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

Abstract

The application relates to an information processing method and device, 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 features and information features of each object in a target period; acquiring object classification labels of each object in each historical period in a first period group; predicting a classification result of each object based on the object features, the information features and the object classification labels; and processing target information based on the classification result of each object. The above scheme can be applied to a vehicle-mounted scene, and the accuracy of information processing can be improved by combining the object classification labels corresponding to each object in multiple historical periods.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of Internet application, and in particular, relate to an information processing method and device, equipment, storage medium and program product. BACKGROUND

[0002] With the continuous development of big data technology and artificial intelligence (AI) technology, the current information processing system can perform personalized information processing for each object.

[0003] In the related art, for a given information, a recommendation system can input the features of each object and the features of the information to be processed into an AI model for prediction, so as to predict a target object suitable for the information from each object, and then process the information corresponding to the target object.

[0004] However, the AI model in the above scheme relies on relatively single information when making a prediction, resulting in low prediction accuracy of the model and further affecting the accuracy of information processing. SUMMARY

[0005] Embodiments of the present application provide an information processing method, device, equipment, storage medium and program product, which can improve the accuracy of information processing. The technical scheme is as follows:

[0006] On the one hand, an information processing method is provided, which comprises:

[0007] obtaining object features of each object in a target period and information features of target information in the target period;

[0008] obtaining object classification labels of each object in each historical period in a first period group, the first period group comprising m historical periods before the target period; the object classification label is used to indicate whether the object produces a target action on the target information; m is an integer greater than or equal to 2;

[0009] based on the object features of each object in the target period, the information features of the target information in the target period, and the object classification labels of each object in each historical period in the first period group, predicting classification results of the objects, the classification results being used to indicate probabilities of the objects producing the target action on the target information in the target period;

[0010] based on the classification results of the objects, performing a first processing on the target information corresponding to a target object in the objects in the target period.

[0011] In one aspect, an information processing method is provided, the method comprising:

[0012] obtaining object feature samples of each object in a first historical period and information feature samples of target information in the first historical period;

[0013] obtaining object classification labels of the each object in each historical period in a second period group, the second period group comprising m historical periods before the first historical period; the object classification label is used to indicate whether the object produces a target action on the target information; m is an integer greater than or equal to 2;

[0014] obtaining a second state transition matrix based on the object classification labels of the each object in each historical period in the second period group; the second state transition matrix is used to indicate a change trend of a probability that the each object produces the target action on the target information in the second period group;

[0015] training an object classification model based on the object features of the each object in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix; the object classification model is used to predict a classification result of the each object, the classification result is used to indicate a probability that the object produces the target action on the target information in a target period.

[0016] In another aspect, an information processing device is provided, the device comprising:

[0017] a first feature obtaining module, configured to obtain object features of each object in a target period and information features of target information in the target period;

[0018] a first label obtaining module, configured to obtain object classification labels of the each object in each historical period in a first period group, the first period group comprising m historical periods before the target period; the object classification label is used to indicate whether the object produces a target action on the target information; m is an integer greater than or equal to 2;

[0019] a prediction module, configured to predict a classification result of the each object based on the object features of the each object in the target period, the information features of the target information in the target period, and the object classification labels of the each object in each historical period in the first period group, the classification result is used to indicate a probability that the object produces the target action on the target information in a target period;

[0020] The processing module is configured to perform first processing on the target information corresponding to a target object in the plurality of objects based on the classification result of the plurality of objects in the target period.

[0021] In a possible implementation, the prediction module is configured to:

[0022] obtain a first state transition matrix based on the object classification labels of the plurality of objects in each historical period in the first period group; the first state transition matrix is used to indicate a change trend of the probability of the plurality of objects producing the target action on the target information in the first period group;

[0023] predict the classification result of the plurality of objects based on the object features of the plurality of objects in the target period, the information features of the target information in the target period, and the first state transition matrix.

[0024] In a possible implementation, the prediction module is configured to:

[0025] obtain an unbiased estimation of the state transition probability of each of the m historical periods in the first period group based on the object classification labels of the plurality of objects in each historical period in the first period group;

[0026] construct the first state transition matrix based on the unbiased estimation of the state transition probability of each of the m historical periods in the first period group.

[0027] In a possible implementation, the prediction module is configured to:

[0028] obtain a first probability that the target information receives the target action in a first target historical period and a second probability that the target information does not receive the target action in the first target historical period according to the object classification label of the plurality of objects in the first target historical period; the first target historical period is any one of the historical periods in the first period group;

[0029] obtain the first probability and the second probability as the unbiased estimation of the state transition probability of the first target historical period.

[0030] In a possible implementation, the prediction module is configured to:

[0031] obtain the first state transition matrix by performing matrix construction in a time sequence from the last to the first based on the unbiased estimation of the state transition probability of each of the m historical periods in the first period group.

[0032] In a possible implementation, the prediction module is configured to:

[0033] inputting the object features of the respective objects in a target period, the information features of the target information in the target period, and the first state transition matrix into an object classification model, to obtain a classification result of the respective objects output by the object classification model;

[0034] The object classification model is a model trained based on object features of the respective objects in a first historical period, information features of the target information in the first historical period, and a second state transition matrix; the second state transition matrix is obtained based on object classification labels of the respective objects in each historical period in a second period group; and the second period group includes m historical periods before the first historical period.

[0035] In a possible implementation, the first historical period is the latest period in the first period group.

[0036] In another aspect, an information processing apparatus is provided, and the apparatus includes:

[0037] a second feature obtaining module configured to obtain object feature samples of respective objects in a first historical period and information feature samples of target information in the first historical period;

[0038] a second label obtaining module configured to obtain object classification labels of the respective objects in each historical period in a second period group, the second period group including m historical periods before the first historical period; the object classification labels are used to indicate whether the objects produce a target action on the target information; and m is an integer greater than or equal to 2;

[0039] a matrix obtaining module configured to obtain a second state transition matrix based on the object classification labels of the respective objects in each historical period in the second period group; the second state transition matrix is used to indicate a change trend of a probability that the respective objects produce the target action on the target information in the second period group;

[0040] a model training module configured to train an object classification model based on the object features of the respective objects in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix; the object classification model is used to predict a classification result of the respective objects, and the classification result is used to indicate a probability that the objects produce the target action on the target information in a target period.

[0041] In a possible implementation, the matrix obtaining module is configured to,

[0042] obtain unbiased estimates of state transition probabilities of m historical periods in the second period group based on the object classification labels of the respective objects in each of the m historical periods in the second period group;

[0043] construct the second state transition matrix based on the unbiased estimates of state transition probabilities of m historical periods in the second period group.

[0044] In a possible implementation, the matrix obtaining module is configured to,

[0045] obtain a third probability that the target information receives the target action in a second target historical period and a fourth probability that the target information does not receive the target action in the second target historical period according to the object classification labels of the respective objects in the second target historical period, the second target historical period being any one of the historical periods in the second period group;

[0046] obtain the third probability and the fourth probability as unbiased estimates of state transition probabilities of the second target historical period.

[0047] In another aspect, a computer device is provided, which includes a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to implement the information processing method described above.

[0048] In another aspect, a computer readable storage medium is provided, which stores at least one computer program, the computer program being loaded and executed by a processor to implement the information processing method described above.

[0049] In another aspect, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the information processing method provided in the various optional implementation manners described above.

