of the MNL model can be stopped, and the MNL model obtained from the last training can be used as the second probabilistic prediction model.

[0163] Optionally, after obtaining the second probability prediction model through the above process, the second probability prediction model can be incrementally trained based on new data. There are no specific restrictions here, and the actual needs shall prevail.

[0164] After obtaining the first object feature information of each promoted object, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object, the obtained first object feature information of each promoted object, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object can be integrated as second data for prediction, thus obtaining multiple second data for prediction.

[0165] It should be noted that the object feature information of the first-tier service, the second-tier service, and the third-tier service corresponding to the target promotion object must be the same as the object feature information of the first-tier service, the second-tier service, and the third-tier service corresponding to the target promotion object obtained by training the second probability prediction model. The first object feature information of the promoted object must correspond to the object feature information of the historical promoted object obtained by training the second probability prediction model.

[0166] For example, taking the integration of the first object feature information of the promoted object A1, the object feature information of the first-tier service corresponding to the target promoted object, the object feature information of the second-tier service corresponding to the target promoted object, and the object feature information of the third-tier service corresponding to the target promoted object as the second data for prediction, assuming that the first object feature information of the promoted object A2 includes: the id of the promoted object A2 is 1002, the gender of the promoted object A2 is female, and the service that the promoted object A2 shows interest in acquiring is the first-tier service of the target promoted object, then the second data for prediction corresponding to the promoted object A2... [The target audience's ID is 1002, the target audience's gender is female, the service the target audience shows interest in is the first-tier service of the target audience, the memory usage of the first-tier service of the target audience is 400, the acquisition cost of the first-tier service of the target audience is 120, the memory usage of the second-tier service of the target audience is 398, the acquisition cost of the second-tier service of the target audience is 100, the memory usage of the third-tier service of the target audience is 396, and the acquisition cost of the third-tier service of the target audience is 80].

[0167] After obtaining multiple second data for prediction in the above manner, each second data for prediction can be input into the obtained second probability prediction model to obtain the second probability of each promoted object choosing the first tier of the target promoted object's service, the second probability of choosing the second tier of the target promoted object's service, and the second probability of choosing the third tier of the target promoted object's service.

[0168] For example, using the second probability prediction model, the following is an example of the predicted second probabilities for promoted objects with IDs 1001 and 1002 to choose the first tier service, the second tier service, and the third tier service of the target promoted object, respectively. Here, OBS_ID represents the ID of the promoted object, PRODUCT_ALT_ID = 0 represents the target promoted object, PRODUCT_PROB represents the first probability of the promoted object choosing the target promoted object (which can be predicted by the first probability prediction model), CONFIG_ALT_ID represents the ID of each tier service of the target promoted object, where "0" represents the ID of the first tier service, "1" represents the ID of the second tier service, "2" represents the ID of the third tier service, and CONFIG_PROB represents the second probability of the promoted object choosing each tier service of the target promoted object.

[0169] OBS ID PRODUCT ALT ID PRODUCT PROB CONFIG ALT ID CONFIG PROB 1001 0 0.25 0 0.4 1001 0 0.25 1 0.2 1001 0 0.25 2 0.4 1002 0 0.27 0 0.6 1002 0 0.27 1 0.1 1002 0 0.27 2 0.3

[0170] It is understandable that the utility of different service tiers of the target promotion object varies for different promoted objects. Therefore, in this embodiment, based on the first object feature information of each promoted object and the object feature information of each service tier corresponding to the target promotion object, a second probability prediction model is used to predict the second probability of each promoted object choosing each service tier of the target promotion object. This can quantify the interaction between the object feature information of each service tier corresponding to the target promotion object and the first object feature information of the promoted object, and quantify the difference in utility contributed by the object feature information of each service tier corresponding to the target promotion object to different promoted objects. Then, based on the second probability of each promoted object choosing each service tier of the target promotion object, the total acquisition cost of the interest performance information corresponding to the target promotion object is predicted, and the quality of the interest performance information is evaluated based on the total acquisition cost, which can make the quality evaluation of the interest performance information more accurate.

[0171] It is also understood that, since the objective of this application embodiment is to evaluate the quality of a batch of interest performance information corresponding to the target promotion object, in order to predict the conversion rate of a batch of interest performance information into promoted objects that will actually acquire the target promotion object, the higher the quality of a batch of interest performance information, the higher the conversion rate of a batch of interest performance information into promoted objects that will actually acquire the target promotion object. Specifically, the quality evaluation is carried out through two dimensions: the expected number of objects to be acquired and the total acquisition cost of the interest performance information corresponding to the target promotion object. Therefore, predicting the first probability of each promoted object choosing each promotion object through the first probability prediction model and the second probability of each promoted object choosing each tier of service of the target promotion object through the second probability prediction model can reduce the prediction of redundant information, that is, reduce the prediction of the probability of each promoted object choosing each tier of service of other promotion objects besides the target promotion object, thus reducing the computational load of the model. Furthermore, by using a first probability prediction model to predict the first probability of each promoted object choosing each promoted object, and by using a second probability prediction model to predict the second probability of each promoted object choosing each tier of service of the target promoted object, each probability prediction model focuses on the difference in information in only one dimension. That is, the first probability prediction model focuses on the difference in the second object feature information of each promoted object, while the second probability prediction model focuses on the difference in the object feature information of each tier of service corresponding to the target promoted object. Compared to the approach of directly predicting the probability of each promoted object choosing each tier of service of each promoted object through a single probability prediction model, this probability prediction model needs to focus on the difference in information in two dimensions, that is, it needs to focus on both the difference in the second object feature information of each promoted object and the difference in the object feature information of each tier of service corresponding to each promoted object. Therefore, the accuracy of the model prediction will be higher.

