CN114463590A

CN114463590A - Information processing method, apparatus, device, storage medium, and program product

Info

Publication number: CN114463590A
Application number: CN202210128368.0A
Authority: CN
Inventors: 钟子宏
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-02-11
Filing date: 2022-02-11
Publication date: 2022-05-10

Abstract

The application relates to an information processing method, an information processing device, information processing equipment, a storage medium and a program product, and relates to the technical field of Internet application. The method comprises the following steps: acquiring object characteristics and information characteristics of each object in a target period; obtaining an object classification label of each object in each historical period in the first period group; predicting the classification result of each object based on the object characteristic information characteristics and the object classification labels; the target information is processed based on the classification result of each object. The scheme can be applied to a vehicle-mounted scene, and the accuracy of information processing can be improved by combining the object classification labels respectively corresponding to the objects in a plurality of historical periods.

Description

Information processing method, apparatus, device, storage medium, and program product

Technical Field

The embodiments of the present application relate to the field of internet application technologies, and in particular, to an information processing method, apparatus, device, storage medium, and program product.

Background

With the continuous development of big data technology and Artificial Intelligence (AI) technology, current information processing systems can perform personalized information processing for each object.

In the related art, for given information, the recommendation system may input the characteristics of each object and the characteristics of information to be processed to the AI model for prediction to predict a target object suitable for the information from each object, and then process the information corresponding to the target object.

However, when the AI model in the above scheme is used for prediction, the information based on the AI model is relatively single, which results in lower accuracy of model prediction and further affects accuracy of information processing.

Disclosure of Invention

The embodiment of the application provides an information processing method, an information processing device, information processing equipment, a storage medium and a program product, which can improve the accuracy of information processing. The technical scheme is as follows:

in one aspect, an information processing method is provided, and the method includes:

acquiring object characteristics of each object in a target period and information characteristics of target information in the target period;

obtaining an object classification label of each object in each history period in a first period group, wherein the first period group comprises m history periods before the target period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

predicting a classification result of each object based on an object feature of each object in a target period, an information feature of the target information in the target period, and an object classification label of each object in each historical period in the first period group, wherein the classification result is used for indicating the probability that the object generates the target action on the target information in the target period;

and executing first processing on the target information corresponding to a target object in the objects in the target period based on the classification result of the objects.

acquiring an object feature sample of each object in a first history period and an information feature sample of target information in the first history period;

obtaining an object classification label of each object in each history period in a second period group, wherein the second period group comprises m history periods before the first history period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

acquiring a second state transition matrix based on the object classification label of each object in each historical period in the second period group; the second state transition matrix is used for indicating the change trend of the probability of generating the target action on the target information in the second periodic group of each object;

training to obtain an object classification model based on the object features of the objects in the first history period, the information features of the target information in the first history period and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability that the object generates the target action on the target information in a target period.

In another aspect, there is provided an information processing apparatus, the apparatus including:

the first characteristic acquisition module is used for acquiring the object characteristics of each object in a target period and the information characteristics of target information in the target period;

a first tag obtaining module, configured to obtain an object classification tag of each object in each history period in a first period group, where the first period group includes m history periods before the target period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

a prediction module, configured to predict a classification result of each object based on an object feature of each object in a target period, an information feature of the target information in the target period, and an object classification label of each object in each historical period in the first period group, where the classification result is used to indicate a probability that the object produces the target action on the target information in the target period;

and the processing module is used for executing first processing on the target information corresponding to the target object in each object in the target period based on the classification result of each object.

In one possible implementation, the prediction module is configured to,

acquiring a first state transition matrix based on the object classification label of each object in each historical period in the first period group; the first state transition matrix is used for indicating the change trend of the probability of the target action generated on the target information by each object in the first period group;

and predicting the classification result of each object based on the object characteristics of each object in a target period, the information characteristics of the target information in the target period and the first state transition matrix.

In one possible implementation, the prediction module is configured to,

obtaining an unbiased estimation of the state transition probability of each of the m history periods in the first period group based on the object classification label of each object in each history period in the first period group;

and constructing the first state transition matrix based on unbiased estimation of the state transition probability of each of the m history cycles in the first cycle group.

In one possible implementation, the prediction module is configured to,

according to the object classification labels of the objects in a first object history period, acquiring a first probability that the object information receives the object action in the first object history period and a second probability that the object information does not receive the object action in the first object history period; the first target history cycle is any one of the history cycles in the first group of cycles;

and acquiring the first probability and the second probability as unbiased estimation of the state transition probability of the first target history period.

In one possible implementation, the prediction module is configured to,

and constructing a matrix according to the sequence from the rear to the front based on the unbiased estimation of the state transition probability of each of the m historical periods in the first period group to obtain the first state transition matrix.

In one possible implementation, the prediction module is configured to,

inputting the object characteristics of each object in a target period, the information characteristics of the target information in the target period and the first state transition matrix into an object classification model, and obtaining the classification result of each object output by the object classification model;

the object classification model is a model obtained by training based on object features of each object in a first history period, information features of the target information in the first history period and a second state transition matrix; the second state transition matrix is obtained based on the object classification labels of the objects in each history period in the second period group; the second period group includes m history periods prior to the first history period.

In one possible implementation, the first history period is a period of the first period group that is the latest in time.

the second characteristic acquisition module is used for acquiring an object characteristic sample of each object in a first history period and an information characteristic sample of target information in the first history period;

a second tag obtaining module, configured to obtain an object classification tag of each object in each history cycle in a second cycle group, where the second cycle group includes m history cycles before the first history cycle; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

a matrix obtaining module, configured to obtain a second state transition matrix based on the object classification label of each object in each history period in the second period group; the second state transition matrix is used for indicating the change trend of the probability of generating the target action on the target information in the second periodic group of each object;

the model training module is used for training and obtaining an object classification model based on the object features of the objects in the first history period, the information features of the target information in the first history period and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability that the object generates the target action on the target information in a target period.

In one possible implementation manner, the matrix obtaining module is configured to,

acquiring unbiased estimates of state transition probabilities of m historical periods in the second periodic group based on the object classification labels of the objects in each historical period in the second periodic group;

and constructing the second state transition matrix based on the unbiased estimation of the state transition probability of each of the m historical periods in the second period group.

according to the object classification labels of the objects in a second object history period, acquiring a third probability that the object information receives the object action in the second object history period and a fourth probability that the object information does not receive the object action in the second object history period; the second target historical period is any one of the historical periods in the second group of periods;

and acquiring the third probability and the fourth probability as unbiased estimation of the state transition probability of the second target history period.

