CN116957668A

CN116957668A - Data processing method and device, electronic equipment and storage medium

Info

Publication number: CN116957668A
Application number: CN202211440557.8A
Authority: CN
Inventors: 赵振岐
Original assignee: Shenzhen Tencent Computer Systems Co Ltd
Current assignee: Shenzhen Tencent Computer Systems Co Ltd
Priority date: 2022-11-17
Filing date: 2022-11-17
Publication date: 2023-10-27

Abstract

The application discloses a data processing method, a device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence, wherein the method comprises the following steps: acquiring first object resource information and preset release parameters of resources to be recalled, wherein the first object resource information represents object attributes of a target object and non-release parameter attributes of the resources to be recalled; inputting the first object resource information and the preset release parameter into a resource recall model to carry out recall identification, so as to obtain first recall index data, wherein a recall index identification function in the resource recall model is a nonlinear positive correlation function, and the recall index identification function takes the sum of the object resource characteristics corresponding to the first object resource information and the preset release parameter as an independent variable, so that the first recall index data and the preset release parameter are positively correlated; and screening out target recall resources corresponding to the target object from the resources to be recalled based on the first recall index data. By utilizing the technical scheme provided by the application, the rationality and the effectiveness of resource recall can be improved.

Description

Data processing method and device, electronic equipment and storage medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to a data processing method, apparatus, electronic device, and storage medium.

Background

With research and progress of artificial intelligence technology, the artificial intelligence technology is widely applied in the resource release process of a resource release system in a plurality of fields, such as deep learning technology in the artificial intelligence technology.

At present, in the resource release process, resource recall is performed by combining resource attribute information of resources to be recalled, object attribute information of resource release objects and a resource recall model, so that target recall resources are selected; then, combining ECPM (effective cost per mille, estimated cost of thousands of times of display) to perform resource sequencing, and further selecting resources to be finally put; however, in the above-mentioned resource delivery process, although the ECPMS that is positively correlated with the delivery parameters (virtual resources to which the corresponding delivery resources are delivered by the resource delivery party) are combined in the resource sequencing stage, the resource recall does not pay attention to the indexes correlated with the delivery parameters, so that whether the resources are recalled or not is not correlated with the delivery parameters of the resource delivery party, even if the delivery party increases the delivery, the probability of the corresponding resources being recalled is not improved, the problem of unreasonable resource recall processing is caused, meanwhile, the enthusiasm of the additional delivery of the resource delivery party is also influenced, not only the ECPM in the resource sequencing stage is damaged, the loss is caused to the whole delivery platform, but also the problem of monotonous quality reduction of the delivered resources in the delivery system is caused. Thus, there is a need to provide a more efficient solution.

Disclosure of Invention

The application provides a data processing method, a device, equipment, a storage medium and a computer program product, which can improve the rationality and the effectiveness of resource recall and can also improve the richness and the diversity of released resources in a release system.

In one aspect, the present application provides a data processing method, the method including:

acquiring first object resource information and preset release parameters of resources to be recalled, wherein the first object resource information represents object attributes of a target object and non-release parameter attributes of the resources to be recalled;

inputting the first object resource information and the preset release parameters into a resource recall model to carry out recall identification to obtain first recall index data, wherein a recall index identification function in the resource recall model is a nonlinear positive correlation function, and the recall index identification function takes the sum of object resource characteristics corresponding to the first object resource information and the preset release parameters as independent variables so as to enable the first recall index data to be in positive correlation with the preset release parameters;

and screening out the target recall resource corresponding to the target object from the to-be-recalled resources based on the first recall index data.

Another aspect provides a data processing apparatus, the apparatus comprising:

the first information acquisition module is configured to acquire first object resource information and preset release parameters of resources to be recalled, wherein the first object resource information represents object attributes of a target object and non-release parameter attributes of the resources to be recalled;

the first recall identification module is configured to perform recall identification on the first object resource information and the preset release parameter input resource recall model to obtain first recall index data, a recall index identification function in the recall index identification layer is a nonlinear positive correlation function, and the recall index identification function takes the sum of the object resource characteristics corresponding to the first object resource information and the preset release parameter as an independent variable, so that the first recall index data and the preset release parameter are positively correlated;

and the target recall resource screening module is configured to perform screening of target recall resources corresponding to the target object from the resources to be recalled based on the first recall index data.

Another aspect provides an electronic device, comprising: a processor;

A memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the data processing method of any of the above.

Another aspect provides a computer readable storage medium, which when executed by a processor of an electronic device, causes the electronic device to perform any of the data processing methods described above.

Another aspect provides a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions to cause the computer device to perform the data processing methods provided in the various alternative implementations described above.

The data processing method, the device, the equipment, the storage medium and the computer program product provided by the application have the following technical effects:

in the recall identification process of inputting the first object resource information and the preset release parameter of the resource to be recalled into the resource recall model, the independent variable of the recall index identification function which is not linearly and positively correlated in the resource recall model is set as the sum of the object resource characteristics corresponding to the first object resource information and the preset release parameter, so that the obtained first recall index data is positively correlated with the preset release parameter on the basis of learning the object and the attribute characteristics of the resource to be recalled, the relevance between the recalled resource and the release (release parameter) of the resource release party is realized, the sensitivity of the recall index data to the release parameter is improved, the target recall resource is screened out from the resource to be recalled on the basis of the first recall index data which is positively correlated with the preset release parameter, the rationality and the effectiveness of the recall process of the resource can be greatly improved, the aggressiveness of the additional release of the resource release party can be effectively improved, and the richness and the diversity of the resources in the release system can be improved while the release system is improved.

Drawings

In order to more clearly illustrate the embodiments of the application or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the application, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of an application environment of a data processing method according to an embodiment of the present application;

FIG. 2 is a schematic flow chart of a data processing method according to an embodiment of the present application;

FIG. 3 is a schematic flow chart of obtaining normalized nonlinear parameters according to an embodiment of the present application;

FIG. 4 is a flowchart of another data processing method according to an embodiment of the present application;

FIG. 5 is a flow chart of generating recall direction analysis results according to an embodiment of the present application;

FIG. 6 is a flow chart of a method for performing direction-of-change consistency analysis to obtain a recall direction analysis result based on a first initial placement parameter, a first adjusted placement parameter, second recall index data, and third recall index data according to an embodiment of the present application;

FIG. 7 is a schematic diagram of recall direction analysis results provided in accordance with an exemplary embodiment;

FIG. 8 is a flow chart of generating an amplitude analysis result according to an embodiment of the present application;

FIG. 9 is a schematic flow chart of a process for performing amplitude consistency analysis to obtain an amplitude analysis result based on fourth recall indicator data, fifth recall indicator data, sixth recall indicator data, seventh recall indicator data, first recommendation indicator data, second recommendation indicator data, third recommendation indicator data, and fourth recommendation indicator data according to an embodiment of the present application;

FIG. 10 is a schematic diagram of a data processing apparatus according to an embodiment of the present application;

FIG. 11 is a block diagram of an electronic device for data processing provided by an embodiment of the present application;

fig. 12 is a block diagram of another electronic device for data processing provided by an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.

It should be noted that the terms "first," "second," and the like in the description and the claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Referring to fig. 1, fig. 1 is a schematic diagram of an application environment of a data processing method according to an embodiment of the present application, where the application environment includes at least a server 100 and a terminal 200.

In an alternative embodiment, the server 100 may be configured to train a resource recall model, and accordingly, may perform resource recall processing based on the resource recall model; further, the server 100 may further order the recalled resources, so as to provide corresponding resource recommendation services for the user at the terminal side, and implement release of corresponding resources. The server 100 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud computing services.