[0172] In an optional embodiment, since the first probability of each promoted object selecting the target promoted object reflects the probability that each promoted object will actually acquire the target promoted object, and the second probability of each promoted object selecting each tier of the target promoted object's services reflects the probability that each promoted object will actually acquire each tier of the target promoted object's services, then the acquisition cost of each tier of the target promoted object's services can be weighted and summed using the first and second probabilities as weights to obtain the total acquisition cost of the target promoted object's interest performance information. This ensures the accuracy of the prediction of the total acquisition cost of the target promoted object's interest performance information to a certain extent. That is, based on the first probability of each promoted object selecting the target promoted object, the second probability of each promoted object selecting each tier of the services, and the acquisition cost of each tier of the target promoted object's services, the total acquisition cost of the target promoted object's interest performance information is predicted, including: using the first probability of each promoted object selecting the target promoted object and the second probability of each promoted object selecting each tier of the services as weights, the acquisition cost of each tier of the services for the target promoted object is weighted and summed to obtain the total acquisition cost of the target promoted object's interest performance information.

[0173] For example, the first probability that each promoted object selects the target promoted object and the second probability that each promoted object selects each service tier can be used as weights. The acquisition cost of each service tier corresponding to the target promoted object can be weighted and summed using the second formula to obtain the total acquisition cost of the interest performance information corresponding to the target promoted object.

[0174] The second formula is: R(T)=∑P t P tk C k

[0175] in, R(T) represents the total cost of acquiring information about the interest expression of the target audience, and P... t P is the first probability of selecting a target audience for the target audience t. tk The second probability, C, of ​​selecting the k-th tier of the target promotion object's service for the promoted object t. k The cost of acquiring the service at the kth tier corresponding to the target audience.

[0176] In some embodiments, predicting the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability and acquisition cost of each promoted object acquiring the target promotion object includes: performing a weighted summation operation on the acquisition cost of the target promotion object using the first probability of each promoted object selecting the target promotion object as the weight, to obtain the total acquisition cost of the interest performance information corresponding to the target promotion object.

[0177] For certain targets of promotion, there is usually only one acquisition cost. Therefore, for these targets, the total acquisition cost of the interest performance information corresponding to the target can be predicted directly based on the first probability of each promoted target acquiring the target promoted target and the acquisition cost of the target promoted target.

[0178] For example, the first probability of each promoted object selecting the target promoted object can be used as a weight, and the acquisition cost of the target promoted object can be weighted and summed through the third formula to obtain the total acquisition cost of the interest performance information corresponding to the target promoted object.

[0179] The third formula is: R(T)=∑P t C

[0180] in, R(T) represents the total cost of acquiring information about the interest expression of the target audience, and P... t Let C be the first probability of selecting the target promotion object for the promoted object t, and let C be the acquisition cost of the target promotion object.

[0181] In 106, the interest performance information is evaluated for quality based on the expected number of targets and the total acquisition cost. The quality evaluation results are used to indicate the degree of promotion of the target target to the promoted targets.

[0182] Once the expected number of targets and the total acquisition cost are obtained, the interest performance information can be evaluated for quality based on these figures. The quality evaluation results are used to indicate the extent to which the target audience is promoted to the promoted audience.

[0183] The degree of promotion is related to the quality of interest expression information. The higher the quality of interest expression information, the higher the degree of promotion, and vice versa.

[0184] For example, the quality assessment result can be a quality assessment level, and the degree of promotion can be a degree of promotion level. The higher the quality assessment level, the higher the quality of the interest expression information; the higher the quality of the interest expression information, the higher the degree of promotion, and the higher the corresponding degree of promotion level. The lower the quality assessment level, the lower the quality of the interest expression information; the lower the quality of the interest expression information, the lower the degree of promotion, and the lower the corresponding degree of promotion level.

[0185] In one possible implementation, the interest performance information is evaluated for quality based on the expected number of objects to be acquired and the total acquisition cost, and the evaluation result is obtained. This includes: constructing a Cartesian coordinate system, wherein the first direction axis of the Cartesian coordinate system represents the quantity and the second direction axis of the Cartesian coordinate system represents the acquisition cost; mapping the expected number of objects to be acquired and the total acquisition cost to the Cartesian coordinate system to obtain the coordinate points corresponding to the expected number of objects to be acquired and the total acquisition cost; and evaluating the interest performance information for quality based on the position of the coordinate points in the Cartesian coordinate system to obtain the evaluation result.

[0186] For example, taking the quality assessment result as the quality assessment level, it is necessary to determine different promotion levels based on different quality assessment levels, and formulate promotion strategies corresponding to different promotion levels to promote the target audience to the target audience corresponding to different batches of interest performance information, so as to achieve a balance between the promotion cost of promoting the target audience and the conversion rate of interest performance information. In this embodiment, the quality assessment of a batch of interest performance information is carried out using two dimensions: the expected number of targets to be acquired for a batch of interest performance information and the total acquisition cost of the target audience for a batch of interest performance information. The Boston Matrix classifies different promotional targets based on two dimensions of information. Therefore, in this embodiment, the method of classifying different promotional targets using the Boston Matrix can be referenced to conduct quality assessment of interest performance information, so as to classify the quality of different batches of interest performance information into different quality assessment levels, thereby determining the interest performance information of different batches. The promotion strategy corresponding to the current information is then implemented, and the target audience is promoted to the target audience corresponding to different batches of interest performance information according to the promotion strategy corresponding to different batches of interest performance information. This aims to achieve a balance between the promotion cost of the target audience and the conversion rate of interest performance information. That is, the expected number of targets and the total acquisition cost corresponding to each batch of interest performance information can be obtained through the above process. A Cartesian coordinate system can be constructed in the same way as constructing the Boston Matrix, and the expected number of targets and the total acquisition cost corresponding to each batch of interest performance information can be mapped to the constructed Cartesian coordinate system to obtain the coordinate points corresponding to the expected number of targets and the total acquisition cost corresponding to each batch of interest performance information. Then, based on the position of the coordinate points corresponding to the expected number of targets and the total acquisition cost corresponding to each batch of interest performance information in the Cartesian coordinate system, the quality of each batch of interest performance information can be evaluated to obtain the quality evaluation level corresponding to each batch of interest performance information. Here, the first direction axis is the horizontal axis and the second direction axis is the vertical axis; or, the first direction axis is the vertical axis and the second direction axis is the horizontal axis.