In another aspect, a computer device is provided, which comprises a processor and a memory, wherein the memory stores at least one computer program, and the at least one computer program is loaded and executed by the processor to realize the information processing method.

In another aspect, a computer-readable storage medium is provided, in which at least one computer program is stored, the computer program being loaded and executed by a processor to implement the information processing method described above.

In another aspect, a computer program product or computer program is 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 the processor executes the computer instructions, so that the computer device executes the information processing method provided in the above-mentioned various alternative implementations.

The technical scheme provided by the application can comprise the following beneficial effects:

the computer device may comprehensively predict a target object matching the target information among the objects in combination with the object characteristics of the objects in the target period, the information characteristics of the target information in the target period, and the object classification tags respectively corresponding to the objects in the plurality of history periods, and perform the first processing on the target information corresponding to the target object. In the scheme, when the computer device determines an object matched with the target information, in addition to combining the object characteristics and the information characteristics, the situation that each object performs the target action on the target object in a plurality of historical periods is combined, and the situation that each object performs the target action on the target object in the plurality of historical periods can reflect the change situation of the probability that the target information receives the target action in the plurality of historical periods as a whole, and the change situation of the probability that the target information receives the target action in the plurality of historical periods can influence the probability that each object generates the target action on the target information in the subsequent period; therefore, the above-described scheme can improve the accuracy of information processing in the case of combining the object classification tags respectively corresponding to the respective objects in a plurality of history cycles.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

FIG. 1 is a schematic diagram illustrating an information processing process according to an exemplary embodiment of the present application;

FIG. 2 is a block diagram illustrating an information handling system in accordance with an exemplary embodiment;

FIG. 3 is a flow diagram illustrating an information processing method according to an exemplary embodiment;

FIG. 4 is a flow diagram illustrating an information processing method according to an exemplary embodiment;

FIG. 5 is a diagram illustrating an information handling framework in accordance with an illustrative embodiment;

FIG. 6 is a flow diagram illustrating an information processing method according to an example embodiment;

FIG. 7 is a block diagram of model training and information processing involved in the embodiment shown in FIG. 6;

FIG. 8 shows a block diagram of an information processing apparatus shown in an exemplary embodiment of the present application;

FIG. 9 shows a block diagram of an information processing apparatus shown in an exemplary embodiment of the present application;

FIG. 10 is a block diagram illustrating the structure of a computer device according to an example embodiment.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

In the information processing method provided by the embodiment of the application, in a scene where an object, the part of which matches with the designated information, is selected from a plurality of objects, in addition to the characteristics of each object and the characteristics of the information, the change condition of the probability that the information receives the target behavior in a plurality of history periods is also considered.

Fig. 1 is a schematic diagram illustrating an information processing procedure according to an exemplary embodiment of the present application, and as shown in fig. 1, in the process of processing target information, a computer device may obtain at least the following three types of information: the object characteristics 110 of each object, the information characteristics 120 of the target information, and the variation 130 of the probability that the target information receives the target behavior over multiple history periods.

Then, the computer device predicts the probability of generating the target behavior for the target information in each subsequent period of each object by combining the object characteristics 110, the information characteristics 120 and the change situation 130 of the probability of receiving the target behavior by the target information in a plurality of historical periods.

Then, the computer device may perform a first process on the target information corresponding to a part of the objects in a subsequent period based on the probability that each object generates a target behavior for the target information.

For example, the first process may include, but is not limited to, transmitting the target information or the associated information of the target information to the terminal of the target object, and the like.

The terminal includes, but is not limited to, a mobile phone, a computer, an intelligent voice interaction device, an intelligent household appliance, a vehicle-mounted terminal, and the like.

FIG. 2 is a block diagram illustrating an information handling system 200 according to an exemplary embodiment, information handling system 200 including: a server 220 and several terminals 240.

The server 220 includes at least one of a server, a plurality of servers, a cloud computing platform, and a virtualization center. The server 220 is used for providing background services for the terminal 240; in this embodiment, the server 220 may be configured to perform an information processing procedure, for example, send the obtained target information to the terminal 240.

The terminal 240 may be a terminal device having an information receiving function, for example, the terminal 240 may be a mobile phone, a tablet computer, an e-book reader, smart glasses, a smart watch, an MP3 player (Moving Picture Experts Group Audio Layer III, motion Picture Experts Group Audio Layer 3), an MP4 player (Moving Picture Experts Group Audio Layer IV, motion Picture Experts Group Audio Layer 4), a laptop computer, a desktop computer, and the like. Among them, the terminal 240 may include an application having an information receiving and presenting function for receiving and presenting information. Optionally, the application may be an application that needs to be downloaded and installed, or may be an application that is to be used at any time, which is not limited in this embodiment of the application.

The terminal 240 is connected to the server 220 through a communication network. Optionally, the communication network is a wired network or a wireless network.

Optionally, the wireless network or wired network described above uses standard communication techniques and/or protocols. The Network is typically the Internet, but may be any Network including, but not limited to, a Local Area Network (LAN), a Metropolitan Area Network (MAN), a Wide Area Network (WAN), a mobile, wireline or wireless Network, a private Network, or any combination of virtual private networks. In some embodiments, data exchanged over a network is represented using techniques and/or formats including HyperText Mark-up Language (HTML), Extensible Mark-up Language (XML), and the like. All or some of the links may also be encrypted using conventional encryption techniques such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec). In other embodiments, custom and/or dedicated data communication techniques may also be used in place of, or in addition to, the data communication techniques described above.

For the sake of understanding, the technical terms related to the various embodiments of the present application will be described first:

and (3) state transition: in the field of machine learning,the state of this stage is often the result of the state of the previous stage and the decision of the previous stage. If the state S of the K-th stage is given_kAnd decision u_k(S_k) Then state S of the K +1 stage_k+1And can be determined.

First order state transition: the state of the present stage is determined by the state of the previous stage and the decision result of the previous stage, i.e. the state S of K +1_k+1State S of phase K_kAnd decision u_k(S_k) Influence.

Second-order state transition: the state of the K +1 stage is determined by the K, K-1 stage state and the result of the decision, i.e., state S of K +1_k+1State S from stage K, K-1_k、S_k-1And decision u_k(S_k)、u_k-1(S_k-1) Influence.

Bayes second-order state transition: state S of K +1_k+1State S from stage K, K-1_k、S_k-1And decision u_k(S_k)、u_k-1(S_k-1) Influence, and can be expressed as the following expression:

u_k+1(S_k+1|S_k,S_k-1)＝α·u_k(S_k)+β·u_k-1(S_k-1)。

m-order state transition: the state of the K +1 stage is determined by K, K-1, …, the K-m stage state and the result of the decision, i.e., state S of K +1_k+1State S from stage K, K-1_k、S_k-1、…、S_k-mAnd decision u_k(S_k)、u_k-1(S_k-1)、…、u_k-m(S_k-m) Influence.