In an alternative embodiment, the terminal 200 may provide a resource recommendation service to the user. Specifically, the terminal 200 may include, but is not limited to, smart phones, desktop computers, tablet computers, notebook computers, smart speakers, digital assistants, augmented reality (augmented reality, AR)/Virtual Reality (VR) devices, smart wearable devices, vehicle terminals, smart televisions, and other types of electronic devices; or software running on the electronic device, such as an application, applet, etc. Operating systems running on the electronic device in embodiments of the present application may include, but are not limited to, android systems, IOS systems, linux, windows, and the like.

In addition, fig. 1 shows only an example of an application environment of the data processing method.

In the embodiment of the present disclosure, the server 100 and the terminal 200 may be directly or indirectly connected through a wired or wireless communication method, which is not limited herein.

In the following description, a data processing method according to an embodiment of the present application is described, and fig. 2 is a schematic flow chart of a data processing method according to an embodiment of the present application, where the method operation steps of the embodiment or the flowchart are provided, but more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps recited in the embodiments is merely one way of performing the order of steps and does not represent a unique order of execution. When implemented in a real system or server product, the methods illustrated in the embodiments or figures may be performed sequentially or in parallel (e.g., in a parallel processor or multithreaded environment). As shown in fig. 2, the method may include:

S201: and acquiring the information of the first object resource and preset release parameters of the resources to be recalled.

In a specific embodiment, the first object resource information may represent an object attribute of the target object and a non-delivery parameter attribute of the resource to be recalled; specifically, the target object may be any user account in the delivery system, and the object attribute may be information capable of reflecting characteristics of the object, for example, user account identification, historical operation information (operation information of resources for delivering system content in a preset historical period), and the like. The resources to be recalled can be resources (a plurality of resources to be put) which need to be put; optionally, the resource to be recalled may be multimedia data introducing the target to be recommended; the target to be recommended can be commodities, shops, living broadcast rooms and the like, and the multimedia data can be static media data such as pictures and texts and dynamic media data such as videos. The preset release parameter may be a virtual resource that the resource release party releases for releasing the resource to be recalled. The non-delivery parameter attribute of the resource to be recalled may be an attribute other than a preset delivery parameter in the resource attribute of the resource to be recalled. The resource attribute of the resource to be recalled can be information capable of reflecting the characteristics of the resource to be recalled, such as a resource identifier of the resource to be recalled (identifier for distinguishing different release resources), an identifier of a target to be recommended corresponding to the resource to be recalled (identifier for distinguishing different targets to be recommended), category information corresponding to the target to be recommended, preset release parameters and the like.

In a specific embodiment, the information meeting the input requirement of the resource recall model can be constructed by combining the object attribute of the target object and the non-release parameter attribute of the resource to be recalled; optionally, the object attribute of the target object and the non-delivery parameter attribute of the resource to be recalled can be used as a whole as input of a resource recall model; correspondingly, the first object resource information can be an object resource vector constructed based on the object attribute of the target object and the non-delivery parameter attribute of the resource to be recalled; optionally, the object attribute of the target object and the non-delivery parameter attribute of the resource to be recalled can be respectively converted into corresponding vectors and then used as the input of a resource recall model; correspondingly, the first object resource information may include an object attribute vector and a non-delivery parameter attribute vector; wherein, the object attribute vector can be a vector constructed based on the object attribute; the non-delivery parameter attribute vector may be a vector constructed based on non-delivery parameter attributes. Specifically, the attribute is converted into a corresponding vector by combining a preset coding algorithm.

S203: and inputting the first object resource information and the preset release parameters into a resource recall model to carry out recall identification, so as to obtain first recall index data.

In a specific embodiment, the resource recall model may be a model obtained by performing resource recall training on a preset deep learning model based on preset training data; specifically, the preset training data may include a plurality of sample object resource information, a delivery parameter of a sample resource corresponding to each sample object resource information, and a recall index data tag corresponding to each sample object resource information. Specifically, the sample object resource information may represent an object attribute of the sample object and a non-delivery parameter attribute of the sample object corresponding to the sample resource; the sample objects corresponding to the plurality of sample object resource information can comprise a positive sample object and a negative sample object, wherein the sample resources corresponding to the positive sample object can be input resources for executing preset interaction operation for the positive sample object; the sample resource corresponding to the negative sample object can be the input resource of the negative sample object after the preset interactive operation is executed; specifically, the preset interaction operation can be set in combination with actual application, such as clicking, checking and other operations.

In a specific embodiment, the preset deep learning model may be a semantic vector recall model of a dual-tower structure, and correspondingly, the first object resource information may include an object attribute vector and a non-delivery parameter attribute vector; optionally, the preset deep learning model may be a convolutional neural network, and correspondingly, the first object resource information may be an object resource vector constructed based on an object attribute of the target object and a non-delivery parameter attribute of the resource to be recalled.

In a specific embodiment, a specific network structure of the preset deep learning model can be set in combination with actual application requirements, and the preset deep learning model can comprise a feature extraction layer to be trained and a recall index identification layer to be trained; accordingly, the resource recall model may include a feature extraction layer and a recall index identification layer. The feature extraction layer may be configured to extract semantic information of object attributes and non-delivery parameter attributes from the object resource information. In the case that the first object resource information includes an object attribute vector and a non-delivery parameter attribute vector, the semantic information extracted from the first object resource information by the feature extraction layer may include an object semantic feature and a resource semantic feature, and the object resource feature corresponding to the first object resource information may be a dot product feature of the object semantic feature and the resource semantic feature. Optionally, in the case that the first object resource information is an object resource vector, the semantic information extracted by the feature extraction layer from the first object resource information may include an object resource semantic feature, and correspondingly, the object resource feature corresponding to the first object resource information may be an object resource semantic feature. The recall index recognition layer can be used for carrying out recall index recognition based on the semantic information extracted by the feature extraction layer and preset delivery parameters so as to obtain recall index data.

In a specific embodiment, in the recall indicator identification layer, recall indicator data may be calculated in conjunction with a recall indicator identification function. Accordingly, the recall indicator identification function may be a function for generating recall indicator data, that is, the dependent variable of the recall indicator identification function is the recall indicator data. Specifically, recall index data may characterize a degree of association between an object and a corresponding released resource; correspondingly, the first recall index data can represent the association degree between the target object and the resource to be recalled.

In a specific embodiment, the recall indicator recognition function in the resource recall model (i.e., the recall indicator recognition function in the recall indicator recognition layer) may be a nonlinear positive correlation function, and the recall indicator recognition function uses the sum of the object resource characteristics corresponding to the first object resource information and the preset release parameter as the independent variable, so that the first recall indicator data and the preset release parameter are positively correlated _。

In an alternative embodiment, as shown in fig. 4, the method may further include:

s207: and acquiring preset weight information.

In a specific embodiment, the preset weight information may include a preset scaling parameter and/or a normalized nonlinear parameter, where the preset scaling parameter is used to constrain a weight of the preset delivery parameter in the recall identification process, and specifically, the preset scaling parameter may be a constant between 0 and 1, and may be set in combination with an actual application requirement. The normalized nonlinear parameter may be used to increase nonlinearity for a preset delivery parameter; specifically, the normalized nonlinear parameter can be obtained by normalizing the nonlinear characteristic of any non-preset delivery parameter. Correspondingly, the preset training data may further include sample weight information (including preset scaling parameters and/or corresponding normalized nonlinear parameters).