[0187] For example, the constructed Cartesian coordinate system is as follows Figure 3As shown, the first region corresponds to the first quality assessment level, the second region corresponds to the second quality assessment level, the third region corresponds to the third quality assessment level, and the fourth region corresponds to the fourth quality assessment level. The first quality assessment level corresponds to the first promotion level, the second quality assessment level corresponds to the second promotion level, the third quality assessment level corresponds to the third promotion level, and the fourth quality assessment level corresponds to the fourth promotion level. The first quality assessment level is higher than the second quality assessment level, the second quality assessment level is higher than the third quality assessment level, the third quality assessment level is higher than the fourth quality assessment level, the first promotion level is higher than the second promotion level, the second promotion level is higher than the third promotion level, and the third promotion level is higher than the fourth promotion level. The coordinates of points P1, P2, P3, P4, P5, P6, P7, and P8 corresponding to the multiple batches of interest performance information DD1, DD2, DD3, DD4, DD5, DD6, DD7, and DD8 in the Cartesian coordinate system are shown below. Figure 3 As shown, it can be determined that interest performance information DD5 corresponds to the first quality assessment level of the first promotion level, indicating the degree of promotion to the target promotion object of the promoted object; interest performance information DD1 and DD2 correspond to the second quality assessment level of the second promotion level, indicating the degree of promotion to the target promotion object of the promoted ...

[0188] In some embodiments, the interest performance information is evaluated for quality based on the expected number of objects to be acquired and the total acquisition cost. After obtaining the quality evaluation result, the method further includes: promoting the target objects to the promoted objects according to the degree of promotion.

[0189] For example, suppose interest performance information DD5 corresponds to the first quality assessment level of the first promotion level for promoting the target promotion object to the first promoted object; interest performance information DD1 and DD2 correspond to the second quality assessment level of the second promotion level for promoting the target promotion object to the second promoted object; interest performance information DD6, DD7, and DD8 correspond to the third quality assessment level of the third promotion level for promoting the target promotion object to the third promoted object; and interest performance information DD3 and DD4 correspond to the fourth quality assessment level of the fourth promotion level for promoting the target promotion object to the fourth promoted object. Then, the target promotion object can be promoted to the promoted object corresponding to interest performance information DD5 according to the first promotion level; to the promoted object corresponding to interest performance information DD1 and DD2 according to the second promotion level; to the promoted object corresponding to interest performance information DD6, DD7, and DD8 according to the third promotion level; and to the promoted object corresponding to interest performance information DD3 and DD4 according to the fourth promotion level.

[0190] In one possible implementation, promoting the target target to the promoted object according to the degree of promotion includes: obtaining the promotion strategy corresponding to the degree of promotion; and promoting the target target to the promoted object according to the promotion strategy.

[0191] In this embodiment, a promotion strategy corresponding to each promotion level can be preset. After obtaining the quality assessment result, the promotion strategy corresponding to the promotion level of the target promotion object indicated by the quality assessment result can be obtained, and the target promotion object can be promoted to the target promotion object according to the promotion strategy.

[0192] Promotion strategies refer to the tactics employed to promote the target audience to the promoted target audience. These strategies can be developed by relevant personnel based on actual needs and include, but are not limited to, pushing content related to the target audience to the devices used by the promoted target audience, such as product introduction articles and videos; pushing promotional information related to the target audience to the devices used by the promoted target audience, such as acquisition cost discounts; and issuing discounted acquisition costs to the devices used by the promoted target audience. Discounted costs refer to the reduced acquisition costs achieved by acquiring the target audience. Promotion strategies corresponding to different levels of promotion can, for example, involve pushing content related to the target audience to the devices used by the promoted target audience at different frequencies. The higher the promotion level, the higher the push frequency. It should be noted that, to avoid excessive pushing of content related to the target audience causing resentment, the push frequency corresponding to the highest promotion level should be appropriately set. Promotion strategies corresponding to different levels of promotion can also include, for example, issuing different discounted acquisition costs to the devices used by the promoted target audience. The higher the promotion level, the higher the discounted cost, meaning the target audience can be acquired at a lower acquisition cost.

[0193] For example, suppose the quality assessment result corresponding to interest performance information DD5 indicates that the target audience is promoted to the target audience corresponding to interest performance information DD5 according to the first promotion level; the quality assessment results corresponding to interest performance information DD1 and DD2 indicate that the target audience is promoted to the target audience corresponding to interest performance information DD1 and DD2 according to the second promotion level; the quality assessment results corresponding to interest performance information DD6, DD7, and DD8 indicate that the target audience is promoted to the target audience corresponding to interest performance information DD6, DD7, and DD8 according to the third promotion level; and the quality assessment results corresponding to interest performance information DD3 and DD4 indicate that the target audience is promoted to the target audience corresponding to interest performance information DD3 and DD4 according to the fourth promotion level. The first promotion level corresponds to the first promotion strategy, the second promotion level corresponds to the second promotion strategy, the third promotion level corresponds to the third promotion strategy, and the fourth promotion level corresponds to the fourth promotion strategy. The first promotion strategy is to push content related to the target audience to the terminals used by the target audience at a first push frequency, the second promotion level corresponds to the third promotion strategy, and the third promotion level corresponds to the fourth promotion strategy. The broad strategy involves pushing content related to the target audience to the devices used by the promoted audience at a second push frequency. The third strategy involves pushing content related to the target audience to the devices used by the promoted audience at a third push frequency. The fourth strategy involves pushing content related to the target audience to the devices used by the promoted audience at a fourth push frequency. Specifically, if the first push frequency is greater than the second push frequency, the second push frequency is greater than the third push frequency, and the third push frequency is greater than the fourth push frequency, then the content related to the target audience can be pushed to the devices used by the promoted audience corresponding to interest performance information DD5 at the first push frequency, to the devices used by the promoted audience corresponding to interest performance information DD1 and DD2 at the second push frequency, to the devices used by the promoted audience corresponding to interest performance information DD6, DD7, and DD8 at the third push frequency, and to the devices used by the promoted audience corresponding to interest performance information DD3 and DD4 at the fourth push frequency.