Bayes m-order state transition: state S of K +1_k+1State S subjected to K, K-1, …, K-m stages_k、S_k-1、…、S_k-mAnd decision u_k(S_k)、u_k-1(S_k-1)、…、u_k-m(S_k-m) Influence, and can be expressed as the following expression:

u_k+1(S_k+1|S_k,S_k-1,…,S_k-m)＝A_1×(m+1)U(S)_(m+1)×1；

wherein A is_1×(m+1)＝(α₁,α₂,…,α_m+1) Representing a 1-row m + 1-column coefficient matrix;

U(S)_(m+1)×1＝(u_k(S_k),u_k-1(S_k-1),…,u_k-m(S_k-m))^Trepresenting a decision matrix of m +1 rows and 1 column.

Sigmoid function: one class is defined as a function of the form:

a two-classification LR (Logistic Regression) algorithm: and (3) mapping the continuous output value of the uncertain range of the linear regression into the range of (0, 1) by introducing a Sigmoid function into the linear regression model, and converting the linear regression model into a probability prediction model.

Gradient descent method of second order state transition: namely gradient descent under the influence of second-order state transition probability in the gradient descent solving process.

Gradient descent method of m-order state transition: namely gradient descent under the influence of m-order state transition probability in the gradient descent solving process.

Unbiased estimation: the mathematical expectation of the estimator is equal to the true value of the estimated parameter.

And (4) estimating the CTR: CTR (Click-Through-Rate) is a commonly used term in the field of internet information (such as advertisements), and generally refers to the Click arrival Rate of a web advertisement (picture advertisement/text advertisement/keyword advertisement/ranking advertisement/video advertisement, etc.), i.e., the actual number of clicks of the advertisement (which may be the number of reached target pages) is divided by the advertisement presentation amount (Show content).

Object characteristics: and (4) behavior records of the objects in the business and data extraction of the objects. The method comprises the following steps: clicking, collecting, paying amount, paying times, active duration, active days and the like of the object in the service.

The commodity characteristics are as follows: attributes of goods (such as target information in this application) and their data refinement. The method comprises the following steps: click rate of goods, payment rate, collection, average payment amount (total payment amount per money/number of paid people), average active duration (total active time per money/number of active people), etc.

And (3) downloading a label: and the object clicks the label corresponding to the downloading behavior of the target information on the activity page. For example, the flag is 1 when the download target information is clicked, and the flag is 0 when the download target information is clicked and not downloaded.

Fig. 3 is a flowchart illustrating an information processing method according to an exemplary embodiment, which may be performed by a computer device, which may be implemented as an information processing device, which may be illustratively implemented as the server 220 shown in fig. 1. As shown in fig. 3, the information processing method may include the steps of:

step 310, acquiring the object characteristics of each object in the target period and the information characteristics of the target information in the target period.

The target period may be a period in which the current time is located, or may be a next period or a plurality of periods after the current time.

The object features in the target period may be features of the respective objects acquired at the start time of the target period. Alternatively, the object feature in the target period may be a feature of each object acquired after completion of a period immediately preceding the target period. Alternatively, the feature of the object in the target period may be a feature of each object acquired before the target period is completed.

Accordingly, the information characteristic in the target period may be a characteristic of the target information acquired at the start time of the target period. Alternatively, the information characteristic in the target period may be a characteristic of the target information acquired after completion of a period immediately preceding the target period. Alternatively, the information characteristic in the target period may be a characteristic of the target information acquired before the target period is completed.

In the embodiment of the application, the computer device can periodically process the target information.

The time lengths of the periods in which the computer device processes the target information may be the same, or the periods in which the computer device processes the target information may be different.

Alternatively, the periods of processing the target information by the computer device may be continuous or discontinuous.

Alternatively, the start-stop time of each cycle of processing the target information by the computer device may be specified by an administrator of the information processing system.

The object may be a terminal used by a user. For example, the objects may be terminals respectively registered with different user accounts. In a general case, each of the objects may be identified by a corresponding user account.

Step 320, obtaining an object classification label of each object in each historical period in a first period group, wherein the first period group comprises m historical periods before a target period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2.

In this embodiment, the object classification tag may be a two-class tag, which is used to indicate that the object generates the target action on the target information, or indicate that the object does not generate the target action on the target information.

Wherein, for an object, the object corresponds to an object classification label in each history period.

For example, if the object a does not generate the target motion for the target information in the history cycle 1 and generates the target motion for the target information in the history cycle 1, the object classification tag of the object a in the history cycle 1 is different from the object classification tag of the object a in the history cycle 2.

The target action may be an action performed based on a user operation, such as performing detail display, downloading, forwarding, and the like on the target information.

And step 330, predicting a classification result of each object based on the object characteristics of each object in the target period, the information characteristics of the target information in the target period and the object classification label of each object in each historical period in the first period group, wherein the classification result is used for indicating the probability that the object generates the target action on the target information in the target period.

The history cycle may refer to a cycle that has been completed in time.

In this embodiment, the classification result may be a single probability value, and the probability value may be used to indicate a probability that a certain object performs a target action on the target information. Alternatively, the classification result may be a probability distribution of two classes, for example, the probability distribution may indicate the probability that a certain object performs a target action on the target information and the probability that the target action is not performed.

In one possible implementation, the computer device may obtain a first state transition matrix based on the object classification label of each object in each history period in the first period group; the first state transition matrix is used for indicating the change trend of the probability of generating target action on the target information in the first period group by each object; the computer device predicts the classification result of each object based on the object characteristics of each object in the target period, the information characteristics of the target information in the target period, and the first state transition matrix.

In the embodiment of the application, the computer device may determine, according to the object classification labels of the objects in m history periods, a variation trend of the probability that the objects generate the target actions on the target information in the m history periods (that is, a variation trend of the probability that the target information receives the target actions in a plurality of history periods) and represent the variation trend in a matrix form, and then predict by combining the object characteristics, the information characteristics, and the matrix.

Step 340, based on the classification result of each object, in the target period, corresponding to the target object in each object, executing a first process on the target information.