In an alternative embodiment, the non-linearity characteristics corresponding to the non-delivery parameter attributes may be combined to generate a normalized non-linearity parameter, and accordingly, as shown in fig. 3, the normalized non-linearity parameter is obtained in the following manner:

s301: acquiring a first resource characteristic, wherein the first resource characteristic is a characteristic of a non-throwing parameter attribute;

s303: performing nonlinear transformation on the first resource characteristics to obtain second resource characteristics;

s305: and carrying out normalization processing on the second resource characteristics to obtain normalized nonlinear parameters.

In a specific embodiment, the first resource characteristic is a characteristic of a non-delivery parameter attribute; optionally, coding the non-delivery parameter attribute to obtain the first resource characteristic; optionally, the feature extraction layer in the resource recall model may extract semantic information and simultaneously implement nonlinear processing on corresponding features, and correspondingly, the second resource features may also be extracted from the semantic information output by the feature extraction layer.

In an alternative embodiment, the second resource feature may be normalized in combination with the activation function sigmoid.

In the embodiment, the non-linear characteristics corresponding to the non-delivery parameter attribute of the resource to be recalled are generated, and the non-linear normalized non-linear parameters are added for the preset delivery parameters, so that the characterization of the resource to be recalled can be better performed on the basis of improving the recall performance of the resource recall model.

Further, in the step S203, inputting the first object resource information and the preset release parameter into the resource recall model to perform recall identification, and obtaining the first recall index data may include:

inputting the first object resource characteristics, preset weight information and preset release parameters into a resource recall model to carry out recall identification, so as to obtain first recall index data;

correspondingly, the recall index identification function can take the sum of the object resource characteristics and the constraint delivery parameters as an independent variable, so that the first recall index data and the preset delivery parameters are positively correlated; specifically, the constraint delivery parameter may be a product of preset weight information and a preset delivery parameter.

In a specific embodiment, taking the resource recall model as an example obtained by training a semantic vector recall model based on a double-tower structure, the recall index identification function may be:

score＝sigmoid(user·ad)；

wherein score may be first recall indicator data, sigmoid () may be an activation function mapping variables between 0, 1; the user may be a semantic feature vector corresponding to the target object, and the ad may be a semantic feature vector corresponding to the resource to be recalled.

Specifically, in order to meet the positive correlation between the first recall index data and the preset release parameter, firstly, splitting the user ad into a feature vector of the preset release parameter and a feature vector of a non-preset release parameter; specifically, assuming that the vector dimension is d, n dimensions (0 < n < d) can be independently obtained to represent the preset delivery parameters, and the remaining d-n dimensions are used for representing the feature vectors of the non-preset delivery parameters. Taking n=1 as an example, it can be expressed specifically as: ad= [ admain, bid ]; correspondingly, since the user does not contain the feature vector of the preset delivery parameter, the user= [ uselmain, 1]; wherein, admain can be resource semantic features (semantic vectors corresponding to non-head delivery parameter attributes), bid is a preset delivery parameter; uselmin may be an object semantic feature.

Further, in order to restrict the weight of the preset delivery parameter in the recall identification process and increase the nonlinearity for the preset delivery parameter, the above ad= [ admain, sigmoid (F) ×scale×bid ], where sigmoid (F) is a normalized nonlinearity parameter; scale is a preset scaling parameter;

correspondingly, score=sigmoid (user_ad)

＝sigmoid(usermain·admain+sigmoid(F)*scale*bid)

Wherein, uselmain·admain can be the object resource feature corresponding to the first object resource information.

In the above embodiment, in the recall identification process, the preset scaling parameter for constraining the weight of the preset release parameter in the recall identification process and/or the normalized nonlinear parameter for adding nonlinearity to the preset release parameter are introduced, so that the fitting capability of the resource recall model to recall index data can be improved on the basis of effectively controlling the influence of the resource release parameter in the recall identification process, and the recall performance of the resource recall model is further improved.

S205: and screening out target recall resources corresponding to the target object from the resources to be recalled based on the first recall index data.

In a specific embodiment, the resources to be recalled include a plurality of release resources, and correspondingly, the first recall index data includes recall index data corresponding to each of the plurality of release resources, and the recall index data corresponding to any release resource can represent a degree of association between the target object and the release resource. Specifically, the larger the recall index data corresponding to any of the released resources, the higher the association degree between the released resources and the target object, and correspondingly, the higher the probability that the released resources are recalled.

In an optional embodiment, the screening, based on the first recall index data, the target recall resource corresponding to the target object from the to-be-recalled resources may include: taking the released resources with the corresponding recall index data larger than or equal to the preset index threshold value in the resources to be recalled as target recall resources; alternatively, the multiple released resources in the resources to be recalled may be sorted in descending order according to the corresponding recall index data, and the released resources sorted in the preset number of bits are used as target recall resources. Specifically, the index threshold and the preset number are set in combination with the actual application requirement.

According to the technical scheme provided by the embodiment of the specification, in the process of inputting the first object resource information and the preset release parameter of the resource to be recalled into the resource recall model for recall identification, the independent variable of the recall index identification function which is not linearly and positively related in the resource recall model is set to be the sum of the object resource characteristics corresponding to the first object resource information and the preset release parameter, the obtained first recall index data and the preset release parameter are positively related on the basis of the learning object and the attribute characteristics of the resource to be recalled, the relevance between the recalled resource and the release (release parameter) of the resource release party is realized, the sensitivity of the recall index data to the release parameter is improved, and the target recall resource is screened from the resource to be recalled based on the first recall index data which is positively related to the preset release parameter, so that the rationality and the effectiveness of the resource recall process can be greatly improved, the addition release enthusiasm of the resource release party can be effectively improved, and the release and diversity of the release resource in the release system can be enriched.

In an optional embodiment, in order to verify recall performance of the resource recall model (including a model of a recall indicator identification function that makes the first recall indicator data positively correlated with the preset delivery parameters) according to an embodiment of the present application, the method may further include: the step of generating recall direction analysis results for reflecting the corresponding recall performance of the resource recall model, as shown in fig. 5, may specifically include:

s501: acquiring a first initial release parameter of a first sample resource, a first adjustment release parameter of the first sample resource and second object resource information corresponding to the first sample resource;

s503: inputting the second object resource information and the first initial input parameter into a resource recall model for recall identification to obtain second recall index data;

s505: inputting the second object resource information and the first adjustment release parameter into a resource recall model for recall identification to obtain third recall index data;

s507: and carrying out change direction consistency analysis based on the first initial delivery parameter, the first adjustment delivery parameter, the second recall index data and the third recall index data to obtain a recall direction analysis result.

In a specific embodiment, the first sample resource may be a sample resource for performing recall performance analysis, and specifically, the first initial delivery parameter may be an initial delivery parameter of the first sample resource, and the first adjusted delivery parameter may be a delivery parameter of the first sample resource after adjustment; specifically, the first initial delivery parameter is different from the first adjustment delivery parameter, and optionally, on the basis of the first initial delivery parameter, the adjustment of the delivery parameter is performed to obtain the first adjustment delivery parameter; the first adjustment delivery parameter can also be obtained by performing the reduction treatment of the delivery parameter on the basis of the first initial delivery parameter. The second object resource information characterizes the non-delivery parameter attribute of the first sample resource and the object attribute of the corresponding sample object of the first sample resource. For specific details of the second object resource information, reference may be made to the first object resource information described above, and details thereof are not repeated herein.