[0194] It should be noted that the above is merely an example provided by the embodiments of this application and is not intended to limit this application. In practical applications, adjustments can be made according to actual needs, and no specific limitations are made here. For example, in practical applications, the first, second, and third regions can correspond to the same quality assessment level, and the fourth region can correspond to another quality assessment level. If the coordinates corresponding to a batch of interest performance information are located in any of the first, second, and third regions, the target promotion object is promoted to the promoted object corresponding to that batch of interest performance information according to the same promotion strategy; if the coordinates corresponding to that batch of interest performance information are located in the fourth region, the target promotion object is promoted to the promoted object corresponding to that batch of interest performance information according to another promotion strategy. Optionally, since the fourth region is a region with a small number of expected purchase objects and a low total acquisition cost, in order to save the promotion cost of promoting the target promotion object, for a batch of interest performance information whose corresponding coordinates are located in the fourth region, the target promotion object may not be promoted to its corresponding promoted object.

[0195] In an optional embodiment, after promoting the target promotion object to the promoted objects according to the promotion strategy, the process may further include: determining the target promoted objects that have obtained the corresponding service level of the target promotion object from the promoted objects; incrementally training the second probability prediction model based on the first object feature information of the target promoted object and the object feature information of each service level corresponding to the target promotion object to obtain the trained second probability prediction model, and then using the trained second probability prediction model to predict the second probability of the promoted object selecting each service level of the target promotion object corresponding to the interest expression information obtained afterward.

[0196] Specifically, the first object feature information of the target promoted object and the object feature information of each service tier corresponding to the target promoted object can be integrated as third sample data. Then, the first probability prediction model can be incrementally trained using this third sample data to obtain the trained third probability prediction model. The specific implementation of "integrating the first object feature information of the target promoted object and the object feature information of each service tier corresponding to the target promoted object as third sample data" can be found in the previous embodiments for obtaining the first or second sample data, and will not be repeated here.

[0197] Understandably, based on the first object feature information of the target promoted object and the object feature information of the corresponding service tier of the target promoted object, the second probability prediction model is incrementally trained so that the second probability prediction model can adapt to new data and thus better predict the second probability of the promoted object choosing each tier of the target promoted object corresponding to the interest performance information obtained later.

[0198] In this embodiment, interest performance information of different promoted objects towards their respective corresponding promoted objects is obtained. This interest performance information represents the different promoted objects' interest in acquiring their respective promoted objects, and the types of promoted objects corresponding to different promoted objects are the same. First object feature information and second object feature information of the promoted objects are obtained. Based on the first and second object feature information, a first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object. The number of promoted objects is obtained, and based on the first probability of each promoted object selecting each promoted object and the number of promoted objects, the expected number of objects to be acquired is predicted. The acquisition cost of the target promoted object among the promoted objects is obtained. Based on the probability and acquisition cost of each target audience selecting the target audience, the total acquisition cost of the interest performance information corresponding to the target audience is predicted. The interest performance information is then evaluated for quality based on the expected number of targets and the total acquisition cost. This evaluation result indicates the degree of promotion of the target audience to the target audience. This allows for quality evaluation of interest performance information, resulting in an evaluation result indicating the degree of promotion of the target audience to the target audience. Subsequently, the target audience can be promoted to the target audience according to this degree of promotion. Compared to a fixed, lower degree of promotion, this can improve the conversion rate of interest performance information.

[0199] Please see Figure 4 , Figure 4 This is a second flowchart illustrating the data processing method provided in this application embodiment. The process may include:

[0200] 201. Obtain interest performance information of different promoted objects for their respective corresponding promoted objects. Interest performance information represents the interest of different promoted objects in obtaining their respective corresponding promoted objects.

[0201] The following will further illustrate the data processing method provided in this application embodiment, taking different cloud storage service software as the target audience and a cloud drive offering four levels of speed-up services as an example. The speed-up service is mainly used to improve the download speed of resources downloaded through the cloud drive, such as files and videos. Different levels of speed-up services improve download speeds differently; the higher the acquisition cost of the speed-up service, the greater the improvement in download speed, and vice versa.

[0202] The target audience can be individuals with biometric information who are interested in acquiring cloud storage service software. These individuals typically express their interest in cloud storage service software through various channels, such as searching for such software using search engines or browsing it through software that showcases it. This expressed interest constitutes the interest expression information.

[0203] After different terminals obtain interest expression information, they can send the obtained interest expression information to a computer device with integrated data processing device, so that the computer device can obtain the interest expression information from different terminals.

[0204] 202. Obtain the first object characteristic information of the promoted object and the second object characteristic information of the promoted object.

[0205] It is understandable that computer devices integrated with data processing units can typically acquire a large amount of interest expression information. However, directly evaluating the quality of this large amount of interest expression information might lead to inaccurate results. Therefore, in this embodiment, the large amount of interest expression information can be divided into multiple batches according to appropriate classification rules. Each batch of interest expression information may include multiple interest expression information items.

[0206] It should be noted that, in this embodiment, if a certain interest expression information represents the interest of a certain promoted object in acquiring its corresponding promoted object, then the promoted object and the promoted object are respectively the promoted object and the promoted object corresponding to the interest expression information.

[0207] For example, after obtaining multiple batches of interest performance information, the first object feature information of the promoted object corresponding to each interest performance information in each batch of interest performance information can be obtained, as well as the second object feature information of the promoted object corresponding to each interest performance information in each batch of interest performance information.

[0208] It should be noted that since different promoted objects may correspond to the same promoted object, once the second object feature information of a promoted object is obtained, the second object feature information of that promoted object will not be obtained again.

[0209] In one possible implementation, after obtaining multiple batches of interest expression information, for each batch of interest expression information, the promotion object corresponding to each interest expression information in each batch of interest expression information can be integrated first, and the same promotion object can be integrated into one promotion object, and then the second object feature information of the integrated promotion object can be obtained.

[0210] It should be noted that since the processing procedure for each batch of interest expression information is the same, the following will use the processing procedure of one batch of interest expression information as an example to illustrate the data processing method provided in this embodiment.