To sum up, in the information processing method provided in the embodiment of the present application, the computer device may comprehensively predict a target object matching the target information in each object by combining the object feature of each object in the target period, the information feature of the target information in the target period, and the object classification label corresponding to each object in a plurality of history periods, and perform the first processing on the target information corresponding to the target object. In the scheme, when the computer device determines an object matched with the target information, in addition to combining the object characteristics and the information characteristics, the situation that each object performs the target action on the target object in a plurality of historical periods is combined, and the situation that each object performs the target action on the target object in the plurality of historical periods can reflect the change situation of the probability that the target information receives the target action in the plurality of historical periods as a whole, and the change situation of the probability that the target information receives the target action in the plurality of historical periods can influence the probability that each object generates the target action on the target information in the subsequent period; therefore, the above-described scheme can improve the accuracy of information processing in the case of combining the object classification tags respectively corresponding to the respective objects in a plurality of history cycles.

The solution in the embodiment shown in fig. 3 described above in the present application may be implemented based on AI, for example, the step of predicting the classification result of each object based on the object features, the information features, and the object classification labels described above may be performed by an object classification model trained based on AI technology.

AI is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the implementation method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making. The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.

Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

FIG. 4 is a flow diagram illustrating an information processing method that may be performed by a computer device, which may be implemented as a model training device, according to an example embodiment. As shown in fig. 4, the information processing method may include the steps of:

step 410, obtaining an object feature sample of each object after the first history period and an information feature sample of the target information after the first history period.

Step 420, obtaining an object classification label of each object in each history period in a second period group, wherein the second period group comprises m history periods before the first history period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2.

Step 430, acquiring a second state transition matrix based on the object classification label of each object in each history period in the second period group; the second state transition matrix is used for indicating the change trend of the probability of generating the target action on the target information in the second periodic group of the objects.

Step 440, training to obtain an object classification model based on the object characteristics of each object after the first history period, the information characteristics of the target information after the first history period, and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability of the object generating a target action on target information in a target period.

In this embodiment of the present application, the sample labeling information used in the model training process may be object classification labels of the respective objects in the first history period.

In summary, in the information processing method provided in the embodiment of the present application, in the model training process, the computer device combines the object features of each object after the first history period, the information features of the target information after the first history period, and the object classification labels respectively corresponding to each object in a plurality of history periods before the first history period, and trains to obtain the object classification model for predicting the target object matching with the target information in each object. Correspondingly, in the scheme, when determining the object matched with the target information in the subsequent prediction process of the computer device, in addition to combining the object characteristics and the information characteristics, the method also combines the situations that each object performs the target action on the target object in a plurality of historical periods, and the situations that each object performs the target action on the target object in the plurality of historical periods can reflect the change situation of the probability that the target information receives the target action in the plurality of historical periods as a whole, and the change situation of the probability that the target information receives the target action in the plurality of historical periods can influence the probability that each object generates the target action on the target information in the subsequent period; therefore, the above-described scheme can improve the accuracy of information processing in the case of combining the object classification tags respectively corresponding to the respective objects in a plurality of history cycles.

Based on the embodiments shown in fig. 3 and fig. 4, taking the example that the first state transition matrix and the second state transition matrix are implemented as the m-order state matrix, the first history cycle is the kth cycle (hereinafter, abbreviated as kth cycle), and the target cycle is the K +1 th cycle, please refer to fig. 5, which is an information processing framework diagram according to an exemplary embodiment.

As shown in fig. 5, in the model training stage, the model training device performs model training through the object feature sample 51 in the K-th stage, the information feature sample 52 in the K-th stage, the object classification label 53 of each object in the K-th stage, and the m-order state matrix 54 in the K-1 to K-m stages to obtain an object classification model.

In the prediction stage, the information processing device uses an object classification model to process the object characteristics 55 of the K +1 th stage, the information characteristics 56 of the K +1 th stage, and the m-order state matrix 57 of the K to K-m +1 th stages, and performs classification prediction to obtain a classification result. The information processing device then performs information processing according to the classification result, for example, selects a target object according to the classification result, and sends the target information or related information of the target information to a terminal corresponding to the target object.

The model training device and the information processing device may be implemented as the same physical computer device, or the model training device and the information processing device may be implemented as different physical computer devices.

In a possible application scenario, the scheme shown in each embodiment of the application can be applied to a CTR prediction scenario of trip service fueling coupon downloading.

When a business scenario for downloading the fueling coupon is constructed, the click behavior of each object in the trip service in the K-1 th period and the downloading behavior of each object to the coupon in the K-1 th period can be used for cross matching, if an object generates the click behavior in the trip service page in the K-1 th period but generates the downloading behavior of the coupon in the K-1 th period, the object is marked to download the coupon in the K-th period (for example, the object classification label of the object in the K-1 th period is marked as 1), otherwise, the object is marked not to download the coupon in the K-th period (for example, the object classification label of the object in the K-th period is marked as 0). When a sample used for training is constructed and input is predicted, an object classification label (a download/non-download label) and characteristics of the K-1 stage object in the travel service and information characteristics of each function point in the travel service small program are used for constructing a training sample and prediction input data in a coupon download business scene.

In the scheme shown in the embodiment of the present application, the additionally added m-order state matrix may be a bayesian m-order state transition matrix. When the method is applied to a fueling coupon downloading service scene, because the preferential fueling activity time is longer and the interval time of preferential operation activities is also longer, a scene of continuous three-period operation activities with a short interval time can be considered, namely, the value m is equal to 3.

For example, the object classification tag of the kth stage may be a coupon download business scenario tag (i.e., download/no download tag); the object characteristics of each period can comprise characteristic data of clicking behaviors of the object users of each period on the travel service small program page, and the information characteristics of each period can be characteristic data of clicking rate, coupon downloading rate, exposure, clicking amount and the like of each function point of the travel service of each period.

The scheme shown in each embodiment of the present application may include the following six stages: the method comprises a sample data set preprocessing stage, a Bayesian m-order state transition probability calculation stage, a model training stage under m-order state transition, a model test evaluation stage under m-order state transition, a model prediction stage under m-order state transition and a classification downloading recommendation stage.

FIG. 6 is a flow diagram illustrating an information processing method that may be performed by a computer device that may include a model training device and an information processing device, according to an example embodiment. As shown in fig. 6, the information processing method may include the steps of:

step 601, the model training device obtains object feature samples of each object in a first history period and information feature samples of target information in the first history period.

In the sample data set preprocessing stage, the model training device can preprocess the sample data set to obtain training and testing samples: and constructing sample data by using the object characteristics, the information characteristics and the object classification labels in the K-th period, and distinguishing the whole sample data (training samples and test samples) into sparse characteristics and dense characteristics. The sparse features may be subjected to onehot processing, and the dense features may be subjected to Principal Component Analysis (PCA) decorrelation processing, normalization (normalization) processing, feature discretization processing, and the like. The model training device randomly divides the processed sparse features, dense features and object classification labels into training samples (ratio a) and test samples (ratio 1-a) according to a certain proportion, for example, the samples can be randomly divided into the training samples according to general experience: the test sample is 8:2 (i.e., training sample and test sample are randomly sliced at an 8:2 ratio).