In a specific embodiment, the specific refinement of S503 and S505 may refer to the above-mentioned inputting the first object resource information and the preset delivery parameter into the resource recall model for recall identification, so as to obtain the relevant refinement of the first recall index data, which is not described herein. Optionally, in the case that the resource recall model takes the second object resource information and the first initial input parameter as input, the recall index identification function takes the sum of the object resource feature corresponding to the second object resource information and the first initial input parameter as an independent variable, so that the second recall index data and the first initial input parameter are positively correlated; optionally, under the condition that the resource recall model takes the second object resource information, the first initial input parameter and the preset weight information as input, the recall index identification function takes the sum of the object resource characteristics corresponding to the second object resource information and the first sample input parameter as an independent variable, so that the second recall index data and the first initial input parameter are positively correlated; the first sample delivery parameter is the product of the preset weight information and the first initial delivery parameter. Optionally, in the case that the resource recall model takes the third object resource information and the first adjustment release parameter as input, the recall index identification function takes the sum of the object resource feature corresponding to the third object resource information and the first adjustment release parameter as an independent variable, so that the second recall index data and the first adjustment release parameter are positively correlated; optionally, in the case that the resource recall model takes the third object resource information, the first adjustment release parameter and the preset weight information as input, the recall index identification function takes the sum of the object resource feature corresponding to the third object resource information and the second sample release parameter as an independent variable, so that the second recall index data and the first adjustment release parameter are positively correlated; the second sample delivery parameter is the product of the preset weight information and the first adjustment delivery parameter.

In a specific embodiment, the recall direction analysis result may be used to indicate a degree of consistency between a recall indicator change direction identified by the resource recall model and a corresponding delivery parameter change direction. Specifically, the better the consistency degree between the recall index change direction identified by the resource recall model and the corresponding release parameter change direction is, the better the recall performance of the resource recall model is correspondingly.

In an optional embodiment, the first sample resource may include a plurality of sample resources, and optionally, as shown in fig. 6, performing the direction-of-change consistency analysis based on the first initial delivery parameter, the first adjustment delivery parameter, the second recall indicator data, and the third recall indicator data, and obtaining the recall direction analysis result may include:

s5071: determining the change direction of the throwing parameter corresponding to each sample resource according to the first initial throwing parameter corresponding to each sample resource and the first adjusting throwing parameter corresponding to each sample resource;

s5073: determining the recall index change direction corresponding to each sample resource according to the second recall index data corresponding to each sample resource and the third recall index data corresponding to each sample resource;

S5075: carrying out change direction consistency comparison on the recall index change direction corresponding to each sample resource and the release parameter change direction corresponding to each sample resource to obtain first consistency indication information corresponding to each sample resource;

s5077: and generating recall direction analysis results according to the first consistency indication information corresponding to the plurality of sample resources.

In a specific embodiment, the change direction of the delivery parameter corresponding to any sample resource may represent the change condition of the first adjusted delivery parameter corresponding to the sample resource relative to the first initial delivery parameter corresponding to the sample resource; optionally, if the first adjustment delivery parameter is greater than the first initial delivery parameter, the delivery parameter may change in direction to be greater; if the first adjusted delivery parameter is smaller than the first initial delivery parameter, the delivery parameter change direction may be smaller.

In a specific embodiment, the recall index change direction corresponding to any sample resource may represent a change condition of the third recall index data corresponding to the sample resource relative to the second recall index data corresponding to the sample resource; optionally, if the third recall index data is greater than the second recall index data, the recall index change direction may be greater; if the third recall index data is smaller than the second recall index data, the recall index change direction may be smaller.

In a specific embodiment, the first consistency indication information corresponding to any sample resource may represent whether the recall indicator change direction corresponding to the sample resource is consistent with the release parameter change direction corresponding to the sample resource; optionally, if the recall index change direction and the release parameter change direction corresponding to a certain sample resource are both larger or smaller, the first consistency indication information corresponding to the sample resource may be the change direction consistency; otherwise, the sample resource corresponding to the first consistency indication information may be inconsistent in the changing direction.

In an optional embodiment, generating the recall direction analysis result according to the first consistency indication information corresponding to the plurality of sample resources may include:

taking the ratio between the number of the second sample resources and the number of the plurality of sample resources as a recall direction analysis result;

or alternatively, the first and second heat exchangers may be,

and taking the ratio between the number of the third sample resources and the number of the plurality of sample resources as a recall direction analysis result.

In a specific embodiment, the second sample resource may be a sample resource whose change direction is inconsistent with the corresponding first consistency indication information in the plurality of sample resources; the third sample resource may be a sample resource with a consistent change direction of the corresponding first consistency indication information in the plurality of sample resources.

In a specific embodiment, when the recall direction analysis result is the ratio between the number of the second sample resources and the number of the plurality of sample resources, the smaller the recall direction analysis result is, the better the consistency degree between the recall index change direction identified by the resource recall model and the corresponding release parameter change direction is, and accordingly, the better the recall performance of the resource recall model is.

In a specific embodiment, when the recall direction analysis result is the ratio between the number of the third sample resources and the number of the plurality of sample resources, the larger the recall direction analysis result is, the better the consistency degree between the recall index change direction identified by the resource recall model and the corresponding release parameter change direction is, and accordingly, the better the recall performance of the resource recall model is. Optionally, if the recall performance of the resource recall model is determined to be insufficient by combining with the recall direction analysis result, the resource recall model can be continuously adjusted and trained until the recall performance requirement is met.

In a specific embodiment, assuming that the recall direction analysis result is a ratio between the number of the second sample resources and the number of the plurality of sample resources, and combining training data of the delivery system a, the delivery system b and the delivery system c, recall performance analysis is performed on the resource recall model and the existing resource recall model (the resource recall model using a dot product between a resource semantic feature and an object semantic feature as an argument of a recall index identification function), as shown in table 1, the resource recall model of the present application has a great recall performance improvement compared with the existing resource recall model.

TABLE 1

In a specific embodiment, in combination with the embodiment of the recall direction analysis result described above as a ratio between the number of second sample resources and the number of multiple sample resources, as shown in fig. 7, fig. 7 is a schematic diagram of a recall direction analysis result provided according to an exemplary embodiment. Wherein, the abscissa represents recall index data, and the ordinate represents recall direction analysis results; as can be seen from fig. 7, among a large amount of training data, samples with poor consistency (larger recall direction analysis result) are mainly samples with recall index data in the vicinity of 0 and 1, while other samples (middle part) with poor consistency are fewer, so that the optimization objective of the application is met, that is, the change space of two extremes (recall index data approach 0 and 1 respectively) is limited; an original approach to 0 is probably due to poor competition, and even if the throwing parameter is greatly improved, the throwing parameter is difficult to win (namely recall index data is difficult to improve); the competition of the device is close to 1, so that the device is strong, and the throwing parameters do not need to be improved; correspondingly, the sensitivity of the middle part is ensured, the competitive power can be effectively improved by improving the throwing parameters, and then the throwing parameters can be better promoted by attracting the resource throwing party.

In the above embodiment, the recall performance of the resource recall model can be effectively represented by combining the release parameters before and after the adjustment of the sample resources and the recall index data determined by the resource recall model based on the release parameters before and after the adjustment, and generating the recall direction analysis result reflecting the consistency degree between the change direction of the release parameters before and after the adjustment and the change direction of the recall index identified by the resource recall model, thereby greatly improving the evaluation accuracy and effectiveness of the recall performance of the resource recall model.