[0211] For example, suppose the acquired interest expression information includes 200 interest expression pieces. Among them, 60 interest expression pieces represent the interest of 60 targeted users in acquiring cloud storage services; 20 interest expression pieces represent the interest of 20 targeted users in acquiring cloud storage service software C1; 20 interest expression pieces represent the interest of 20 targeted users in acquiring cloud storage service software C2; ​​30 interest expression pieces represent the interest of 30 targeted users in acquiring cloud storage service software C3; and 40 interest expression pieces represent the interest of 40 targeted users in acquiring cloud storage service software C3. The acquisition interest of storage service software C4 is represented by 30 interest expression information to indicate the acquisition interest of 30 promoted objects in cloud storage service software C5. Then, the first object feature information of each of the 200 promoted objects, as well as the second object feature information of cloud storage, cloud storage service software C1, cloud storage service software C2, cloud storage service software C3, cloud storage service software C4, and cloud storage service software C5 can be obtained.

[0212] 203. Based on the first object feature information and the second object feature information, predict the first probability of each promoted object selecting each promoted object through the first probability prediction model.

[0213] For example, using the first object characteristic information of a particular target among the 200 promoted targets, the second object characteristic information of the cloud storage service, the second object characteristic information of cloud storage service software C1, C2, C3, C4, and C5, a first probability prediction model can be used to predict the probability that the promoted target will choose the cloud storage service, the first probability that it will choose cloud storage service C1, the first probability that it will choose cloud storage service C2, the first probability that it will choose cloud storage service C3, the first probability that it will choose cloud storage service C4, and the first probability that it will choose cloud storage service C5. Similarly, the first object characteristic information, the second object characteristic information of the cloud storage service, and the second object characteristic information of cloud storage service C1 can be used to predict the probability that the promoted target will choose the cloud storage service, the second probability that it will choose cloud storage service C1, the second probability that it will choose cloud storage service C2, the second probability that it will choose cloud storage service C3, the second probability that it will choose cloud storage service C4, and the second probability that it will choose cloud storage service C5. The second object characteristic information of cloud storage service software C2, cloud storage service software C3, cloud storage service software C4, and cloud storage service software C5 are used to predict the first probability of other promoted objects choosing cloud storage, cloud storage service software C1, cloud storage service software C2, cloud storage service software C3, cloud storage service software C4, and cloud storage service software C5 through a first probability prediction model. This yields the first probability of each of the 200 promoted objects choosing cloud storage, cloud storage service software C1, cloud storage service software C2, cloud storage service software C3, cloud storage service software C4, and cloud storage service software C5.

[0214] 204. Obtain the number of promoted objects. Based on the first probability of each promoted object selecting each promoted object and the number of promoted objects, calculate the expected number of objects to be obtained through the first formula.

[0215] Once we obtain the probability that each of the 200 target users chooses a cloud storage service, the probability that they choose cloud storage service software C1, the probability that they choose cloud storage service software C2, the probability that they choose cloud storage service software C3, the probability that they choose cloud storage service software C4, the probability that they choose cloud storage service software C5, and the number of target users (200) are obtained, and the expected number of users can be calculated using the first formula.

[0216] The first formula is:

[0217] Where E(j) is the expected number of objects to be obtained, and P it Let c be the probability of selecting the i-th promotion target for the t-th promotion target, such as the probability of selecting a cloud storage service, the probability of selecting cloud storage service software C1, the probability of selecting cloud storage service software C2, the probability of selecting cloud storage service software C3, the probability of selecting cloud storage service software C4, and the probability of selecting cloud storage service software C5. j This refers to the number of people being promoted, i.e., the amount of information expressing their interests.

[0218] 205. Based on the first object feature information and the object feature information of each service tier corresponding to the target promotion object in the promotion object, the second probability of each promoted object selecting each service tier of the target promotion object is predicted through the second probability prediction model.

[0219] For example, using the first object feature information of a certain promoted object among the 200 promoted objects and the object feature information of each service tier corresponding to the cloud storage, a second probability prediction model can be used to predict the second probability of that promoted object choosing the first, second, third, and fourth service tiers of the cloud storage. Similarly, using the first object feature information of other promoted objects and the object feature information of each service tier corresponding to the cloud storage, a second probability prediction model can be used to predict the second probability of other promoted objects choosing the first, second, third, and fourth service tiers of the cloud storage. This yields the second probability of each of the 200 promoted objects choosing the first, second, third, and fourth service tiers of the cloud storage.

[0220] 206. Obtain the acquisition cost of each service tier corresponding to the target promotion object, and calculate the total acquisition cost of the interest performance information corresponding to the target promotion object through the second formula based on the first probability of each promoted object selecting the target promotion object, the second probability of each promoted object selecting each service tier of the target promotion object, and the acquisition cost of each service tier corresponding to the target promotion object.

[0221] The second formula is: R(T)=∑P t P tk C k

[0222] in, R(T) represents the target audience, such as the total cost of acquiring interest-based information related to cloud storage, and P represents the total cost of acquiring such information. t For the target audience t, select the target audience for promotion, such as the first probability of selecting a cloud storage service, P. tk For the promoted object t, the second probability of selecting the k-th tier of service corresponding to the target promoted object, such as the speed-up service of the k-th tier corresponding to cloud storage, is C. k The cost of acquiring the service corresponding to the k-th tier for the target audience, such as the speed-up service for cloud storage corresponding to the k-th tier.

[0223] 207. Construct a Cartesian coordinate system, where the first direction axis of the Cartesian coordinate system represents the quantity and the second direction axis of the Cartesian coordinate system represents the acquisition cost.

[0224] 208. Map the expected number of objects to be acquired and the total acquisition cost to a Cartesian coordinate system to obtain the coordinate points corresponding to the expected number of objects to be acquired and the total acquisition cost.

[0225] 209. Based on the position of the coordinate point in the Cartesian coordinate system, the quality of interest expression information is evaluated to obtain the quality evaluation result. The quality evaluation result is used to indicate the degree of promotion to the target audience.

[0226] 210. Obtain the promotion strategy corresponding to the promotion level, and promote the target audience to the promoted audience according to the promotion strategy.

[0227] The quality assessment results include the quality assessment level, and the promotion level includes the promotion level.