Here, onehot processing refers to using an N-bit status register to encode N states, each state being represented by its own independent register bit and only one bit being active at any time. For example: sex characteristics: the onehot codes for [ "male", "female" ] correspond to: male- > 10; female- > 01. The Chinese characteristics are as follows: the onehot codes for [ "china", "usa" france "] correspond to: china- > 100; U.S. > 010; france- > 001; the motion characteristics are as follows: the onehot codes corresponding to [ "football", "basketball", "badminton", "table tennis" ] are: football- > 1000; basketball- > 0100; badminton- > 0010; Ping-Pong-ball- > 0001. When a sample is [ "male", "china", "table tennis" ], the complete feature digitized onehot results are: [1,0,1,0,0,0,0,0,1].

The PCA described above is a widely used data dimension reduction algorithm. The principal idea of PCA is to map n-dimensional features onto k-dimensions, where k-dimensions are completely new orthogonal features, also called principal components, and k-dimensional features reconstructed on the basis of the original n-dimensional features.

Wherein, the object characteristics of the K stage can include: basic attribute data such as gender, age, region, and the like; active attribute data such as active days, active duration, active function quantity, interval of registration time and current time days and the like; recharging attribute data such as recharging amount, consumption amount, recharging times, recharging days, interval between the first recharging and the current time days and the like; function clicks, gift bag/type of gift coupon (quantity, number, value), type of gift bag/gift coupon used (quantity, value), type of gift bag/gift coupon expired (quantity, value), etc.

In step 602, the model training apparatus obtains an object classification label of each object in each historical period in the second period group.

Wherein the second period group comprises m history periods before the first history period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2.

In the embodiment of the present application, taking a coupon downloading service scenario as an example, the object classification label (label) of each period may be 1 or 0. For example, 1 represents a click and download (positive sample), and 0 represents a click and not download (negative sample).

Step 603, the model training device obtains a second state transition matrix based on the object classification label of each object in each history period in the second period group.

The second state transition matrix is used for indicating the change trend of the probability of generating the target action on the target information in the second periodic group of each object.

In one possible implementation manner, the obtaining the second state transition matrix based on the object classification label of each object in each history period in the second period group includes:

acquiring unbiased estimation of the state transition probability of each of the m historical periods in the second periodic group based on the object classification label of each object in each historical period in the second periodic group;

and constructing a second state transition matrix based on the unbiased estimation of the state transition probability of each of the m historical periods in the second period group.

In one possible implementation manner, obtaining an unbiased estimation of the state transition probability of each of the m history cycles in the second period group based on the object classification label of each object in each history cycle in the second period group includes:

according to the object classification labels of the objects in the second object history period, acquiring a third probability that the object information receives the object action in the second object history period and a fourth probability that the object information does not receive the object action in the second object history period; the second target history period is any one history period in the second period group;

In the bayesian m-order state transition probability calculation stage, the model training device may determine the m-order state transition probability of the training and testing stages:

object Classification Label label (Y) Using stage K-1_k-1) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As first order state transition probability (p)_0,k-1，1-p_0,k-1) Unbiased estimation of (2); object Classification Label label (Y) Using stage K-2_k-2) Calculating the ratio of positive and negative samples (negative sample ratio of

Positive sample fraction of

) As second order state transition probability (p)_0,k-2，1-p_0,k-2) Unbiased estimation of (2); similarly, the object classification label (Y) of low K-m phase is utilized_k-m) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As m-order state transition probability (p)_0,k-m，1-p_0,k-m) OfAnd (6) estimating deviation. Thereby constructing an unbiased estimation matrix of m-order state transition probability:

for example, taking m-3 as an example in a coupon download business scenario, in the training and testing phase, three periods of classification labels Y of K-3, K-2 and K-1 are used_k-3、Y_k-2、Y_k-1Respectively calculating the proportion of positive samples and negative samples of the three periods of K-3, K-2 and K-1: k-3 order negative sample ratio of

A positive sample fraction of

The K-2 order negative sample ratio is

A positive sample fraction of

The K-1 order negative sample ratio is

A positive sample fraction of

Step 604, training the model training device to obtain an object classification model based on the object features of the objects in the first history period, the information features of the target information in the first history period, and the second state transition matrix.

The object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability that the object generates a target action on target information in a target period.

In a possible implementation manner, when performing model training, the model training device may first obtain an object classification model W by using object features of each object in a first history period and information features of target information in the first history period, and substitute the object classification model W into the second state transition matrix to perform gradient descent solution in the training process.

The labeling data used in the training process of the object classification model W may be object classification labels of the respective objects in the first history period.

For example, the model training device may input the object features of each object in the first history period and the information features of the target information in the first history period to the object classification model, substitute the object features into the second state transition matrix to obtain the predicted object classification prediction result of each object in the first history period, and update the parameters of the object classification model according to the difference between the object classification prediction result of each object in the first history period and the object classification label of each object in the first history period.

In an embodiment of the present application, the model training phase includes a model training sub-phase and a model test evaluation sub-phase. In the sub-stage of model training, the model training device can substitute the Bayesian second-order state transition probability model M1 derived by mathematics based on training samples into the K-2, K-1, … and K-M order state transition probabilities p_0,k-1，1-p_0,k-1，p_0,k-2，1-p_0,k-2，…，p_0,k-m，1-p_0,k-mAnd obtaining a model weight (the model weight measures the contribution of the characteristic X to Y) by a gradient descent method under the m-order state transition (namely, gradient descent under the influence of the m-order state transition probability in the gradient descent solving process), and obtaining an object classification model W.

Mathematical studies have demonstrated that the state outcome at stage K has an explicit effect on the predicted outcome at stage K + 1.

Wherein the probability model (M1) predicted as a positive sample is:

the probability model predicted as a negative sample (M2) is:

wherein, Y_t-1The probability of taking two states of 0 or 1 for e {0,1} is p₀，1-p₀。

A model weight representing a t period;

a feature representing the t period; y is_tA class label representing the t period. w is a_YRepresenting the tag weight coefficient.

In the model test evaluation phase, for the trained model W, the model training device can use the test sample to perform test and substitute the test sample into K-2, K-1, … and K-m order state transition probability p_0,k-1，1-p_0,k-1，p_0,k-2，1-p_0,k-2，…，p_0,k-m，1-p_0,k-mAnd calculating classification probability and evaluation indexes (indexes such as recall ratio, precision ratio and Area Under the Curve (AUC)) Under the test sample through a formula M3, and if the evaluation indexes achieve the evaluation effect, storing the model W. If the model evaluation is not passed, the training test steps are repeated until the model reaches the evaluation index.