[0050] The technical solutions provided in the present application can include the following beneficial effects:

[0051] The computer device can combine the object features of each object in the target period, the information features of the target information in the target period, and the object classification labels of each object in the plurality of historical periods respectively corresponding to the target object, comprehensively predict the target object matching the target information in each object, and perform the first processing on the target information corresponding to the target object. In the scheme, when the computer device determines the object matching the target information, in addition to combining the object features and the information features, the situation of each object performing the target action on the target object in the plurality of historical periods is also combined, and the situation of each object performing the target action on the target object in the plurality of historical periods can reflect the change of the probability of the target information receiving the target behavior in the plurality of historical periods as a whole, and the change of the probability of the target information receiving the target behavior in the plurality of historical periods will affect the probability of each object generating the target action on the target information in the subsequent period; therefore, the above scheme can improve the accuracy of information processing by combining the object classification labels of each object in the plurality of historical periods respectively corresponding to the target object.

[0052] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0053] The accompanying drawings incorporated in and forming a part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

[0054] Figure 1 A schematic diagram of an information processing process according to an exemplary embodiment of the present application is shown;

[0055] Figure 2 is a structural schematic diagram of an information processing system according to an exemplary embodiment;

[0056] Figure 3 is a flowchart of an information processing method according to an exemplary embodiment;

[0057] Figure 4 is a flowchart of an information processing method according to an exemplary embodiment;

[0058] Figure 5 is a framework diagram of an information processing method according to an exemplary embodiment;

[0059] Figure 6 is a flowchart of an information processing method according to an exemplary embodiment;

[0060] Figure 7 is Figure 6 The framework diagram of model training and information processing according to the embodiment shown;

[0061] Figure 8 A block diagram of an information processing device according to an example embodiment is shown;

[0062] Figure 9 A block diagram of an information processing device according to an example embodiment is shown;

[0063] Figure 10 is a structural block diagram of a computer device according to an example embodiment. DETAILED DESCRIPTION

[0064] The example embodiments will be described in detail herein with reference to the drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following example embodiments are not meant to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0065] The information processing method provided by the embodiments of the present disclosure, in the scenario of selecting part of objects matching a specified information from a plurality of objects, in addition to considering the characteristics of each object and the characteristics of the information, also considers the change of the probability of the target information receiving a target behavior in a plurality of historical periods.

[0066] Figure 1 A schematic diagram of an information processing process according to an example embodiment of the present disclosure is shown, as Figure 1 shown, in the process of processing the target information, the computer device can obtain at least the following three kinds of information: the object characteristics 110 of each object, the information characteristics 120 of the target information, and the change of the probability of the target information receiving a target behavior in a plurality of historical periods 130.

[0067] Then, the computer device combines the object characteristics 110, the information characteristics 120, and the change of the probability of the target information receiving a target behavior in a plurality of historical periods 130 to predict the probability of each object respectively generating a target behavior on the target information in the subsequent period.

[0068] Then, the computer device can execute a first processing on the target information corresponding to part of the objects in the subsequent period based on the probability of each of the above objects respectively generating a target behavior on the target information.

[0069] For example, the first processing can include, but is not limited to, sending the target information or associated information of the target information to the terminal of the target object, and the like.

[0070] The terminal includes, but is not limited to, a mobile phone, a computer, a smart voice interaction device, a smart home appliance, a vehicle-mounted terminal, and the like.

[0071] Figure 2 Fig. 1 is a structural schematic diagram of an information processing system 200 according to an exemplary embodiment. The information processing system 200 includes a server 220 and a plurality of terminals 240.

[0072] The server 220 includes at least one of a single server, a plurality of servers, a cloud computing platform, and a virtualization center. The server 220 is configured to provide a background service for the terminals 240. In the embodiment of the present application, the server 220 can be configured to perform an information processing process, such as sending the obtained target information to the terminals 240.

[0073] The terminal 240 can be a terminal device having an information receiving function. For example, the terminal 240 can be a mobile phone, a tablet computer, an electronic book reader, smart glasses, a smart watch, an MP3 player, an MP4 player, a laptop computer, a desktop computer, and the like. The terminal 240 can include an application program having an information receiving and displaying function, so as to perform information receiving and displaying. Optionally, the application program can be an application program that needs to be downloaded and installed, or an application program that can be used immediately after being clicked, and the present application is not limited in this regard.

[0074] The terminal 240 and the server 220 are connected through a communication network. Optionally, the communication network is a wired network or a wireless network.

[0075] Optionally, the wireless networks or wired networks described above use standard communications technologies and / or protocols. The networks typically carry Internet traffic, but can also include private networks, such as local area networks (LANs) or wide area networks (WANs). Such networks can also include public networks such as the Internet. The networks can utilize a variety of technologies and / or protocols including, but not limited to, standard communications technologies and / or protocols. In some embodiments, data exchanged over the one or more networks is represented using technologies and / or formats including, but not limited to, Hypertext Markup Language (HTML), Extensible Markup Language (XML), and others. In addition, all or some links can be encrypted using conventional encryption technologies, such as, but not limited to, Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), and others. In some embodiments, custom and / or dedicated data communications technologies can be employed as dial and / or supplement the aforementioned data communication technologies.

[0076] For the sake of understanding, the following first introduces the technical related terms involved in various embodiments of the present application:

[0077] State transition: in the field of machine learning, the state of this stage is often the result of the previous stage state and the previous stage decision. If the state S k and the decision u k (S k ) of the Kth stage are given, then the state S k+1 of the K+1th stage can be determined.

[0078] First-order state transition: the state of this stage is determined by the result of the previous stage state and the previous stage decision, that is, the state S k+1 of K+1 is affected by the state S k of the Kth stage and the decision u k (S k ).

[0079] Second-order state transition: the state of K+1 stage is determined by the result of K, K-1 stage state and decision, that is, the state S k+1 of K+1 is affected by the state S k , S k-1 of K, K-1 stage and the decision u k (Sk ), u k-1 (S k-1 ) influence.

[0080] Bayesian second order state transition: the state S k+1 at stage K+1 is influenced by the states S k , S k-1 at stages K, K-1 and the decision u k (S k ), u k-1 (S k-1 ) and can be expressed as follows:

[0081] u k+1 (S k+1 | S k , S k-1 ) = a • u k (S k ) + β • u k-1 (S k-1 ).

[0082] m order state transition: the state at stage K+1 is determined by the states at stages K, K-1,..., K-m and the decisions, i.e. the state S k+1 at stage K+1 is influenced by the states S k , S k-1 ,..., S k-m at stages K, K-1,..., K-m and the decisions u k (S k ), u k-1 (S k-1 ),..., u k-m (S k-m ).

[0083] Bayesian m order state transition: the state S k+1 at stage K+1 is influenced by the states S k , S k-1 ,..., S k-m at stages K, K-1,..., K-m and the decisions u k (S k ), u k-1 (S k-1 ),..., u k-m (S k-m ) and can be expressed as follows:

[0084] u k+1 (S k+1 | S k , S k-1 ,..., S k-m ) = A 1×(m+1) U(S) (m+1)×1 ;

[0085] wherein A 1×(m+1) = (a1, a2, …, am+1) represents a 1 row m+1 column coefficient matrix; m+1

[0086] U(S) (m+1)×1 = (u k (S k ), u k-1 (S k-1 ), …, u k-m (S k-m )) T represents a m+1 row 1 column decision matrix.