[0228] For example, the constructed Cartesian coordinate system is as follows Figure 5 As shown, the first area corresponds to the first quality assessment level, the second area corresponds to the second quality assessment level, the third area corresponds to the third quality assessment level, and the fourth area corresponds to the fourth quality assessment level. The third quality assessment result indicates that the degree of promotion to the target audience is the third promotion level. The promotion strategy corresponding to the third promotion level is to push content related to the target audience to the terminals used by the target audience at the third promotion frequency. Assume that the coordinates of a batch of interest expression information are located in a Cartesian coordinate system as follows: Figure 5 As shown, the quality assessment result of this batch of interest performance information can be determined as the third quality assessment level of the third promotion level, indicating the degree of promotion to the target audience. Then, content related to the target audience can be pushed to the terminal used by the target audience corresponding to this batch of interest performance information at the third promotion frequency, so that the target audience can understand the advantages of the target audience, thereby making it easier for the target audience to choose the target audience and pay the corresponding acquisition cost to acquire the target audience.

[0229] In this embodiment, interest expression information of different promoted objects towards their respective corresponding promoted objects is obtained. This interest expression information represents the interest shown by different promoted objects in acquiring their respective corresponding promoted objects, and the types of promoted objects corresponding to different promoted objects are the same. First object feature information and second object feature information of the promoted objects are obtained. Based on the first and second object feature information, a first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object. The number of promoted objects is obtained, and based on the first probability of each promoted object selecting each promoted object and the number of promoted objects, the expected number of objects to be acquired is predicted. The acquisition of the target promoted object among the promoted objects is then determined. The cost is calculated, and based on the probability and acquisition cost of each target audience selecting the target audience, the total acquisition cost of the interest performance information corresponding to the target audience is predicted. The interest performance information is then evaluated for quality based on the expected number of targets and the total acquisition cost. This quality evaluation result indicates the degree of promotion of the target audience to the target audience. This allows for quality evaluation of interest performance information to obtain a quality evaluation result indicating the degree of promotion of the target audience to the target audience. Subsequently, the target audience can be promoted to the target audience according to this degree of promotion. Compared to a fixed, lower degree of promotion, this can improve the conversion rate of interest performance information.

[0230] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application. The data processing device includes: a first information acquisition module 301, a second information acquisition module 302, a probability prediction module 303, a quantity prediction module 304, a cost prediction module 305, and a quality assessment module 306.

[0231] The first information acquisition module 301 is used to acquire interest performance information of different promoted objects for their respective corresponding promoted objects. The interest performance information represents the acquisition interest of different promoted objects for their respective corresponding promoted objects. The types of promoted objects corresponding to different promoted objects are the same.

[0232] The second information acquisition module 302 is used to acquire the first object feature information of the promoted object and the second object feature information of the promoted object.

[0233] The probability prediction module 303 is used to predict the first probability of each promoted object selecting each promoted object based on the first object feature information and the second object feature information, through the first probability prediction model.

[0234] The quantity prediction module 304 is used to obtain the number of promoted objects and predict the expected number of objects to be obtained based on the first probability of each promoted object selecting each promoted object and the number of promoted objects.

[0235] The cost prediction module 305 is used to obtain the acquisition cost of the target promotion object among the promotion objects, and predict the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability and acquisition cost of each promoted object selecting the target promotion object.

[0236] The quality assessment module 306 is used to assess the quality of interest performance information based on the expected number of targets and the total acquisition cost, and obtain the quality assessment result. The quality assessment result is used to indicate the degree of promotion of the target target to the promoted targets.

[0237] Optionally, the target audience offers services at multiple tiers, with different acquisition costs for each tier. The probability prediction module 303 is also used to perform:

[0238] Based on the first object feature information and the object feature information of each service tier corresponding to the target promotion object, the second probability prediction model is used to predict the second probability of each promoted object selecting each service tier.

[0239] The cost forecasting module 305 is specifically used for the following:

[0240] The cost of acquiring each service tier corresponding to the target audience;

[0241] Based on the first probability that each promoted object selects the target promoted object, the second probability that each promoted object selects a service at each tier, and the acquisition cost of the target promoted object for each tier of service, the total acquisition cost of the interest performance information corresponding to the target promoted object is predicted.

[0242] Optionally, the cost prediction module 305 is specifically used to perform:

[0243] Using the first probability of each promoted object selecting the target promoted object and the second probability of each promoted object selecting each service tier as weights, the acquisition cost of each service tier corresponding to the target promoted object is weighted and summed to obtain the total acquisition cost of the interest expression information corresponding to the target promoted object.

[0244] Optionally, the quality assessment module 306 is specifically used to perform:

[0245] Construct a Cartesian coordinate system, where the first direction axis of the Cartesian coordinate system represents the quantity, and the second direction axis of the Cartesian coordinate system represents the acquisition cost;

[0246] Map the expected number of objects to be acquired and the total acquisition cost to a Cartesian coordinate system to obtain the coordinate points corresponding to the expected number of objects to be acquired and the total acquisition cost.

[0247] Based on the position of the coordinate point in the Cartesian coordinate system, the quality of the interest expression information is evaluated, and the quality evaluation result is obtained.

[0248] Optionally, the quantity prediction module 304 is specifically used to perform:

[0249] The probability of each promoted object selecting each other is added together to obtain the sum of probabilities.

[0250] The desired number of objects is obtained by dividing the sum of probabilities and the number of objects to be promoted.

[0251] Optionally, the data processing apparatus further includes an object promotion module, which is used to perform:

[0252] The target audience is promoted to the promoted audience based on the degree of promotion.

[0253] Optionally, the object promotion module is specifically used to perform:

[0254] Obtain the corresponding promotion strategy based on the promotion level;

[0255] Promote the target audience to the target audience according to the promotion strategy.

[0256] In this embodiment, the first information acquisition module 301 acquires interest performance information of different promoted objects for their respective corresponding promoted objects. This interest performance information represents the interest shown by different promoted objects in acquiring their respective corresponding promoted objects, and the types of promoted objects corresponding to different promoted objects are the same. The second information acquisition module 302 acquires first object feature information of the promoted object and second object feature information of the promoted object. The probability prediction module 303, based on the first and second object feature information, predicts the first probability of each promoted object selecting each promoted object using a first probability prediction model. The quantity prediction module 304 acquires the quantity of promoted objects and, based on the first probability of each promoted object selecting each promoted object and the quantity of promoted objects, predicts the expected quantity of objects to be acquired. The cost prediction module... Block 305 obtains the acquisition cost of the target promotion object among the promotion objects, and predicts the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability of each promoted object selecting the target promotion object and the acquisition cost; the quality assessment module 306 performs a quality assessment on the interest performance information based on the expected number of acquired objects and the total acquisition cost, and obtains a quality assessment result. The quality assessment result is used to indicate the degree of promotion of the target promotion object to the promoted object. It can realize the quality assessment of interest performance information and obtain a quality assessment result indicating the degree of promotion of the target promotion object to the promoted object. Then, the target promotion object can be promoted to the promoted object according to the degree of promotion. Compared with the scheme of promoting the target promotion object to the promoted object according to a fixed lower degree of promotion, it can improve the conversion rate of interest performance information.