The embodiment of the application provides a Bayesian M-order state transition-based binary algorithm calculation model, namely a probability model (M3) with positive prediction:

wherein, the stage state of K-j (j is 1, 2, …, m) is represented as Y_k-jIs in the field of {0,1}, and the probability of taking 0 or 1 as two states is p respectively_0,k-j，1-p_0,k-j；W_XRepresenting a characteristic XModel weight, X_kCharacteristic of information representing the kth period, W_YRepresents the tag weight coefficient, W_Y ^k-jIndicates that the label Y is marked when the k-j period is reached_k-j＝i_jThen, the kth-j tag coefficients; wherein the content of the first and second substances,

after the model training device trains the object classification model, the object classification model can be deployed to the information processing device.

In step 605, the information processing apparatus acquires the object characteristics of each object in the target period and the information characteristics of the target information in the target period.

In the sample data set preprocessing stage, the information processing device may also construct input data used for prediction. For example, the information processing apparatus may construct input data used for prediction using the object feature and the information feature of the K +1 th stage and the object classification tags of the K-th to K-m +1 th stages, and distinguish the input data used for prediction into a sparse-type feature and a dense-type feature. Wherein, the sparse type characteristic is subjected to onehot processing, and the dense type characteristic is subjected to PCA decorrelation processing, normalization (standardization) processing, characteristic discretization processing and the like.

In step 606, the information processing apparatus obtains an object classification label for each object in each history period in the first period group.

Wherein the first period group comprises m history periods before the target period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2.

In one possible implementation, the first history period is the latest in time period in the first period group.

That is to say, in the embodiment of the present application, the model training device may construct the second state transition matrix used in the training process based on the object classification labels in the K-1 th to K-m th periods, and perform model training by combining the object features and the information features in the K-th period; and in the information processing process, a first state transition matrix used in the prediction process is constructed based on the object classification labels from the K-th stage to the K-m + 1-th stage, and the classification result of the K + 1-th stage is predicted by combining the object characteristics and the information characteristics of the K + 1-th stage.

In step 607, the information processing apparatus acquires a first state transition matrix based on the object classification tags of the respective objects in each history period in the first period group.

The first state transition matrix is used for indicating the change trend of the probability of target action generated on target information by each object in the first period group.

In one possible implementation, obtaining the first state transition matrix based on the object classification label of each object in each history period in the first period group includes:

acquiring unbiased estimation of state transition probabilities of m historical periods in the first period group based on the object classification labels of the objects in each historical period in the first period group;

a first state transition matrix is constructed based on unbiased estimates of the state transition probabilities for each of the m history cycles in the first cycle group.

In one possible implementation manner, obtaining an unbiased estimation of the state transition probability of each of the m history cycles in the first cycle group based on the object classification label of each object in each history cycle in the first cycle group includes:

according to the object classification labels of the objects in the first object history period, acquiring a first probability that the object information receives the object action in the first object history period and a second probability that the object information does not receive the object action in the first object history period; the first target history cycle is any one of the history cycles in the first group of cycles;

In one possible implementation manner, constructing the first state transition matrix based on an unbiased estimation of the state transition probability of each of the m history cycles in the first cycle group includes:

and constructing a matrix according to the sequence from the rear to the front based on the unbiased estimation of the state transition probability of each of the m historical periods in the first period group to obtain a first state transition matrix.

The m-order state transition probability matrix used in the prediction stage may be determined as follows:

using the object Classification Label label (Y) of phase K_k) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As first order state transition probability (p)_0,k，1-p_0,k) Unbiased estimation of (2); object Classification Label label (Y) Using stage K-1_k-1) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As second order state transition probability (p)_0,k-1，1-p_0,k-1) Unbiased estimation of (2); object Classification Label label (Y) Using stage K-2_k-2) Calculating the ratio of positive and negative samples (negative sample ratio of

Positive sample fraction of

) As third order state transition probability (p)_0,k-2，1-p_0,k-2) Unbiased estimation of (d). Similarly, the object classification label (Y) of stage K-m +1 is used_k-m+1) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As m-order state transition probability (p)_0,k-m+1，1-p_0,k-m+1) Unbiased estimation of (d). Thereby constructing an unbiased estimation matrix of m-order state transition probability:

for example, also taking m as 3 in the scenario of coupon download service as an example, in the prediction stage, the information processing device uses the classification label Y of three periods of K-2, K-1 and K_k-2、Y_k-1、Y_kRespectively calculating the proportion of positive samples and negative samples of the three periods of K-2, K-1 and K: the K-2 order negative sample ratio is

A positive sample fraction of

The K-1 order negative sample ratio is

A positive sample fraction of

Negative sample ratio of order K of

A positive sample fraction of

Wherein N is_kIndicating the number of negative examples in the classification label of the kth order.

In addition, the case where m is 2 is mainly applied to the case where the current cycle phase is affected by the previous two cycles, i.e., the state S of the K +1 th cycle_k+1State S subjected to phase K, K-1_k、S_k-1And decision u_k(S_k)、u_k-1(S_k-1) And (4) influence. The object classification model may be:

in step 608, the information processing apparatus inputs the object characteristics of each object in the target period, the information characteristics of the target information in the target period, and the first state transition matrix into the object classification model, and obtains the classification result of each object output by the object classification model.

The classification result is used for indicating the probability that the object generates the target action on the target information in the target period.

Taking the coupon downloading service scenario as an example, in the prediction stage, the information processing device may use the trained model W, use the input data in the prediction process, and substitute the input data into K, K-1, …, and K-m +1 order state transition probabilities p_0,k，1-p_0,k，p_0,k-1，1-p_0,k-1，…，p_0,k-m+1，1-p_0,k-m+1The probability of each object downloading the coupon is calculated by the model M3.

In step 609, the information processing apparatus performs a first process on the target information corresponding to the target object in each object in the target period based on the classification result of each object.

Taking a coupon downloading service scene as an example, in a classification recommendation stage, the information processing device divides a classification result obtained by predicting the characteristic data in the K +1 period according to a certain threshold (for example, 0.5) into positive and negative samples (wherein the positive sample is an object willing to download the coupon and is marked as 1, and the negative sample is an object unwilling to download the coupon and is marked as 0), and carries out fueling coupon recommendation on the object divided into 1.

Taking m as 3 in a coupon download service scenario as an example, please refer to fig. 7, which shows a framework diagram of model training and information processing according to an embodiment of the present application. As shown in FIG. 7, the framework includes a training evaluation portion of the model, as well as a model prediction portion.