In practical application, if the delivery parameters are improved, under the condition that other conditions are unchanged, how much should the recall index data identified by the resource recall model be improved is reasonable? This problem is illustrated by way of example. In general, the game industry is much higher than the local life industry in terms of the parameters of resources. When both raise the same delivery parameters at the same time, how much should the recall index data identified by the resource recall model be raised respectively? For this problem, the present application can perform amplitude consistency analysis in combination with recommendation index data (e.g., ECPM) corresponding to resource ranking (fine ranking recommendation). Optionally, the method further comprises: the step of generating the amplitude analysis result for reflecting the recall performance corresponding to the recall model of the resource, as shown in fig. 8, may specifically include:

S801: acquiring sample release parameters, first recommendation index data, second recommendation index data, third recommendation index data, fourth recommendation index data, third object resource information corresponding to fourth sample resources in each sample resource pair and fourth object resource information corresponding to fifth sample resources in each sample resource pair corresponding to each sample resource pair;

s803: inputting the third object resource information and the second initial release parameter into a resource recall model to carry out recall identification, so as to obtain fourth recall index data;

s805: inputting the third object resource information and the second adjustment input parameters into a resource recall model for recall identification to obtain fifth recall index data;

s807: inputting the fourth object resource information and the third initial input parameter into a resource recall model to carry out recall identification, so as to obtain sixth recall index data;

s809: inputting the fourth object resource information and the third adjustment input parameter into a resource recall model for recall identification to obtain seventh recall index data;

s811: and carrying out amplitude consistency analysis based on the fourth recall index data, the fifth recall index data, the sixth recall index data, the seventh recall index data, the first recommendation index data, the second recommendation index data, the third recommendation index data and the fourth recommendation index data to obtain an amplitude analysis result.

In a particular embodiment, the plurality of sample resource pairs may be sample resource pairs for performing recall performance analysis, each sample resource pair comprising two different sample resources. Specifically, the sample delivery parameters corresponding to each sample resource pair may include a second initial delivery parameter of a fourth sample resource in the sample resource pair, a third initial delivery parameter of a fifth sample resource in the sample resource pair, a second adjusted delivery parameter of the fourth sample resource, and a third adjusted delivery parameter of the fifth sample resource.

In a specific embodiment, the second initial delivery parameter may be an initial delivery parameter of the fourth sample resource, and the second adjusted delivery parameter may be a delivery parameter of the fourth sample resource after adjustment; specifically, the second initial delivery parameter is different from the second adjustment delivery parameter, and optionally, on the basis of the second initial delivery parameter, the adjustment of the delivery parameter is performed to obtain the second adjustment delivery parameter; the adjustment of the delivery parameters can also be performed on the basis of the second initial delivery parameters to obtain second adjusted delivery parameters. The third object resource information characterizes the non-delivery parameter attribute of the fourth sample resource and the object attribute of the corresponding sample object of the fourth sample resource. For specific details of the third object resource information, reference may be made to the first object resource information described above, and details thereof are not repeated herein.

In a specific embodiment, the third initial delivery parameter may be an initial delivery parameter of the fifth sample resource, and the third adjusted delivery parameter may be a delivery parameter of the fifth sample resource after adjustment; specifically, the third initial delivery parameter is different from the third adjustment delivery parameter, and optionally, on the basis of the third initial delivery parameter, the step-up processing of the delivery parameter can be performed to obtain the third adjustment delivery parameter; the adjustment of the delivery parameters can also be performed on the basis of the third initial delivery parameters to obtain third adjusted delivery parameters. The fourth object resource information characterizes the non-delivery parameter attribute of the fifth sample resource and the object attribute of the corresponding sample object of the fifth sample resource. For details of the fourth object resource information, reference may be made to the first object resource information described above, and details thereof are not repeated here.

In a specific embodiment, the first recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second initial throwing parameter, the second recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second adjusted throwing parameter, the third recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third initial throwing parameter, and the fourth recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third adjusted throwing parameter.

In a specific embodiment, the steps S803, S805, S807 and S809 may refer to the steps of inputting the first object resource information and the preset delivery parameter into the resource recall model to perform recall identification, so as to obtain relevant refinement of the first recall index data, which is not described herein.

In a specific embodiment, the amplitude analysis result may be used to indicate a degree of consistency between the recall indicator change amplitude identified by the resource recall model and the corresponding recommended indicator change amplitude.

In an alternative embodiment, as shown in fig. 9, the performing the amplitude consistency analysis based on the fourth recall indicator data, the fifth recall indicator data, the sixth recall indicator data, the seventh recall indicator data, the first recommendation indicator data, the second recommendation indicator data, the third recommendation indicator data, and the fourth recommendation indicator data, to obtain the amplitude analysis result includes:

s8131: according to the first recall comparison result corresponding to each sample resource pair and the second recall comparison result corresponding to each sample resource pair, carrying out consistency analysis on the comparison result to obtain second consistency indication information corresponding to each sample resource pair;

s8133: according to the first recommendation comparison result corresponding to each sample resource pair and the second recommendation comparison result corresponding to each sample resource pair, carrying out comparison result consistency analysis to obtain third consistency indication information corresponding to each sample resource pair;

S8135: and generating an amplitude analysis result according to the second consistency indication information corresponding to the plurality of sample resource pairs and the third consistency indication information corresponding to the plurality of sample resource pairs.

In a specific embodiment, the first recall comparison result may represent a size relationship of the sixth recall index data corresponding to each sample resource pair relative to the fourth recall index data corresponding to each sample resource pair; optionally, if the sixth recall index data corresponding to a certain sample resource pair is greater than or equal to the fourth recall index data corresponding to the sample resource pair, and accordingly, the first recall comparison result may be greater than or equal to the fourth recall index data; conversely, the first recall comparison result may be less than;

in a specific embodiment, the second recall comparison result may represent a size relationship of the seventh recall index data corresponding to each sample resource pair relative to the fifth recall index data corresponding to each sample resource pair; optionally, if the seventh recall index data corresponding to a certain sample resource pair is greater than or equal to the fifth recall index data corresponding to each sample resource pair, the corresponding second recall comparison result may be greater than or equal to the fifth recall index data; conversely, the second recall comparison result may be less than.

In a specific embodiment, the second consistency indication information corresponding to any sample resource pair may indicate whether the first recall comparison result corresponding to the sample resource pair and the second recall comparison result corresponding to each sample resource pair are consistent. Specifically, if the first recall comparison result corresponding to a certain sample resource pair and the second recall comparison result corresponding to each sample resource pair are both greater than, or are both equal to, or are both less than, and accordingly, the second consistency indication information corresponding to the sample resource pair may be the consistency of the comparison results; otherwise, the sample resource pair corresponding to the second consistency indicating information may be inconsistent as a result of the comparison.

In a specific embodiment, the first recommendation comparison result characterizes a size relationship of the third recommendation index data corresponding to each sample resource pair relative to the first recommendation index data corresponding to each sample resource pair; optionally, if the third recommendation index data corresponding to a certain sample resource pair is greater than or equal to the first recommendation index data corresponding to each sample resource pair, and accordingly, the first recommendation comparison result may be greater than or equal to the first recommendation index data; conversely, the first recommendation comparison result may be less than.

In a specific embodiment, the second recommendation comparison result characterizes a size relationship of the fourth recommendation index data corresponding to each sample resource pair relative to the second recommendation index data corresponding to each sample resource pair; optionally, if the fourth recommendation index data corresponding to a certain sample resource pair is greater than or equal to the two recommendation index data corresponding to each sample resource pair, and correspondingly, the second recommendation comparison result may be greater than or equal to the second recommendation index data; conversely, the second recommended comparison result may be less than.

In a specific embodiment, the third consistency indication information corresponding to any sample resource pair may indicate whether the first recommended comparison result corresponding to the sample resource pair and the second recommended comparison result corresponding to each sample resource pair are consistent. Specifically, if the first recommended comparison result corresponding to a certain sample resource pair and the second recommended comparison result corresponding to each sample resource pair are both greater than or both equal to or both less than, and accordingly, the third consistency indication information corresponding to the sample resource pair may be the consistency of the comparison results; otherwise, the sample resource pair corresponding to the third consistency indicating information may be inconsistent as a result of the comparison.