[0257] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed on a computer, causes the computer to perform the data processing method provided in this embodiment.

[0258] This application also provides a computer device, including a memory and a processor, wherein the processor executes the data processing method provided in this embodiment by calling a computer program stored in the memory.

[0259] Furthermore, embodiments of this application also provide a computer device, which can be a terminal or a server, such as... Figure 7 As shown, it illustrates a structural schematic diagram of the computer device involved in the embodiments of this application, specifically:

[0260] The computer device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 7The computer device structure shown does not constitute a limitation on the computer device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0261] in:

[0262] The processor 401 is the control center of the computer device. It connects various parts of the computer device via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402, thereby providing overall control of the computer device. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0263] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0264] The computer device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0265] The computer device may also include an input unit 404, which can be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0266] Although not shown, the computer device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the computer device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to realize various functions, as follows:

[0267] Acquire interest information of different promoted objects for their respective corresponding promoted objects. Interest information represents the interest shown by different promoted objects in acquiring their respective corresponding promoted objects. The types of promoted objects corresponding to different promoted objects are the same.

[0268] Obtain the first object characteristic information of the promoted object, and the second object characteristic information of the promoted object;

[0269] Based on the first object feature information and the second object feature information, the first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object.

[0270] Obtain the number of promoted objects, and predict the expected number of objects to be obtained based on the first probability of each promoted object selecting each promoted object and the number of promoted objects;

[0271] Obtain the acquisition cost of the target promotion object among the promotion objects, and predict the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability and acquisition cost of each promoted object selecting the target promotion object;

[0272] Based on the expected number of targets and the total acquisition cost, the interest performance information is evaluated to obtain the quality evaluation results. The quality evaluation results are used to indicate the degree of promotion to the target audience.

[0273] Optionally, the target audience offers services at multiple tiers, with different acquisition costs for each tier. Before predicting the total acquisition cost of the target audience's interest performance information based on the probability and acquisition cost of each target audience selecting the target audience, the process also includes:

[0274] Based on the first object feature information and the object feature information of each service tier corresponding to the target promotion object, the second probability prediction model is used to predict the second probability of each promoted object selecting each service tier.

[0275] The cost of acquiring the target audience within the target audience includes:

[0276] The cost of acquiring each service tier corresponding to the target audience;

[0277] Based on the probability and acquisition cost of each promoted object selecting the target promoted object, predict the total acquisition cost of the interest performance information corresponding to the target promoted object, including:

[0278] Based on the first probability that each promoted object selects the target promoted object, the second probability that each promoted object selects a service at each tier, and the acquisition cost of the target promoted object for each tier of service, the total acquisition cost of the interest performance information corresponding to the target promoted object is predicted.

[0279] Optionally, based on the first probability that each promoted object selects the target promoted object, the second probability that each promoted object selects a service at each tier, and the acquisition cost of the service at each tier corresponding to the target promoted object, the total acquisition cost of the interest performance information corresponding to the target promoted object is predicted, including:

[0280] Using the first probability of each promoted object selecting the target promoted object and the second probability of each promoted object selecting each service tier as weights, the acquisition cost of each service tier corresponding to the target promoted object is weighted and summed to obtain the total acquisition cost of the interest expression information corresponding to the target promoted object.

[0281] Optionally, the interest performance information is evaluated for quality based on the expected number of targets and the total acquisition cost, resulting in a quality evaluation result, including:

[0282] Construct a Cartesian coordinate system, where the first direction axis of the Cartesian coordinate system represents the quantity, and the second direction axis of the Cartesian coordinate system represents the acquisition cost;

[0283] Map the expected number of objects to be acquired and the total acquisition cost to a Cartesian coordinate system to obtain the coordinate points corresponding to the expected number of objects to be acquired and the total acquisition cost.

[0284] Based on the position of the coordinate point in the Cartesian coordinate system, the quality of the interest expression information is evaluated, and the quality evaluation result is obtained.

[0285] Optionally, based on the first probability of each promoted object selecting each promoted object and the number of promoted objects, the expected number of objects to be acquired is predicted, including:

[0286] Add the first probability of each promoted object selecting each promoted object to obtain the sum of probabilities;

[0287] The desired number of objects is obtained by dividing the sum of probabilities and the number of objects to be promoted.

[0288] Optionally, based on the expected number of targets and the total acquisition cost, the interest performance information is evaluated for quality. After obtaining the quality evaluation results, the following steps are also included:

[0289] The target audience is promoted to the promoted audience based on the degree of promotion.

[0290] Optionally, the target audience is promoted to the promoted audience according to the degree of promotion, including:

[0291] Obtain the promotion strategy corresponding to the promotion level;

[0292] Promote the target audience to the target audience according to the promotion strategy.

[0293] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be accomplished by instructions, or by instructions controlling related hardware. These instructions can be stored in a storage medium and loaded and executed by a processor.

[0294] Therefore, embodiments of this application provide a storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the data processing methods provided in embodiments of this application. For example, the instructions can execute the following steps:

[0295] Acquire interest information of different promoted objects for their respective corresponding promoted objects. Interest information represents the interest shown by different promoted objects in acquiring their respective corresponding promoted objects. The types of promoted objects corresponding to different promoted objects are the same.

[0296] Obtain the first object characteristic information of the promoted object, and the second object characteristic information of the promoted object;

[0297] Based on the first object feature information and the second object feature information, the first probability prediction model is used to predict the first probability of each promoted object selecting each promoted object.