[0087] Sigmoid function: a class of functions defined as follows:

[0088]

[0089] Binary classification LR (Logistic Regression) algorithm: by introducing a Sigmoid function in a linear regression model, the continuous output value of the uncertain range of linear regression is mapped to the (0, 1) range, and the linear regression model is converted into a probability prediction model.

[0090] Gradient descent method of second-order state transition: that is, the gradient descent solving process is affected by the gradient descent under the influence of the second-order state transition probability.

[0091] Gradient descent method of m-order state transition: that is, the gradient descent solving process is affected by the gradient descent under the influence of the m-order state transition probability.

[0092] Unbiased estimation: the mathematical expectation of the estimator is equal to the true value of the estimated parameter.

[0093] CTR estimation: CTR (Click-Through-Rate) is a commonly used term in the field of Internet information (such as advertisements), usually referring to the click-through rate of network advertisements (picture advertisements / text advertisements / keyword advertisements / ranking advertisements / video advertisements, etc.), that is, the actual number of clicks (which can be the number of target pages reached) divided by the number of advertisements displayed (Show content).

[0094] Object characteristics: the behavior records and data extraction of objects in the business. Including: the click, collection, payment amount, payment times, active time, active days, etc. of the object in the business.

[0095] ​Item feature: attribute and data extraction of an item (e.g., target information in the present application). It includes: click rate, payment rate, collection, average payment amount (total payment amount per item / number of people who paid), average active duration (total active time per item / number of active people), etc.

[0096] Download label: label corresponding to the behavior of an object clicking and downloading target information on an activity page. For example, if the target information is clicked and downloaded, it is marked as 1, and if the target information is clicked but not downloaded, it is marked as 0.

[0097] Figure 3 Fig. 1 is a flowchart of an information processing method according to an example embodiment. The information processing method can be executed by a computer device, which can be implemented as an information processing device. As shown in Fig. 2, the computer device can be implemented as a server 220. As shown in Fig. 3, the information processing method can include the following steps: Figure 1 Figure 3

[0098] Step 310: Obtain object features of each object in a target period and information features of target information in the target period.

[0099] The target period can be the period in which the current time is located, or the next period or multiple periods after the current time.

[0100] The object features in the target period can be the features of each object obtained at the beginning of the target period. Alternatively, the object features in the target period can be the features of each object obtained after the completion of the previous period of the target period. Alternatively, the object features in the target period can be the features of each object obtained before the completion of the target period.

[0101] Correspondingly, the information features in the target period can be the features of the target information obtained at the beginning of the target period. Alternatively, the information features in the target period can be the features of the target information obtained after the completion of the previous period of the target period. Alternatively, the information features in the target period can be the features of the target information obtained before the completion of the target period.

[0102] In the embodiments of the present application, the computer device can periodically process the target information.

[0103] The duration of each period in which the computer device processes the target information can be the same, or the duration of each period in which the computer device processes the target information can be different.

[0104] ​​Optionally, the computer device can process the target information continuously or discontinuously between each cycle.

[0105] Optionally, the start and end time of each cycle of the computer device processing the target information can be specified by a manager of the information processing system.

[0106] The objects can be terminals used by users, for example. The objects can be terminals respectively logged in with different user accounts. In general, the objects can be identified by the corresponding user accounts.

[0107] In step 320, an object classification label of each object in each historical cycle in a first cycle group is obtained, the first cycle group including m historical cycles before the target cycle, the object classification label being used to indicate whether the object produces the target action on the target information, and m being an integer greater than or equal to 2.

[0108] In the embodiments of the present application, the object classification label can be a binary classification label, which is used to indicate that the object produces the target action on the target information or that the object does not produce the target action on the target information.

[0109] For an object, the object has an object classification label in each historical cycle.

[0110] For example, if the object A does not produce the target action on the target information in the historical cycle 1 and produces the target action on the target information in the historical cycle 1, the object classification label of the object A in the historical cycle 1 is different from the object classification label of the object A in the historical cycle 2.

[0111] The target action can be an action performed based on the operation of the user, such as detail display, download, forwarding, and the like.

[0112] In step 330, a classification result of each object is predicted based on the object feature of each object in the target cycle, the information feature of the target information in the target cycle, and the object classification label of each object in each historical cycle in the first cycle group, the classification result being used to indicate the probability of the object producing the target action on the target information in the target cycle.

[0113] The historical cycle can be a cycle that has been completed in time.

[0114] In the embodiments of the present application, the classification result can be a single probability value, which can be used to indicate the probability of the object performing the target behavior on the target information. Alternatively, the classification result can also be a binary classification probability distribution, which can indicate the probability of the object performing the target behavior on the target information and the probability of not performing the target behavior.

[0115] In a possible implementation, the computer device can obtain a first state transition matrix based on the object classification labels of the objects in each historical period in the first period group, where the first state transition matrix is used to indicate the trend of the probability of the objects performing the target action on the target information in the first period group. The computer device predicts the classification result of each object based on the object features of the objects in the target period, the information features of the target information in the target period, and the first state transition matrix.

[0116] In the embodiments of the present application, the computer device can determine the trend of the probability of the objects performing the target action on the target information in the m historical periods (or the trend of the probability of the target information receiving the target behavior in the multiple historical periods) according to the object classification labels of the objects in the m historical periods, and represent the trend in the form of a matrix, which is combined with the object features, the information features, and the matrix for subsequent prediction.

[0117] In step 340, based on the classification results of the objects, the target information is processed in the target period corresponding to the target object in the objects.

[0118] In summary, the information processing method provided in the embodiments of the present application can combine the object features of the objects in the target period, the information features of the target information in the target period, and the object classification labels of the objects in the multiple historical periods, to comprehensively predict the target object matching the target information in the objects, and perform the first processing on the target information corresponding to the target object. In this scheme, when the computer device determines the object matching the target information, the object features and the information features are combined, and the situations of the objects performing the target action on the target object in the multiple historical periods are also combined. The situations of the objects performing the target action on the target object in the multiple historical periods can reflect the change of the probability of the target information receiving the target behavior in the multiple historical periods, and the change of the probability of the target information receiving the target behavior in the multiple historical periods can affect the probability of the objects performing the target action on the target information in the subsequent period. Therefore, the above scheme can improve the accuracy of information processing by combining the object classification labels of the objects in the multiple historical periods.

[0119] The aboveFigure 3 The scheme in the illustrated embodiment can be implemented based on AI. For example, the steps of predicting the classification results of each object based on the object features, information features, and object classification labels described above can be performed by an object classification model trained based on AI technology.

[0120] AI is a theory, method, technology 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 knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is the design principle and implementation method of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making. Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc. several major directions.

[0121] Machine learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It is a specialized study of how computers simulate or implement human learning behavior to acquire new knowledge or skills, reorganize existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent, and its applications are widespread in various fields of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning, etc.

[0122] Figure 4 is a flowchart of an information processing method according to an example embodiment. The information processing method can be executed by a computer device, which can be implemented as a model training device. As shown in Figure 4 The information processing method can include the following steps:

[0123] Step 410, obtaining object feature samples of each object after the first historical period, and information feature samples of the target information after the first historical period.