As shown in fig. 7, in the training evaluation portion of the model, the model training device calculates to obtain a state transition probability matrix 702 used for training by using object classification labels 701 from the K-1 th stage to the K-3 rd stage; the training sample 706 and the test sample 707 are obtained by dividing the K-th stage object feature 703, the K-th stage information feature 704 and the K-th stage object classification label 705. In the training sub-stage, a training sample 706 is utilized to perform sample processing, and model training is performed through the processed training sample and a state transition probability matrix 702 used for training; in the evaluation sub-stage, the test sample 707 is used to process the sample, and the processed test sample and the model of the state transition probability matrix 702 used in training are used to evaluate, if the model evaluation is qualified, the prediction stage is entered, otherwise, the training and evaluation are performed again.

In the prediction part of the model, the information processing equipment calculates to obtain a state transition probability matrix 709 used for prediction by using object classification labels 708 from the K th stage to the K-2 th stage; then, a prediction input 712 is constructed by using the object characteristics 710 at the K +1 th stage and the information characteristics 711 at the K +1 th stage; the prediction input 712 is processed again, model prediction is performed through the processed prediction input and the state transition probability matrix 709 used in prediction, classification results of all objects are obtained, and coupons are issued for the objects with positive classification results.

Taking a service downloading scene of the fueling coupon as an example, the recommendation method based on the traditional two-classification algorithm, the state fusion two-classification recommendation method and the recommendation method based on the Bayesian m-order state transition two-classification recommendation method are compared.

The scheme based on the traditional machine learning binary algorithm comprises the following steps: and constructing sample data by using the object characteristics, the information characteristics and the object classification labels label in the K-1 stage, and randomly dividing the sample data into training samples and testing samples according to a certain proportion. Model training is carried out on training samples by using a traditional LR (low-level random-access) binary classification model, model evaluation is carried out by using test samples, and when the model accords with the model evaluation index, the model W is saved_lr. If the model evaluation does not meet the requirements, the process is repeated by adjusting parameters and the like until the model reaches the requirementsTo an evaluation criterion. And then, performing model prediction by using the object characteristics and the information characteristics in the K-th period, dividing the classification result into 1 and 0 according to a set threshold value, and recommending the fueling coupon for the object marked as 1.

The binary recommendation scheme based on state fusion comprises the following steps: and constructing sample data by using the object characteristics, the information characteristics and the object classification label of the K-1 stage, and randomly dividing the sample data into a training sample and a test sample according to a certain proportion. Model training W using conventional LR two-class model for training samples_lr. Object Classification Label label (Y) Using stage K-1_k-1) Calculating the ratio of positive and negative samples (negative sample ratio of

A positive sample fraction of

) As first order state transition probability (p)_0,k-1，1-p_0,k-1) And model evaluation is performed using the test sample substituted into model M1, and model W is saved when the model meets the model evaluation criterion_lr. If the model evaluation does not meet the requirements, the process is repeated by adjusting parameters and the like until the model meets the evaluation standard. And then using the object characteristics and the information characteristics in the K-th period and the classification result predicted by the positive and negative sample ratio calculation model in the K-th period, dividing the classification result into 1 and 0 according to a set threshold value according to the classification result, and recommending the fueling coupon for the object marked as 1.

And (3) effect comparison: the Bayesian m-order state transition two-classification algorithm provided by the application is compared with the classification effect of the traditional two-classification algorithm and the state fusion two-classification algorithm in the oil-filling coupon downloading scene. Specifically, see table 1 below:

TABLE 1

The scheme shown in the embodiment of the application adopts an m-order state transfer optimization algorithm, and not only is the problem that the classification effect is poor due to high probability concentration degree of prediction of the traditional two-classification algorithm solved.

According to the scheme, the state fusion optimization two-classification algorithm is expanded from the first-order state transition to the m-order state transition, the classification effect can be further improved, meanwhile, the state transition optimization algorithm can be expanded to the maximum extent, and the algorithm optimization of state transition of any order can be adapted.

The two-classification probability model provided by the scheme shown in the embodiment of the application theoretically ensures the realizability of the optimization algorithm through strict mathematical derivation. The model calculation formula is obtained by deduction based on a Bayesian method and by combining a joint probability method with mathematical technologies such as a traditional LR model and unbiased estimation, has theoretical strictness, and ensures the rationality and practical feasibility of a probability calculation model.

The gradient descent method of the scheme shown in the embodiment of the application adopts a gradient descent algorithm derived from a derived m-order state transition classification probability model, so that the model weight is influenced by the m-order state transition probability, the model weight under the m-order state transition probability is further obtained, and the functional relation of the m-order state transition to the model can be more accurately reflected.

The scheme shown in the embodiment of the application is easy to be used in service scenes of various classification algorithms, such as: and service scenes related to the classification algorithm, such as digital marketing, refined operation and the like.

For example, in the digital marketing or refinement operation, the service scenarios include, but are not limited to, the following service scenarios: the method comprises the following steps of loss early warning, payment conversion scene, backflow scene, CTR estimation scene, flow conversion scene and the like in the business such as game business, advertisement business, e-commerce business, preferential refueling, designated driving service and the like.

The classification algorithm-related business scenarios may include: the method comprises a picture classification prediction algorithm, a time series classification prediction algorithm, a series of recommendation algorithms (commodity recommendation, place recommendation, shop recommendation, gift certificate recommendation and the like), a series of estimation algorithms with classification labels, vehicle overspeed estimation, road section congestion estimation, weather condition estimation, system resource distribution estimation and the like.

The algorithm optimization method of the scheme shown in the embodiment of the application can be integrated with machine learning, deep learning and other two-classification algorithms, and the two-classification effect can be obviously improved through algorithm optimization.

In summary, in the information processing method provided in the embodiment of the present application, in the model training process, the computer device combines the object features of each object after the first history period, the information features of the target information after the first history period, and the object classification labels respectively corresponding to each object in a plurality of history periods before the first history period, and trains to obtain the object classification model for predicting the target object matching with the target information in each object. Correspondingly, in the scheme, when determining the object matched with the target information in the subsequent prediction process of the computer device, in addition to combining the object characteristics and the information characteristics, the method also combines the situations that each object performs the target action on the target object in a plurality of historical periods, and the situations that each object performs the target action on the target object in the plurality of historical periods can reflect the change situation of the probability that the target information receives the target action in the plurality of historical periods as a whole, and the change situation of the probability that the target information receives the target action in the plurality of historical periods can influence the probability that each object generates the target action on the target information in the subsequent period; therefore, the above scheme can improve the accuracy of information processing when combining the object classification labels corresponding to the objects in a plurality of history periods.