[0298] Obtain the number of promoted objects, and predict the expected number of objects to be obtained based on the first probability of each promoted object selecting each promoted object and the number of promoted objects;

[0299] Obtain the acquisition cost of the target promotion object among the promotion objects, and predict the total acquisition cost of the interest performance information corresponding to the target promotion object based on the first probability and acquisition cost of each promoted object selecting the target promotion object;

[0300] Based on the expected number of targets and the total acquisition cost, the interest performance information is evaluated to obtain the quality evaluation results. The quality evaluation results are used to indicate the degree of promotion to the target audience.

[0301] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0302] Since the instructions stored in the storage medium can execute the steps of any of the data processing methods provided in the embodiments of this application, the beneficial effects that any of the data processing methods provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0303] According to one aspect of this application, a computer program content or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the data processing method provided in the above-described embodiments.

[0304] The data processing method, apparatus, computer equipment, and storage medium provided in the embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A data processing method, characterized by, The method comprises the following steps: obtaining interest expression information of different promoted objects for respective corresponding promoted objects, wherein the interest expression information represents the interest of different promoted objects in obtaining respective corresponding promoted objects, and the types of the corresponding promoted objects of different promoted objects are the same; obtaining first object feature information of the promoted objects and second object feature information of the promoted objects; predicting, by a first probability prediction model, a first probability of each promoted object selecting each promoted object according to the first object feature information and the second object feature information; obtaining the number of promoted objects, and predicting an expected number of objects according to the first probability of each promoted object selecting each promoted object and the number of promoted objects; obtaining the cost of a target promoted object in the promoted objects, and predicting a total cost of the target promoted object corresponding to the interest expression information according to the first probability of each promoted object selecting the target promoted object and the cost; performing quality evaluation on the interest expression information according to the expected number of objects and the total cost to obtain a quality evaluation result, wherein the quality evaluation result is used to indicate the promotion degree of the target promoted object to the promoted objects.

2. The data processing method according to claim 1, characterized in that, The target promoted object provides services in multiple gears, the cost of the target promoted object corresponding to services in different gears is different, and before predicting the total cost of the target promoted object corresponding to the interest expression information according to the first probability of each promoted object selecting the target promoted object and the cost, the method further comprises the following steps: predicting, by a second probability prediction model, a second probability of each promoted object selecting services in each gear according to the first object feature information and object feature information of the target promoted object corresponding to services in each gear; The method further comprises the following steps: obtaining the cost of the target promoted object corresponding to services in each gear; The method further comprises the following steps: predicting the total cost of the target promoted object corresponding to the interest expression information according to the first probability of each promoted object selecting the target promoted object, the second probability of each promoted object selecting services in each gear, and the cost of the target promoted object corresponding to services in each gear.

3. The data processing method according to claim 2, characterized in that, The method further comprises the following steps: predicting the total cost of the target promoted object corresponding to the interest expression information according to the first probability of each promoted object selecting the target promoted object, the second probability of each promoted object selecting services in each gear, and the cost of the target promoted object corresponding to services in each gear. The total acquisition cost of the target promotion object corresponding to each level of service is obtained by performing a weighted summation operation on the acquisition cost of the target promotion object corresponding to each level of service, with each first probability of each promotion object being selected by each promotion object and each second probability of each promotion object selecting the service of each level as weights.

4. The data processing method according to any one of claims 1 to 3, characterized in that, The quality evaluation result is obtained by performing quality evaluation on the interest expression information according to the expected acquisition object quantity and the total acquisition cost. A plane rectangular coordinate system is constructed, wherein a first direction axis of the plane rectangular coordinate system represents quantity, and a second direction axis of the plane rectangular coordinate system represents acquisition cost. The expected acquisition object quantity and the total acquisition cost are mapped to the plane rectangular coordinate system to obtain a coordinate point corresponding to the expected acquisition object quantity and the total acquisition cost. The quality evaluation result is obtained by performing quality evaluation on the interest expression information according to the position of the coordinate point in the plane rectangular coordinate system.

5. The data processing method according to any one of claims 1 to 3, characterized in that, The expected acquisition object quantity is predicted according to the first probability of each promotion object being selected by each promotion object and the quantity of the promotion object, including: The first probability of each promotion object being selected by each promotion object is added to obtain a probability sum. The expected acquisition object quantity is obtained by performing a division operation on the probability sum and the quantity of the promotion object.

6. The data processing method according to any one of claims 1 to 3, characterized in that, After the quality evaluation result is obtained by performing quality evaluation on the interest expression information according to the expected acquisition object quantity and the total acquisition cost, the method further includes: The target promotion object is promoted to the promotion object according to the promotion degree.

7. The data processing method according to claim 6, characterized in that, The target promotion object is promoted to the promotion object according to the promotion degree, including: An acquisition strategy corresponding to the promotion degree is obtained. The target promotion object is promoted to the promotion object according to the promotion strategy.

8. A data processing apparatus, characterized by, The method includes: A first information acquisition module is configured to acquire interest expression information of different promotion objects for respective corresponding promotion objects, wherein the interest expression information represents acquisition interest of different promotion objects in acquiring respective corresponding promotion objects, and the types of the corresponding promotion objects of different promotion objects are the same. A second information acquisition module is configured to acquire first object feature information of the promotion object and second object feature information of the promotion object. A probability prediction module is configured to predict a first probability of each promotion object being selected by each promotion object according to the first object feature information and the second object feature information through a first probability prediction model. A quantity prediction module is configured to acquire a quantity of the promotion object, and predict an expected acquisition object quantity according to the first probability of each promotion object being selected by each promotion object and the quantity of the promotion object. A cost prediction module is configured to acquire an acquisition cost of a target promotion object in the promotion object, and predict a total acquisition cost of the target promotion object corresponding to the interest expression information according to the first probability of each promotion object being selected by the target promotion object and the acquisition cost. A quality evaluation module is configured to perform quality evaluation on the interest expression information according to the expected number of acquisition objects and the total acquisition cost, and obtain a quality evaluation result, which is used to indicate a promotion degree of promoting the target promotion object to the promoted object.

9. A computer device, comprising: The computer readable storage medium is configured to store a computer program, and the computer program is loaded by the processor to execute the data processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium is configured to store a computer program, and the computer program is loaded by the processor to execute the data processing method according to any one of claims 1 to 7.

Images