[0124] In step 420, the object classification label of each object in each historical period in a second period group is obtained, the second period group including m historical periods before the first historical period, the object classification label being used to indicate whether the object produces a target action on the target information, m being an integer greater than or equal to 2.

[0125] In step 430, a second state transition matrix is obtained based on the object classification label of each object in each historical period in the second period group, the second state transition matrix being used to indicate a change trend of the probability of each object producing a target action on the target information in the second period group.

[0126] In step 440, an object classification model is trained based on the object feature of each object after the first historical period, the information feature of the target information after the first historical period, and the second state transition matrix, the object classification model being used to predict a classification result of each object, the classification result being used to indicate the probability of the object producing a target action on the target information in a target period.

[0127] In the embodiments of the present application, the sample label information used in the model training process can be the object classification label of each object in the first historical period.

[0128] To sum up, the information processing method provided in the embodiments of the present application, in the model training process, the computer device combines the object feature of each object after the first historical period, the information feature of the target information after the first historical period, and the object classification label of each object in the plurality of historical periods before the first historical period, respectively, to train an object classification model for predicting a target object matching the target information in each object. Accordingly, in this scheme, when the computer device determines the object matching the target information in the subsequent prediction process, it combines not only the object feature and the information feature, but also the case that each object performs a target action on the target object in the plurality of historical periods, and the case that each object performs a target action on the target object in the plurality of historical periods can reflect the change of the probability that the target information receives a target behavior in the plurality of historical periods as a whole, and the change of the probability that the target information receives a target behavior in the plurality of historical periods will affect the probability that each object produces a target action on the target information in the subsequent period; therefore, the above scheme can improve the accuracy of information processing by combining the object classification label of each object in the plurality of historical periods, respectively.

[0129] Based on the above Figure 3 and Figure 4 The embodiments shown in the above are taken as an example that the first state transition matrix and the second state transition matrix are implemented as m-order state matrices, the first historical period is the Kth period (hereinafter referred to as the Kth period), and the target period is the K+1th period, please refer to Figure 5which is an information processing framework diagram according to an exemplary embodiment.

[0130] As shown in the model training phase, the model training device first trains the model by using the object feature samples in the Kth period 51, the information feature samples in the Kth period 52, the object classification labels of each object in the Kth period 53, and the m-order state matrices in the K-1th to K-mth periods 54, to obtain an object classification model. Figure 5 In the prediction phase, the information processing device uses the object classification model to process the object features in the K+1th period 55, the information features in the K+1th period 56, and the m-order state matrices in the Kth to K-m+1th periods 57, to make a classification prediction and obtain a classification result. The information processing device subsequently performs information processing according to the classification result, for example, selects a target object according to the classification result, and sends target information or related information of the target information to the terminal corresponding to the target object.

[0131] The model training device and the information processing device can be implemented as the same entity computer device, or the model training device and the information processing device can be implemented as different entity computer devices.

[0132] In a possible application scenario, the scheme shown in the embodiments of the present application can be applied to the CTR prediction scenario of travel service refueling coupon downloading.

[0133] In the construction of the refueling coupon downloading business scenario, the click behavior of each object on the travel service in the K-1th period and the downloading behavior of each object on the coupon in the Kth period are cross-matched. If an object produces a click behavior in the travel service page in the K-1th period, but produces a coupon downloading behavior in the Kth period, it is marked as downloading the coupon in the Kth period (for example, the object classification label of the object in the Kth period is marked as 1), otherwise it is marked as not downloading the coupon in the Kth period (for example, the object classification label of the object in the Kth period is marked as 0). In the construction of the samples used for training and the input for prediction, the object classification label (download / non-download label) and the features of the object in the K-1th period on the travel service, the information features of each function point in the travel service applet are used to construct the training samples and the prediction input data in the coupon downloading business scenario.

[0134] In the scheme shown in the embodiments of the present application, the m-order state matrix additionally added can be a Bayesian m-order state transition matrix. When applied to the refueling coupon downloading business scenario, since the activity time of the refueling coupon is relatively long, and the interval time of the discount operation activity is also relatively long, the scenario of the operation activities of the three consecutive periods with a relatively short interval time can be considered, that is, the value m=3.

[0135] In the scheme shown in the embodiments of the present application, the m-order state matrix additionally added can be a Bayesian m-order state transition matrix. When applied to the refueling coupon downloading business scenario, since the activity time of the refueling coupon is relatively long, and the interval time of the discount operation activity is also relatively long, the scenario of the operation activities of the three consecutive periods with a relatively short interval time can be considered, that is, the value m=3.

[0136] For example, the object classification label of the Kth period can be a coupon download business scenario label (i.e., a download / non-download label); the object features of each period can include the click behavior feature data of the object users of each period on the travel service applet page, and the information features of each period can be the click rate, coupon download rate, exposure, click volume, and the like of each function point of the travel service of each period.

[0137] The scheme shown in each embodiment of the present application can include the following six stages: sample data set preprocessing, Bayesian m-order state transition probability calculation stage, model training stage under m-order state transition, model test evaluation stage under m-order state transition, model prediction stage under m-order state transition, and classification download recommendation stage.

[0138] Figure 6 is a flowchart of an information processing method according to an exemplary embodiment. The information processing method can be executed by a computer device, which can include a model training device and an information processing device. As shown in Figure 6 The information processing method can include the following steps:

[0139] In step 601, the model training device obtains object feature samples of each object in a first historical period and information feature samples of target information in the first historical period.

[0140] In the sample data set preprocessing stage, the model training device can preprocess the sample data set to obtain training and test samples: using the object features, information features, and object classification labels of the Kth period to construct sample data, and dividing the overall sample data (training samples + test samples) into sparse features and dense features. The sparse features can be onehot processed, and the dense features can be processed for principal component analysis (PCA) decorrelation, normalization (standardization), feature discretization, and the like. The model training device randomly divides the processed sparse features and dense features and the object classification labels into training samples (a proportion of a) and test samples (a proportion of 1-a) according to a certain proportion, for example, the samples can be randomly divided into training samples: test samples = 8:2 (i.e., randomly divided into training samples and test samples according to a proportion of 8:2) according to general experience.

[0141] The onehot processing refers to using an N-bit state register to encode N states, each state has an independent register bit, and only one bit is valid at any time. For example, the gender feature: ["male", "female"] corresponds to the onehot encoding: male -> 10; female -> 01. The country feature: ["China", "America", "France"] corresponds to the onehot encoding: China -> 100; America -> 010; France -> 001. The sports feature: ["soccer", "basketball", "badminton", "table tennis"] corresponds to the onehot encoding: soccer -> 1000; basketball -> 0100; badminton -> 0010; table tennis -> 0001. When a sample is ["male", "China", "table tennis"], the complete feature digitization onehot result is: [1, 0, 1, 0, 0, 0, 0, 0, 1].

[0142] The PCA is a widely used data dimension reduction algorithm. The main idea of PCA is to map n-dimensional features to k-dimensional features, where the k-dimensional features are new orthogonal features, also known as principal components, which are k-dimensional features reconstructed on the basis of the original n-dimensional features.

[0143] The object features in the Kth period can include: basic attribute data such as gender, age, and region; active attribute data such as active days, active duration, active function quantity, and the number of days between registration time and current time; recharge attribute data such as recharge amount, consumption amount, recharge times, recharge days, and the number of days between first recharge and current time; coupon attribute data such as function clicks, gift package / coupon types (quantity, times, value), used gift package / coupon types (quantity, value), and expired gift package / coupon types (quantity, value).