In an optional embodiment, generating the amplitude analysis result according to the second consistency indication information corresponding to the plurality of sample resource pairs and the third consistency indication information corresponding to the plurality of sample resource pairs may include:

taking the ratio between the number of the first sample resource pairs and the number of the plurality of sample resource pairs as an amplitude analysis result;

or alternatively, the first and second heat exchangers may be,

and taking the ratio between the number of the second sample resource pairs and the number of the plurality of sample resource pairs as an amplitude analysis result.

In a specific embodiment, the first pair of sample resources includes a pair of sample resources for which the second correspondence indication information and the third correspondence indication information are inconsistent as a result of the comparison, and a pair of sample resources for which the second correspondence indication information and the third correspondence indication information are inconsistent as a result of the comparison.

In a specific embodiment, the second consistency indication information is inconsistent as a result of the comparison, and the corresponding pair of sample resources whose third consistency indication information is inconsistent as a result of the comparison may be a first type of amplitude inconsistent sample; the second consistency indication information is consistent with the comparison result, and the corresponding third consistency indication information is inconsistent with the comparison result, and the sample resource pair can be a second class of inconsistent-amplitude samples.

In a specific embodiment, the fourth sample resource in a certain sample resource pair is a, the fifth sample resource is B, the first recommendation index data is ECPM (a), the second recommendation index data is ECPM '(a), the third recommendation index data is ECPM (B), the fourth recommendation index data is ECPM' (B), the score (a) is fourth recall index data, and the score (B) is fifth recall index data; score '(a) is sixth recall indicator data, and score' (B) is seventh recall indicator data; assuming ECPM (A) < ECPM (B) and score (A) < score (B), the first recommended comparison result may be greater than and the first recall comparison result is greater than; after the delivery parameters are adjusted, ECPM '(A) is less than ECPM' (B), but score '(A) is more than or equal to score' (B), namely the second recommendation comparison result can be more than, and the second recall comparison result is less than or equal to; correspondingly, the third consistency indication information is consistent with the comparison result, the second consistency indication information is inconsistent with the comparison result, and the sample resource pair formed by the fourth sample resource A and the fifth sample resource B is a first-class amplitude inconsistent sample.

In another specific embodiment, the fourth sample resource in a certain sample resource pair is C, the fifth sample resource is D, the first recommendation index data is ECPM (C), the second recommendation index data is ECPM '(C), the third recommendation index data is ECPM (D), the fourth recommendation index data is ECPM' (D), the score (C) is fourth recall index data, and the score (D) is fifth recall index data; score '(C) is sixth recall indicator data, and score' (D) is seventh recall indicator data; assuming ECPM (C) < ECPM (D) and score (C) < score (D), the first recommended comparison result may be greater than and the first recall comparison result is greater than; after the delivery parameters are adjusted, ECPM '(C) is not less than ECPM' (D), but score '(C) is less than core' (D), namely the second recommendation comparison result can be less than or equal to the second recommendation comparison result, and the second recall comparison result is less than the first recall comparison result; the corresponding third consistency indicating information is inconsistent in comparison result, the second consistency indicating information is inconsistent in comparison result, and the corresponding sample resource pair consisting of the fourth sample resource C and the fifth sample resource D is a second class of amplitude inconsistent sample.

In a specific embodiment, the second pair of sample resources may include a pair of sample resources whose corresponding second coherence indication information is consistent with the comparison result and whose corresponding third coherence indication information is consistent with the comparison result.

In a specific embodiment, when the amplitude analysis result is the ratio between the number of the first sample resource pairs and the number of the plurality of sample resource pairs, the smaller the amplitude analysis result is, the better the consistency degree between the recall index change amplitude identified by the resource recall model and the corresponding recommended index change amplitude is, and accordingly, the better the recall performance of the resource recall model is.

In a specific embodiment, when the amplitude analysis result is the ratio between the number of the second sample resource pairs and the number of the plurality of sample resource pairs, the larger the amplitude analysis result is, the better the consistency degree between the recall index change amplitude identified by the resource recall model and the corresponding recommended index change amplitude is, and accordingly, the better the recall performance of the resource recall model is.

Optionally, if the recall performance of the resource recall model is determined to be insufficient by combining the amplitude analysis result, the resource recall model can be continuously adjusted and trained until the recall performance requirement is met.

In the embodiment, the recall performance of the resource recall model is represented by combining the amplitude analysis result which can represent the consistency degree between the recall index change amplitude identified by the resource recall model and the corresponding recommended index change amplitude, so that the evaluation accuracy and the effectiveness performance of the recall performance of the resource recall model can be greatly improved.

The embodiment of the application also provides a data processing device, as shown in fig. 10, where the device includes:

a first information obtaining module 1010 configured to perform obtaining first object resource information and preset release parameters of a resource to be recalled, where the first object resource information characterizes an object attribute of a target object and a non-release parameter attribute of the resource to be recalled;

a first recall identification module 1020 configured to perform recall identification of the first object resource information and the preset release parameter input into a resource recall model to obtain first recall index data, wherein a recall index identification function in the recall index identification layer is a nonlinear positive correlation function, and the recall index identification function uses a sum of object resource characteristics corresponding to the first object resource information and the preset release parameter as an independent variable, so that the first recall index data and the preset release parameter are positively correlated;

And a target recall resource screening module 1030 configured to perform screening, based on the first recall index data, the target recall resource corresponding to the target object from the resources to be recalled.

In an alternative embodiment, the apparatus further comprises:

the device comprises a preset weight information acquisition module, a preset parameter identification module and a parameter identification module, wherein the preset weight information acquisition module is configured to execute acquisition of preset weight information, the preset weight information comprises preset scaling parameters and/or normalized nonlinear parameters, the preset scaling parameters are used for restraining weights of the preset delivery parameters in a recall identification process, and the normalized nonlinear parameters are used for adding nonlinearities to the preset delivery parameters;

the first recall identification module 1020 is specifically configured to perform inputting the first object resource feature, the preset weight information and the preset delivery parameter into the resource recall model for recall identification, so as to obtain the first recall index data;

the recall index identification function takes the sum of the object resource characteristics and constraint delivery parameters as an independent variable, so that the first recall index data and the preset delivery parameters are positively correlated; the constraint delivery parameters are products of the preset weight information and the preset delivery parameters.

In an optional embodiment, the preset weight information acquiring module includes:

a first resource feature acquisition unit configured to perform acquisition of a first resource feature, the first resource feature being a feature of the non-delivery parameter attribute;

the nonlinear transformation unit is configured to perform nonlinear transformation on the first resource characteristics to obtain second resource characteristics;

and the normalization processing unit is configured to perform normalization processing on the second resource characteristics to obtain the normalized nonlinear parameters.

In an alternative embodiment, the apparatus further comprises:

the second information acquisition module is configured to acquire first initial delivery parameters of first sample resources, first adjustment delivery parameters of the first sample resources and second object resource information corresponding to the first sample resources;

the second recall identification module is configured to input the second object resource information and the first initial delivery parameter into the resource recall model for recall identification, so as to obtain second recall index data;

the third recall identification module is configured to input the second object resource information and the first adjustment delivery parameter into the resource recall model for recall identification, so as to obtain third recall index data;

The change direction consistency analysis module is configured to perform change direction consistency analysis based on the first initial delivery parameter, the first adjustment delivery parameter, the second recall index data and the third recall index data to obtain a recall direction analysis result, and the recall direction analysis result is used for indicating the consistency degree between the recall index change direction identified by the resource recall model and the corresponding delivery parameter change direction.