It is understood that in the specific implementation of the present application, data related to a user, such as a user account, needs to be approved or agreed by the user when the above implementation of the present application is applied to a specific product or technology, and the collection, use and processing of the related data need to comply with relevant laws and regulations and standards of relevant countries and regions.

Fig. 8 is a block diagram showing an information processing apparatus according to an exemplary embodiment of the present application, and as shown in fig. 8, the information processing apparatus includes:

a first feature obtaining module 801, configured to obtain object features of each object in a target period and information features of target information in the target period;

a first tag obtaining module 802, configured to obtain an object classification tag of each object in each history period in a first period group, where the first period group includes m history periods before the target period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

a prediction module 803, configured to predict a classification result of each object based on an object feature of each object in a target period, an information feature of the target information in the target period, and an object classification label of each object in each historical period in the first period group, where the classification result is used to indicate a probability that the object produces the target action on the target information in the target period;

a processing module 804, configured to perform, in the target period, a first process on the target information corresponding to a target object in the objects based on the classification result of the objects.

In one possible implementation, the prediction module 803 is configured to,

Fig. 9 shows a block diagram of an information processing apparatus shown in an exemplary embodiment of the present application, which includes, as shown in fig. 9:

a second feature obtaining module 901, configured to obtain an object feature sample of each object in a first history period and an information feature sample of target information in the first history period;

a second tag obtaining module 902, configured to obtain an object classification tag of each object in each history period in a second period group, where the second period group includes m history periods before the first history period; the object classification label is used for indicating whether the object generates a target action on the target information; m is an integer greater than or equal to 2;

a matrix obtaining module 903, configured to obtain a second state transition matrix based on the object classification label of each object in each history period in the second period group; the second state transition matrix is used for indicating the change trend of the probability of generating the target action on the target information in the second periodic group of each object;

a model training module 904, configured to train and obtain an object classification model based on object features of the objects in the first history period, information features of the target information in the first history period, and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability of the object generating the target action on the target information in a target period.

In one possible implementation manner, the matrix obtaining module 903 is configured to,

Fig. 10 shows a block diagram of a computer device 1000 according to an exemplary embodiment of the present application. The computer device may be implemented as a server in the above-mentioned aspects of the present application. The computer apparatus 1000 includes a Central Processing Unit (CPU) 1001, a system Memory 1004 including a Random Access Memory (RAM) 1002 and a Read-Only Memory (ROM) 1003, and a system bus 1005 connecting the system Memory 1004 and the Central Processing Unit 1001. The computer device 1000 also includes a mass storage device 1006 for storing an operating system 1009, application programs 1010, and other program modules 1011.

The mass storage device 1006 is connected to the central processing unit 1001 through a mass storage controller (not shown) connected to the system bus 1005. The mass storage device 1006 and its associated computer-readable media provide non-volatile storage for the computer device 1000. That is, the mass storage device 1006 may include a computer-readable medium (not shown) such as a hard disk or Compact Disc-Only Memory (CD-ROM) drive.

Without loss of generality, the computer-readable media may comprise computer storage media and communication media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), flash Memory or other solid state Memory technology, CD-ROM, Digital Versatile Disks (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage, or other magnetic storage devices. Of course, those skilled in the art will appreciate that the computer storage media is not limited to the foregoing. The system memory 1004 and mass storage device 1006 described above may be collectively referred to as memory.

The computer device 1000 may also operate as a remote computer connected to a network through a network, such as the internet, in accordance with various embodiments of the present disclosure. That is, the computer device 1000 may be connected to the network 1008 through the network interface unit 1007 connected to the system bus 1005, or may be connected to other types of networks or remote computer systems (not shown) using the network interface unit 1007.

The memory further includes at least one computer program, which is stored in the memory, and the central processing unit 1001 implements all or part of the steps in the information processing method shown in each of the above embodiments by executing the at least one computer program.

In an exemplary embodiment, a computer-readable storage medium is also provided for storing at least one computer program, which is loaded and executed by a processor to implement all or part of the steps of the above-described information processing method. For example, the computer-readable storage medium may be a Read-Only Memory (ROM), a Random Access Memory (RAM), a Compact Disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.

In an exemplary embodiment, a computer program product or a computer program is also provided, which comprises at least one computer program that is loaded by a processor and executes all or part of the steps of the above-mentioned information processing method.

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

It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. An information processing method, characterized in that the method comprises:

2. The method of claim 1, wherein predicting the classification result of the respective object based on the object characteristics of the respective object in a target period, the information characteristics of the target information in the target period, and the object classification label of the respective object in each historical period in the first period group comprises:

3. The method of claim 2, wherein obtaining a first state transition matrix based on the object classification label of the respective object in each historical period of the first period group comprises:

4. The method of claim 3, wherein obtaining an unbiased estimate of the state transition probability for each of the m historical periods in the first periodic group based on the object classification label of the respective object within each historical period in the first periodic group comprises:

5. The method of claim 3, wherein constructing the first state transition matrix based on unbiased estimates of state transition probabilities for each of the m history cycles in the first group of cycles comprises:

6. The method of claim 2, wherein predicting the classification result of each object based on the object feature of each object in the target period, the information feature of the target information in the target period, and the first state transition matrix comprises:

7. The method of claim 6, wherein the first historical period is a period of the first group of periods that is the latest in time.

8. An information processing method, characterized in that the method comprises:

training and obtaining an object classification model based on the object features of the objects in the first history period, the information features of the target information in the first history period and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability that the object generates the target action on the target information in a target period.

9. The method of claim 8, wherein obtaining a second state transition matrix based on the object classification labels of the respective objects in each historical period of the second periodic set comprises:

10. The method of claim 9, wherein obtaining an unbiased estimate of the state transition probability for each of the m historical periods in the second periodic set based on the object classification label of the respective object within each historical period in the second periodic set comprises:

11. An information processing apparatus characterized in that the apparatus comprises:

12. An information processing apparatus characterized in that the apparatus comprises:

the model training module is used for training and obtaining an object classification model based on the object characteristics of each object in the first history period, the information characteristics of the target information in the first history period and the second state transition matrix; the object classification model is used for predicting a classification result of each object, and the classification result is used for indicating the probability that the object generates the target action on the target information in a target period.

13. A computer device comprising a processor and a memory, the memory storing at least one computer program that is loaded and executed by the processor to implement the information processing method according to any one of claims 1 to 10.

14. A computer-readable storage medium, in which at least one computer program is stored, the computer program being loaded and executed by a processor to implement the information processing method according to any one of claims 1 to 10.

15. A computer program product, characterized in that it comprises at least one computer program which is loaded and executed by a processor to implement the information processing method according to any one of claims 1 to 10.