[0144] In step 602, the model training device obtains object classification labels of each object in each historical period in the second period group.

[0145] The second period group includes m historical periods before the first historical period; the object classification label is used to indicate whether the object produces a target action on the target information; and m is an integer greater than or equal to 2.

[0146] In the embodiments of the present application, taking the coupon download business scenario as an example, the object classification label (label) of each period can be 1 or 0. For example, 1 represents clicking and downloading (positive sample), and 0 represents clicking but not downloading (negative sample).

[0147] In step 603, the model training device obtains a second state transition matrix based on the object classification labels of each object in each historical period in the second period group.

[0148] The second state transition matrix is ​​used to indicate the changing trend of the probability of generating target actions in response to target information within the second period group of each object.

[0149] In one possible implementation, a second state transition matrix is ​​obtained based on the object classification labels of each object within each historical period of the second period group, including:

[0150] Based on the object classification labels of each object in each historical period in the second period group, obtain unbiased estimates of the state transition probabilities of each of the m historical periods in the second period group.

[0151] The second state transition matrix is ​​constructed based on the unbiased estimates of the state transition probabilities of each of the m historical periods in the second periodic group.

[0152] In one possible implementation, based on the object classification labels of each object within each historical period of the second period group, unbiased estimates of the state transition probabilities for each of the m historical periods in the second period group are obtained, including:

[0153] Based on the object classification labels of each object in the second target historical period, obtain the third probability that the target information receives the target action in the second target historical period, and the fourth probability that the target information does not receive the target action in the second target historical period; the second target historical period is any historical period in the second period group.

[0154] The third and fourth probabilities are obtained as unbiased estimates of the state transition probabilities of the second target historical period.

[0155] During the Bayesian m-order state transition probability calculation phase, the model training device can determine the m-order state transition probabilities for both the training and testing phases:

[0156] Using the object classification label (Y) in period K-1 k-1 Calculate the ratio of positive to negative samples (the proportion of negative samples is...). The proportion of positive samples is ) as the first-order state transition probability (p 0,k-1 ,1-p 0,k-1 Unbiased estimation of ); using the object classification label (Y) in period K-2. k-2 Calculate the ratio of positive to negative samples (the proportion of negative samples is...). The proportion of positive samples is ) as the second-order state transition probability (p 0,k-2 ,1-p 0,k-2 The unbiased estimate of ) is obtained; similarly, the object classification label (Y) of the low Km period is used. k-m Calculate the ratio of positive to negative samples (the proportion of negative samples is...). The proportion of positive samples is ) as an unbiased estimate of the m-order state transition probability (p 0,k -m, 1-p 0,k-m ) of the m-order state transition probability. Thus, the unbiased estimation matrix of the m-order state transition probability is constructed:

[0157]

[0158] For example, taking m=3 in the coupon download service scenario, in the training and testing stages, the classification labels Y k-3 , Y k-2 , Y k-1 of K-3, K-2 and K-1 periods are used to calculate the positive and negative sample proportions of K-3, K-2 and K-1 periods, respectively. The proportion of negative samples of K-3 order is The proportion of positive samples is The proportion of negative samples of K-2 order is The proportion of positive samples is The proportion of negative samples of K-1 order is The proportion of positive samples is

[0159] In step 604, the model training device trains an object classification model based on the object features of each object in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix.

[0160] The object classification model is used to predict the classification results of each object, and the classification results are used to indicate the probability of the object performing a target action on the target information in the target period.

[0161] In one possible implementation, when the model training device performs model training, it can first use the object classification model W trained based on the object features of each object in the first historical period and the information features of the target information in the first historical period, and in the training process, the second state transition matrix is substituted to solve the gradient descent.

[0162] The labeled data used in the training process of the object classification model W can be the object classification labels of each object in the first historical period.

[0163] For example, the model training device can input the object features of each object in the first historical period and the information features of the target information in the first historical period into the object classification model, substitute the second state transition matrix, and obtain the predicted object classification prediction results of each object in the first historical period. Then, according to the difference between the object classification prediction results of each object in the first historical period and the object classification labels of each object in the first historical period, the parameters of the object classification model are updated.

[0164] In the embodiments of the present application, the model training stage includes a model training sub-stage and a model test evaluation sub-stage. In the model training sub-stage, the model training device can use the Bayesian second-order state transition probability model M1 derived by mathematical deduction based on the training samples to substitute K-2, K-1, …, 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-m , and obtain the model weight (the model weight measures the contribution of feature X to Y) by the gradient descent method under the m-order state transition (i.e., the gradient descent solving process is affected by the gradient descent under the m-order state transition probability), to obtain the object classification model W.

[0165] Mathematical studies have shown that the state result of the Kth period has an explicit influence on the prediction result of the K+1th period.

[0166] The probability model (M1) for predicting positive samples is:

[0167]

[0168] The probability model (M2) for predicting negative samples is:

[0169]

[0170] where Y t-1 The probabilities of taking 0 or 1 of the two states are p0 and 1-p0, respectively. denotes the model weight of the t period; denotes the feature of the t period; Y t denotes the classification label of the t period. w Y denotes the label weight coefficient.

[0171] In the model test evaluation stage, the model training device can use the test samples to test the trained model W, and substitute K-2, K-1, …, 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-m , calculate the classification probability and evaluation indicators (recall rate, precision rate, area under curve (AUC), etc.) of the test samples by formula M3, and if the evaluation indicators meet the evaluation effect, save the model W. If the model evaluation fails, repeat the above training and testing steps until the model meets the evaluation indicators.

[0172] In this application, an embodiment proposes a binary classification algorithm calculation model based on Bayesian m-order state transition, namely, the probability model (M3) for predicting a positive result:

[0173]

[0174] Wherein, the stage state of Kj (j = 1, 2, ..., m) is represented by Y. k-j ∈{0,1}, the probabilities of taking the two states 0 or 1 are p respectively. 0,k-j ,1-p 0,k-j W X The model weights represent the features X. k W represents the information feature of the k-th period. Y W represents the label weight coefficient. Y k-j This indicates that when the label Y is in the kj-th period k-j =i j At time, the kj-th label coefficient; where,

[0175]

[0176] After the object classification model is trained by the model training equipment, it can be deployed to the information processing equipment.

[0177] Step 605: The information processing device acquires the object characteristics of each object within the target period, as well as the information characteristics of the target information within the target period.

[0178] During the sample dataset preprocessing stage, the information processing device can also construct the input data used for prediction. For example, the information processing device can use the object features and information features of period K+1, and the object classification labels from period K to period K-m+1, to construct the input data used for prediction, and distinguish the input data used for prediction into sparse features and dense features. Among them, sparse features are processed using one-hot encoding, while dense features are processed using PCA to decorrelate, normalization (standardization), and feature discretization.

[0179] Step 606: The information processing device obtains the object classification labels for each object within each historical period in the first period group.

[0180] The first period group mentioned above includes m historical periods prior to the target period; the object classification label is used to indicate whether the object has generated a target action in response to the target information; m is an integer greater than or equal to 2.

[0181] In one possible implementation, the first historical period is the latest period in the first period group.