In an alternative embodiment, the change direction consistency analysis module includes:

a delivery parameter change direction determining unit configured to perform the first initial delivery parameter corresponding to each sample resource and the first adjustment delivery parameter corresponding to each sample resource, and determine a delivery parameter change direction corresponding to each sample resource;

a recall index change direction determination unit configured to perform determining a recall index change direction corresponding to each sample resource according to the second recall index data corresponding to each sample resource and the third recall index data corresponding to each sample resource;

the change direction consistency comparison unit is configured to perform change direction consistency comparison on the recall index change direction corresponding to each sample resource and the delivery parameter change direction corresponding to each sample resource, so as to obtain first consistency indication information corresponding to each sample resource;

And a recall direction analysis result generation unit configured to perform generation of the recall direction analysis result according to the first consistency indication information corresponding to the plurality of sample resources.

In an alternative embodiment, the recall direction analysis result generation unit includes:

a first recall direction analysis result determination unit configured to perform, as the recall direction analysis result, a ratio between a number of second sample resources and a number of the plurality of sample resources, the second sample resources being sample resources in which corresponding first coincidence indication information in the plurality of sample resources is inconsistent in a change direction;

or alternatively, the first and second heat exchangers may be,

and the second recall direction analysis result determining unit is configured to execute a ratio between the number of third sample resources and the number of the plurality of sample resources as the recall direction analysis result, wherein the third sample resources are sample resources with consistent change directions according to the corresponding first consistency indication information in the plurality of sample resources.

In an alternative embodiment, the apparatus further comprises:

a third information obtaining module configured to obtain sample input parameters, first recommendation index data, second recommendation index data, third recommendation index data, fourth recommendation index data, third object resource information corresponding to fourth sample resources in each sample resource pair, and fourth object resource information corresponding to fifth sample resources in each sample resource pair corresponding to each sample resource pair; the sample delivery parameters comprise a second initial delivery parameter of the fourth sample resource, a third initial delivery parameter of the fifth sample resource, a second adjusted delivery parameter of the fourth sample resource and a third adjusted delivery parameter of the fifth sample resource; the first recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second initial throwing parameter, the second recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second adjusting throwing parameter, the third recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third initial throwing parameter, and the fourth recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third adjusting throwing parameter;

The fourth recall identification module is configured to input the third object resource information and the second initial delivery parameter into the resource recall model for recall identification, so as to obtain fourth recall index data;

the fifth recall identification module is configured to input the third object resource information and the second adjustment delivery parameter into the resource recall model for recall identification, so as to obtain fifth recall index data;

the sixth recall identification module is configured to input the fourth object resource information and the third initial delivery parameter into the resource recall model for recall identification, so as to obtain sixth recall index data;

the seventh recall identification module is configured to input the fourth object resource information and the third adjustment delivery parameter into the resource recall model for recall identification, so as to obtain seventh recall index data;

and the amplitude consistency analysis module is configured to perform amplitude consistency analysis based on the fourth recall index data, the fifth recall index data, the sixth recall index data, the seventh recall index data, the first recommendation index data, the second recommendation index data, the third recommendation index data and the fourth recommendation index data to obtain an amplitude analysis result, wherein the amplitude analysis result is used for indicating the consistency degree between the recall index change amplitude identified by the resource recall model and the corresponding recommendation index change amplitude.

In an alternative embodiment, the amplitude consistency analysis module includes:

the first consistency analysis unit is configured to perform consistency analysis of the comparison result according to the first recall comparison result corresponding to each sample resource pair and the second recall comparison result corresponding to each sample resource pair, and obtain second consistency indication information corresponding to each sample resource pair; the first recall comparison result represents the size relation of the sixth recall index data corresponding to each sample resource pair relative to the fourth recall index data corresponding to each sample resource pair; the second recall comparison result represents the size relation of the seventh recall index data corresponding to each sample resource pair relative to the fifth recall index data corresponding to each sample resource pair;

the second consistency analysis unit is configured to perform consistency analysis of the comparison result according to the first recommendation comparison result corresponding to each sample resource pair and the second recommendation comparison result corresponding to each sample resource pair, and obtain third consistency indication information corresponding to each sample resource pair; the first recommendation comparison result represents the size relation of the third recommendation index data corresponding to each sample resource pair relative to the first recommendation index data corresponding to each sample resource pair; the second recommendation comparison result represents the size relation of the fourth recommendation index data corresponding to each sample resource pair relative to the second recommendation index data corresponding to each sample resource pair;

And an amplitude analysis result generation unit configured to perform generation of the amplitude analysis result from the second coincidence indication information corresponding to the plurality of sample resource pairs and the third coincidence indication information corresponding to the plurality of sample resource pairs.

In an alternative embodiment, the amplitude analysis result generation unit includes:

a first amplitude analysis result determination unit configured to perform, as the amplitude analysis result, a ratio between the number of first sample resource pairs including the sample resource pairs for which the second coincidence indication information is inconsistent and the corresponding third coincidence indication information is consistent, and the sample resource pairs for which the second coincidence indication information is consistent and the corresponding third coincidence indication information is inconsistent, as the comparison result;

or alternatively, the first and second heat exchangers may be,

and a second amplitude analysis result determination unit configured to perform, as the amplitude analysis result, a ratio between the number of second sample resource pairs including the sample resource pairs for which the second coincidence indication information is coincident as a comparison result and the number of the plurality of sample resource pairs for which the corresponding third coincidence indication information is coincident as a comparison result.

The specific manner in which the various modules perform the operations in the apparatus of the above embodiments have been described in detail in connection with the embodiments of the method, and will not be described in detail herein.

Fig. 11 is a block diagram of an electronic device for data processing, which may be a terminal, according to an embodiment of the present application, and an internal structure diagram thereof may be as shown in fig. 11. The electronic device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the electronic device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a data processing method. The display screen of the electronic equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the electronic equipment, and can also be an external keyboard, a touch pad or a mouse and the like.

Fig. 12 is a block diagram of another electronic device for data processing, which may be a server, according to an exemplary embodiment, and an internal structure diagram thereof may be as shown in fig. 12. The electronic device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the electronic device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a data processing method.

It will be appreciated by those skilled in the art that the structures shown in fig. 11 or 12 are merely block diagrams of partial structures related to the present disclosure and do not constitute limitations of the electronic device to which the present disclosure is applied, and that a particular electronic device may include more or fewer components than shown in the drawings, or may combine certain components, or have different arrangements of components.

In an exemplary embodiment, there is also provided an electronic device including: a processor; a memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement a data processing method as in the embodiments of the present disclosure.

In an exemplary embodiment, a computer-readable storage medium is also provided, which when executed by a processor of an electronic device, enables the electronic device to perform the page display method in the embodiments of the present disclosure.

In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions to cause the computer device to perform the data processing methods provided in the various alternative implementations described above.

It will be appreciated that in the specific embodiment of the present application, related data such as historical operation information of a user is involved, when the above embodiments of the present application are applied to specific products or technologies, user permission or consent is required to be obtained, and the collection, use and processing of related data is required to comply with related laws and regulations and standards of related countries and regions.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.

Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

It is to be understood that the present disclosure is not limited to the precise arrangements and instrumentalities shown in the drawings, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method of data processing, the method comprising:

2. The method according to claim 1, wherein the method further comprises:

acquiring preset weight information, wherein the preset weight information comprises preset scaling parameters and/or normalized nonlinear parameters, the preset scaling parameters are used for restraining weights of the preset delivery parameters in a recall identification process, and the normalized nonlinear parameters are used for adding nonlinearities to the preset delivery parameters;

inputting the first object resource information and the preset release parameters into a resource recall model for recall identification, wherein obtaining first recall index data comprises the following steps:

inputting the first object resource characteristics, the preset weight information and the preset delivery parameters into the resource recall model for recall identification, and obtaining the first recall index data;

3. The method of claim 2, wherein the normalized nonlinear parameter is obtained by:

acquiring a first resource characteristic, wherein the first resource characteristic is the characteristic of the non-delivery parameter attribute;

performing nonlinear transformation on the first resource characteristics to obtain second resource characteristics;

and carrying out normalization processing on the second resource characteristics to obtain the normalized nonlinear parameters.