[0182] That is, in the embodiment of the present application, the model training device can construct a second state transition matrix used in the training process based on the object classification labels of the K-1th period to the K-mth period, and combine the object features and information features of the Kth period to perform model training; while in the information processing process, a first state transition matrix used in the prediction process is constructed based on the object classification labels of the Kth period to the K-m+1th period, and the object features and information features of the K+1th period are combined to predict the classification result of the K+1th period.

[0183] In step 607, the information processing device obtains a first state transition matrix based on the object classification labels of each object in each historical period in the first period group.

[0184] The first state transition matrix is used to indicate the change trend of the probability of each object producing a target action on the target information in the first period group.

[0185] In a possible implementation, obtaining the first state transition matrix based on the object classification labels of each object in each historical period in the first period group comprises:

[0186] Obtaining an unbiased estimate of the state transition probability of each of the m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group;

[0187] Constructing the first state transition matrix based on the unbiased estimate of the state transition probability of each of the m historical periods in the first period group.

[0188] In a possible implementation, obtaining an unbiased estimate of the state transition probability of each of the m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group comprises:

[0189] According to the object classification label of each object in the first target historical period, obtaining a first probability that the target information receives a target action in the first target historical period, and a second probability that the target information does not receive a target action in the first target historical period; the first target historical period is any one of the historical periods in the first period group;

[0190] The first probability and the second probability are obtained as an unbiased estimate of the state transition probability of the first target historical period.

[0191] In a possible implementation, constructing the first state transition matrix based on the unbiased estimate of the state transition probability of each of the m historical periods in the first period group comprises:

[0192] Based on the unbiased estimation of the state transition probability of each of the m historical periods in the first period group, the matrix is constructed in the order from the last to the first in time, and the first state transition matrix is obtained.

[0193] The determination manner of the m-order state transition probability matrix used in the prediction stage can be as follows:

[0194] The object classification label label(Y k ) of the Kth period is used to calculate the positive and negative sample ratio (the negative sample ratio is The positive sample ratio is ) as the unbiased estimation of the first-order state transition probability (p 0,k , 1-p 0,k ); The object classification label label(Y k-1 ) of the K-1th period is used to calculate the positive and negative sample ratio (the negative sample ratio is The positive sample ratio is ) as the unbiased estimation of the second-order state transition probability (p 0,k-1 , 1-p 0,k-1 ); The object classification label label(Y k-2 ) of the K-2th period is used to calculate the positive and negative sample ratio (the negative sample ratio is The positive sample ratio is ) as the unbiased estimation of the third-order state transition probability (p 0,k-2 , 1-p 0,k-2 ). Similarly, the object classification label label(Y k-m+1 ) of the K-m+1th period is used to calculate the positive and negative sample ratio (the negative sample ratio is The positive sample ratio is ) as the unbiased estimation of the m-order state transition probability (p 0,k-m+1 , 1-p 0,k-m+1 ). Thus, the m-order state transition probability unbiased estimation matrix is constructed:

[0195]

[0196] For example, still taking m=3 in the coupon download business scenario as an example, in the prediction stage, the information processing device uses the classification labels Y k-2 , Y k-1 , Y k of the K-2th, K-1th and Kth periods to calculate the positive and negative sample ratios of the K-2th, K-1th and Kth periods respectively: the K-2th negative sample ratio is The positive sample ratio is The K-1th negative sample ratio is The positive sample ratio is The Kth negative sample ratio is The positive sample ratio is wherein, N k denotes the number of negative samples in the Kth classification label.

[0197] In addition, the case of m=2 is mainly applied to the current cycle stage being affected by the previous two cycles, i.e., the state S k+1 being affected by the states S k and S k-1 of the Kth and K-1th cycles. The decision u k (S k ), u k-1 (S k-1 ) is affected. The object classification model can be:

[0198]

[0199] In step 608, the information processing device inputs the object features of each object in the target cycle, the information features of the target information in the target cycle, and the first state transition matrix into the object classification model, and obtains the classification results of each object output by the object classification model.

[0200] The above classification result is used to indicate the probability of the object generating the target action on the target information in the target cycle.

[0201] Taking the coupon downloading business scenario as an example, in the prediction stage, the information processing device can use the input data in the prediction process and substitute the Kth, K-1th, …, K-m+1th 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+1 into the trained model W, and calculate the probability of each object downloading the coupon through the model M3.

[0202] In step 609, the information processing device performs the first processing on the target information corresponding to the target object in each object in the target cycle based on the classification results of each object.

[0203] Taking the coupon downloading business scenario as an example, in the classification recommendation stage, the information processing device divides the classification results obtained by using the K+1th feature data in the prediction stage into positive and negative samples according to a certain threshold (such as 0.5) (wherein the positive sample is the object willing to download the coupon, marked as 1, and the negative sample is the object not willing to download the coupon, marked as 0), and recommends the oil coupon to the object divided into 1.

[0204] Taking m=3 in the coupon downloading business scenario as an example, please refer to Figure 7, which shows a framework of model training and information processing related to embodiments of the present application. As shown in Figure 7 , the framework includes a model training and evaluation part, and a model prediction part.

[0205] As shown in Figure 7 , in the model training and evaluation part, the model training device calculates a state transition probability matrix 702 used for training using the object classification labels 701 of the K-1th to K-3rd periods; and divides the training samples 706 and the test samples 707 using the object features 703 of the Kth period, the information features 704 of the Kth period, and the object classification labels 705 of the Kth period. In the training sub-stage, the training samples 706 are processed, and the model is trained using the processed training samples and the state transition probability matrix 702 used for training; in the evaluation sub-stage, the test samples 707 are processed, and the model is evaluated using the processed test samples and the state transition probability matrix 702 used for training. If the model evaluation is qualified, the prediction stage is entered, otherwise the training and evaluation are performed again.

[0206] In the model prediction part, the information processing device calculates a state transition probability matrix 709 used for prediction using the object classification labels 708 of the Kth to K-2nd periods; then constructs a prediction input 712 using the object features 710 of the K+1th period and the information features 711 of the K+1th period; and processes the prediction input 712, performs model prediction using the processed prediction input and the state transition probability matrix 709 used for prediction, obtains the classification results of each object, and issues coupons to the objects with positive classification results.

[0207] Next, taking the fuel coupon download business scenario as an example, the traditional two-classification algorithm recommendation method, the state fusion two-classification recommendation method, and the Bayesian m-order state transition two-classification recommendation method proposed in the present application are compared.

[0208] The traditional machine learning two-classification algorithm scheme includes: using the object features, information features, and object classification labels label of the K-1th period to construct sample data, and randomly dividing the sample data into training samples and test samples according to a certain proportion. The traditional LR two-classification model is used to train the model using the training samples, and the model is evaluated using the test samples. When the model meets the model evaluation indicators, the model W lr is saved. If the model evaluation does not meet the requirements, the above process is repeated by adjusting the parameters until the model meets the evaluation standard. Then, the object features and information features of the Kth period are used for model prediction, and according to the classification results, the objects are divided into 1 and 0 according to the set threshold, and the objects marked as 1 are recommended for fuel coupons.

[0209] The state fusion-based binary classification recommendation scheme includes: constructing sample data using object features, information features, and object classification labels from period K-1, and randomly dividing them into training and test samples according to a certain ratio. The training samples are then trained using a traditional LR binary classification model. lr Using the object classification label (Y) from period K-1. k-1 Calculate the ratio of positive to negative samples (the proportion of negative samples is...). The proportion of positive samples is ) as the first-order state transition probability (p 0,k-1 ,1-p 0,k-1 The unbiased estimate of W is obtained, and the test samples are substituted into model M1 for model evaluation. When the model meets the model evaluation index, model W is saved. lr If the model evaluation does not meet the requirements, the above process is repeated by adjusting parameters until the model meets the evaluation criteria. Then, the object features, information features, and the positive and negative sample ratios in period K are used to calculate the classification results predicted by the model. Based on the classification results, the models are divided into 1 and 0 according to the set thresholds, and fuel coupons are recommended to objects labeled as 1.

[0210] Performance Comparison: The proposed Bayesian m-order state transition binary classification algorithm is compared with traditional binary classification algorithms and state fusion binary classification algorithms in the scenario of downloading fuel coupons. See Table 1 below for details:

[0211] Table 1

[0212]

[0213] The solution shown in this application uses an m-order state transition optimization algorithm, which solves the problem that the traditional binary classification algorithm has a high concentration of predicted probabilities, resulting in poor classification performance.

[0214] The scheme shown in this application extends the state fusion optimization binary classification algorithm from a first-order state transition to an m-order state transition, which can further improve the classification effect. At the same time, it can also maximize the expansion of the state transition optimization algorithm and adapt to the algorithm optimization of any order of state transition.

[0215] The binary classification probability model proposed in the embodiments of this application, through rigorous mathematical derivation, theoretically guarantees the feasibility of the optimization algorithm. The model's calculation formula is based on the Bayesian method, derived using joint probability combined with traditional LR models, unbiased estimation, and other mathematical techniques, thus possessing theoretical rigor and ensuring the rationality and practical feasibility of the probability calculation model.

[0216] The gradient descent method of the scheme shown in the embodiments of the present application is a gradient descent algorithm derived from the m-order state transition classification probability model, therefore, the model weight is affected by the m-order state transition probability, and the model weight under the m-order state transition probability is obtained, which can more accurately reflect the function relationship of the m-order state transition to the model.

[0217] The scheme shown in the embodiments of the present application is easy to be used in various classification algorithm business scenarios, such as digital marketing, fine operation and other business scenarios related to classification algorithms.

[0218] For example, in digital marketing or fine operation, the business scenarios include but are not limited to the following business scenarios: loss warning, payment conversion scenarios, backflow scenarios, CTR estimation scenarios, traffic conversion scenarios, etc. in game business, advertising business, e-commerce business, preferential refueling, driving service and other businesses.

[0219] The business scenarios related to classification algorithms can include: picture classification, time series classification prediction algorithm, a series of recommendation algorithms (such as commodity recommendation, location recommendation, store recommendation, gift coupon recommendation, etc.), vehicle overspeed estimation, road congestion estimation, weather condition estimation, system resource distribution estimation and other series of estimation algorithms with classification labels.

[0220] The algorithm optimization method of the scheme shown in the embodiments of the present application can be integrated with binary classification algorithms such as machine learning and deep learning, and the binary classification effect can be significantly improved through algorithm optimization.

[0221] In summary, the information processing method provided by the embodiments of the present application, in the model training process, the computer device combines the object features of each object after the first historical period, the information features of the target information after the first historical period, and the object classification labels corresponding to each object in the plurality of historical periods before the first historical period, and trains to obtain an object classification model for predicting a target object in each object that matches the target information. Correspondingly, in this scheme, when the computer device determines the object that matches the target information in the subsequent prediction process, in addition to combining the object features and the information features, it also combines the cases where each object performs a target action on the target object in the plurality of historical periods, and the cases where each object performs a target action on the target object in the plurality of historical periods can reflect the change of the probability that the target information receives a target behavior in the plurality of historical periods as a whole, and the change of the probability that the target information receives a target behavior in the plurality of historical periods will affect the probability that each object generates a target action on the target information in the subsequent period; therefore, the above scheme can improve the accuracy of information processing by combining the object classification labels corresponding to each object in the plurality of historical periods.

[0222] It can be understood that in the specific embodiments of the present application, the user account and other user-related data are involved. When the above embodiments of the present application are applied to specific products or technologies, the user's permission or consent needs to be obtained, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of the country and region.

[0223] Figure 8 A block diagram of an information processing apparatus according to an example embodiment of the present application is shown, which includes Figure 8

[0224] A first feature acquisition module 801 is configured to acquire object features of each object in a target period and information features of target information in the target period.

[0225] A first label acquisition module 802 is configured to acquire an object classification label of each object in each historical period in a first period group, the first period group including m historical periods before the target period; the object classification label is used to indicate whether the object produces a target action on the target information; m is an integer greater than or equal to 2.

[0226] A prediction module 803 is configured to predict a classification result of each object based on the object features of each object in the target period, the information features 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; the classification result is used to indicate a probability of the object producing the target action on the target information in the target period.

[0227] A processing module 804 is configured to perform a first processing on the target information corresponding to a target object in the target period based on the classification result of each object.

[0228] In a possible implementation, the prediction module 803 is configured to,

[0229] acquire 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 to indicate a trend of change in the probability of the object producing the target action on the target information in the first period group;

[0230] predict the classification result of each object based on the object features of each object in the target period, the information features of the target information in the target period, and the first state transition matrix.

[0231] In a possible implementation, the prediction module 803 is configured to, ​

[0232] obtain unbiased estimates of state transition probabilities of the m historical periods in the first period group based on the object classification labels of the respective objects in each of the historical periods in the first period group;

[0233] construct the first state transition matrix based on the unbiased estimates of state transition probabilities of the m historical periods in the first period group.

[0234] In a possible implementation, the prediction module 803 is configured to,

[0235] obtain a first probability that the target information receives the target action in a first target historical period and a second probability that the target information does not receive the target action in the first target historical period according to the object classification labels of the respective objects in the first target historical period; the first target historical period is any one of the historical periods in the first period group;

[0236] obtain the first probability and the second probability as unbiased estimates of state transition probabilities of the first target historical period.

[0237] In a possible implementation, the prediction module 803 is configured to,

[0238] obtain the first state transition matrix by constructing a matrix in a time order from later to earlier based on the unbiased estimates of state transition probabilities of the m historical periods in the first period group.

[0239] In a possible implementation, the prediction module 803 is configured to,

[0240] input the object features of the respective objects in a target period, the information features of the target information in the target period, and the first state transition matrix into an object classification model, and obtain classification results of the respective objects output by the object classification model;

[0241] The object classification model is a model trained based on object features of the respective objects in a first historical period, information features of the target information in the first historical period, and a second state transition matrix; the second state transition matrix is obtained based on object classification labels of the respective objects in each of historical periods in a second period group; and the second period group includes m historical periods before the first historical period.

[0242] In a possible implementation, the first historical period is the latest period in the first period group.

[0243] Figure 9A block diagram of an information processing device according to an example embodiment of the present application is shown in FIG. 1. As shown in FIG. 1, the information processing device includes: Figure 9

[0244] A second feature acquisition module 901 is configured to acquire object feature samples of each object in a first historical period and information feature samples of target information in the first historical period.

[0245] A second label acquisition module 902 is configured to acquire object classification labels of each object in each historical period in a second period group, the second period group including m historical periods before the first historical period, the object classification label being used to indicate whether the object produces a target action on the target information, and m being an integer greater than or equal to 2.