4. A method according to any one of claims 1 to 3, wherein the method further comprises:

acquiring a first initial delivery parameter of a first sample resource, a first adjustment delivery parameter of the first sample resource and second object resource information corresponding to the first sample resource _；

Inputting the second object resource information and the first initial input parameter into the resource recall model for recall identification to obtain second recall index data;

inputting the second object resource information and the first adjustment delivery parameters into the resource recall model for recall identification to obtain third recall index data;

and carrying out change direction consistency analysis based on the first initial delivery parameter, the first adjustment delivery parameter, the second recall index data and the third recall index data to obtain a recall direction analysis result, wherein the recall direction analysis result is used for indicating the consistency degree between the recall index change direction identified by the resource recall model and the corresponding delivery parameter change direction.

5. The method of claim 4, wherein the first sample resources comprise a plurality of sample resources, and wherein performing a direction of change consistency analysis based on the first starting delivery parameter, the first adjusted delivery parameter, the second recall indicator data, and the third recall indicator data, obtaining recall direction analysis results comprises:

determining a change direction of the delivery parameter corresponding to each sample resource according to the first initial delivery parameter corresponding to each sample resource and the first adjustment delivery parameter corresponding to each sample resource;

determining the recall index change direction corresponding to each sample resource according to the second recall index data corresponding to each sample resource and the third recall index data corresponding to each sample resource;

carrying out change direction consistency comparison on the recall index change direction corresponding to each sample resource and the delivery parameter change direction corresponding to each sample resource to obtain first consistency indication information corresponding to each sample resource;

and generating the recall direction analysis result according to the first consistency indication information corresponding to the plurality of sample resources.

6. The method of claim 5, wherein generating the recall direction analysis result from the first consistency indication information corresponding to the plurality of sample resources comprises:

taking the ratio between the number of second sample resources and the number of the plurality of sample resources as the recall direction analysis result, wherein the second sample resources are sample resources with inconsistent change directions corresponding to the first consistency indication information in the plurality of sample resources;

or alternatively, the first and second heat exchangers may be,

and taking the ratio of the number of third sample resources to the number of the plurality of sample resources as the recall direction analysis result, wherein the third sample resources are sample resources with consistent change directions corresponding to the first consistency indication information in the plurality of sample resources.

7. A method according to any one of claims 1 to 3, wherein the method further comprises:

acquiring sample release parameters, first recommendation index data, second recommendation index data, third recommendation index data, fourth recommendation index data, third object resource information corresponding to fourth sample resources in each sample resource pair and fourth object resource information corresponding to fifth sample resources in each sample resource pair corresponding to each sample resource pair; the sample delivery parameters comprise a second initial delivery parameter of the fourth sample resource, a third initial delivery parameter of the fifth sample resource, a second adjusted delivery parameter of the fourth sample resource and a third adjusted delivery parameter of the fifth sample resource; the first recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second initial throwing parameter, the second recommended index data is recommended index data obtained by throwing the fourth sample resource based on the second adjusting throwing parameter, the third recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third initial throwing parameter, and the fourth recommended index data is recommended index data obtained by throwing the fifth sample resource based on the third adjusting throwing parameter;

Inputting the third object resource information and the second initial delivery parameter into the resource recall model for recall identification to obtain fourth recall index data;

inputting the third object resource information and the second adjustment delivery parameters into the resource recall model for recall identification to obtain fifth recall index data;

inputting the fourth object resource information and the third initial delivery parameter into the resource recall model to carry out recall identification, so as to obtain sixth recall index data;

inputting the fourth object resource information and the third adjustment delivery parameter into the resource recall model for recall identification to obtain seventh recall index data;

and carrying out amplitude consistency analysis based on the fourth recall index data, the fifth recall index data, the sixth recall index data, the seventh recall index data, the first recommendation index data, the second recommendation index data, the third recommendation index data and the fourth recommendation index data to obtain an amplitude analysis result, wherein the amplitude analysis result is used for indicating the consistency degree between the recall index change amplitude identified by the resource recall model and the corresponding recommendation index change amplitude.

8. The method of claim 7, wherein the performing a magnitude consistency analysis based on the fourth recall indicator data, the fifth recall indicator data, the sixth recall indicator data, the seventh recall indicator data, the first recommendation indicator data, the second recommendation indicator data, the third recommendation indicator data, and the fourth recommendation indicator data, the obtaining a magnitude analysis result comprises:

according to the first recall comparison result corresponding to each sample resource pair and the second recall comparison result corresponding to each sample resource pair, carrying out comparison result consistency analysis to obtain second consistency indication information corresponding to each sample resource pair; the first recall comparison result represents the size relation of the sixth recall index data corresponding to each sample resource pair relative to the fourth recall index data corresponding to each sample resource pair; the second recall comparison result represents the size relation of the seventh recall index data corresponding to each sample resource pair relative to the fifth recall index data corresponding to each sample resource pair;

according to the first recommended comparison result corresponding to each sample resource pair and the second recommended comparison result corresponding to each sample resource pair, carrying out comparison result consistency analysis to obtain third consistency indication information corresponding to each sample resource pair; the first recommendation comparison result represents the size relation of the third recommendation index data corresponding to each sample resource pair relative to the first recommendation index data corresponding to each sample resource pair; the second recommendation comparison result represents the size relation of the fourth recommendation index data corresponding to each sample resource pair relative to the second recommendation index data corresponding to each sample resource pair;

And generating the amplitude analysis result according to the second consistency indication information corresponding to the plurality of sample resource pairs and the third consistency indication information corresponding to the plurality of sample resource pairs.

9. The method of claim 8, wherein generating the amplitude analysis result from the second coherence indication information corresponding to the plurality of pairs of sample resources and the third coherence indication information corresponding to the plurality of pairs of sample resources comprises:

taking the ratio between the number of the first sample resource pairs and the number of the plurality of sample resource pairs as the amplitude analysis result, wherein the first sample resource pairs comprise sample resource pairs with inconsistent comparison results and inconsistent comparison results corresponding to the second consistency indication information, and sample resource pairs with consistent comparison results corresponding to the third consistency indication information, and sample resource pairs with inconsistent comparison results corresponding to the second consistency indication information and inconsistent comparison results corresponding to the third consistency indication information;

or alternatively, the first and second heat exchangers may be,

and taking the ratio of the number of second sample resource pairs to the number of the plurality of sample resource pairs as the amplitude analysis result, wherein the second sample resource pairs comprise sample resource pairs with the same comparison result as the second consistency indication information corresponding to the plurality of sample resource pairs, and the corresponding third consistency indication information is sample resource pairs with the same comparison result.

10. A data processing apparatus, the apparatus comprising:

11. An electronic device, comprising:

a processor;

A memory for storing the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the data processing method of any of claims 1 to 9.

12. A computer readable storage medium, characterized in that instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the data processing method of any one of claims 1 to 9.

13. A computer program product comprising computer instructions, characterized in that the above computer instructions are stored in a computer readable storage medium, from which computer readable storage medium a processor of a computer device reads the above computer instructions, which computer instructions are executed by a processor to cause the computer device to perform the data processing method according to any of claims 1 to 9.