[0246] A matrix acquisition module 903 is configured to acquire a second state transition matrix based on the object classification labels of each object in each historical period in the second period group, the second state transition matrix being used to indicate a change trend of a probability that the object produces the target action on the target information in the second period group.

[0247] A model training module 904 is configured to train an object classification model based on the object features of each object in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix, the object classification model being used to predict a classification result of the object, the classification result being used to indicate a probability that the object produces the target action on the target information in a target period.

[0248] In a possible implementation, the matrix acquisition module 903 is configured to,

[0249] acquire unbiased estimates of state transition probabilities of the m historical periods in the second period group based on the object classification labels of each object in each historical period in the second period group;

[0250] construct the second state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the second period group.

[0251] In a possible implementation, the matrix acquisition module 903 is configured to,

[0252] ​According to the object classification labels of the respective objects in a second target historical period, a third probability that the target information receives the target action in the second target historical period and a fourth probability that the target information does not receive the target action in the second target historical period are obtained; the second target historical period is any one historical period in the second cycle group;

[0253] The third probability and the fourth probability are obtained as unbiased estimates of state transition probabilities of the second target historical period.

[0254] Figure 10 A structural block diagram of a computer device 1000 is shown, which is an example of an embodiment of the present application. The computer device can be implemented as a server in the above-mentioned scheme of the present application. The computer device 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.

[0255] 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 medium provide non-volatile storage for the computer device 1000. That is, the mass storage device 1006 can include a computer readable medium (not shown) such as a hard disk or a compact disc read-only memory (CD-ROM) drive.

[0256] Without loss of generality, the computer readable medium can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, 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 Disc (DVD), or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices. It should be understood by those skilled in the art that computer storage media does not limit to the foregoing examples. The system memory 1004 and the mass storage device 1006 mentioned above can be collectively referred to as memory.

[0257] According to various embodiments of the present disclosure, the computer device 1000 can also operate in connection with a remote computer through a network such as the Internet. That is, the computer device 1000 can connect to the network 1008 through the network interface unit 1007 connected to the system bus 1005, or in other words, the network interface unit 1007 can be used to connect to other types of networks or remote computer systems (not shown).

[0258] The memory also includes at least one computer program stored therein, and the central processing unit 1001 implements all or part of the steps of the information processing method shown in the various embodiments above by executing the at least one computer program.

[0259] 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 information processing method described above. For example, the computer readable storage medium can 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 and an optical data storage device, etc.

[0260] In an exemplary embodiment, a computer program product or computer program is also provided, the computer program product comprising at least one computer program which is loadable into the working memory of a processor and which causes the processor to carry out all or part of the steps of the information processing method described above.

[0261] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the specification and examples be considered as exemplary only, with the true scope and spirit of the application being indicated by the following claims.

[0262] It is to be understood that the application is not limited to the precise construction described in the specification above and shown in the drawings, and that various modifications and changes can be made by those skilled in the art without departing from the scope of the application. The scope of the application is limited only by the claims that follow.

Claims

1. An information processing method characterized by comprising: The method comprises: obtaining object features of each object in a target period and information features of target information in the target period; the target information comprises a coupon; the object features comprise object click behavior feature data on a travel service applet page; and the information features comprise click rates, coupon download rates, exposure amounts and click amounts of each function point of the travel service; obtaining object classification labels of each object in each historical period in a first period group; the first period group comprises m historical periods before the target period; the object classification labels are used to indicate whether the object generates a target action on the target information; m is an integer greater than or equal to 2; and the target action comprises downloading a coupon; obtaining unbiased estimates of state transition probabilities of m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group; constructing a first state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the first period group; the first state transition matrix is used to indicate a change trend of a probability that the object generates the target action on the target information in the first period group; and predicting classification results of the objects based on the object features of each object in the target period, the information features of the target information in the target period and the first state transition matrix; the classification results are used to indicate the probability that the object generates the target action on the target information in the target period; performing a first processing on the target information corresponding to a target object in the objects in the target period based on the classification results of the objects.

2. The method of claim 1, wherein, The method comprises: obtaining unbiased estimates of state transition probabilities of m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group; constructing a first state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the first period group; the first state transition matrix is used to indicate a change trend of a probability that the object generates the target action on the target information in the first period group; and predicting classification results of the objects based on the object features of each object in the target period, the information features of the target information in the target period and the first state transition matrix; the classification results are used to indicate the probability that the object generates the target action on the target information in the target period; obtaining a first probability that the target information receives the target action in a first target historical period and a second probability that the target information does not receive the target action in the first target historical period according to the object classification labels of each object in the first target historical period; the first target historical period is any one of the historical periods in the first period group; 3. The method of claim 1, wherein, obtaining the first probability and the second probability as unbiased estimates of the state transition probabilities of the first target historical period. The method comprises:

4. The method of claim 1, wherein, obtaining unbiased estimates of state transition probabilities of m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group; constructing a first state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the first period group; the first state transition matrix is used to indicate a change trend of a probability that the object generates the target action on the target information in the first period group; and predicting classification results of the objects based on the object features of each object in the target period, the information features of the target information in the target period and the first state transition matrix; the classification results are used to indicate the probability that the object generates the target action on the target information in the target period; obtaining the first probability and the second probability as unbiased estimates of the state transition probabilities of the first target historical period. inputting the object features of the objects in a target period, the information features of the target information in the target period, and the first state transition matrix into an object classification model, to obtain a classification result of the objects output by the object classification model; The object classification model is a model trained based on object features of the objects in a first historical period, information features of the target information in the first historical period, and a second state transition matrix; the second state transition matrix is obtained based on object classification labels of the objects in each historical period in a second period group; and the second period group includes m historical periods before the first historical period.

5. The method of claim 4, wherein, The first historical period is the latest period in the first period group.

6. An information processing method characterized by comprising: The method comprises: obtaining object feature samples of the objects in a first historical period and information feature samples of the target information in the first historical period; the target information comprises a coupon; the object feature samples comprise object click behavior feature data on a travel service applet page, and the information feature samples comprise click rates, coupon download rates, exposure amounts, and click amounts of each function point of the travel service; obtaining object classification labels of the objects in each historical period in a second period group; the second period group includes m historical periods before the first historical period; the object classification labels are used to indicate whether the objects perform a target action on the target information; m is an integer greater than or equal to 2; and the target action comprises downloading a coupon; obtaining unbiased estimates of state transition probabilities of m historical periods in the second period group based on the object classification labels of the objects in each historical period in the second period group; and constructing a second state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the second period group; the second state transition matrix is used to indicate a change trend of a probability that the objects perform the target action on the target information in the second period group. training an object classification model based on the object features of the objects in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix; the object classification model is used to predict a classification result of the objects, and the classification result is used to indicate a probability that the objects perform the target action on the target information in a target period.

7. The method of claim 6, wherein, The method comprises: obtaining unbiased estimates of state transition probabilities of m historical periods in the second period group based on the object classification labels of the objects in each historical period in the second period group; and constructing a second state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the second period group; the second state transition matrix is used to indicate a change trend of a probability that the objects perform the target action on the target information in the second period group. obtaining a third probability that the target information receives the target action in a second target historical period and a fourth probability that the target information does not receive the target action in the second target historical period according to the object classification labels of the objects in the second target historical period; the second target historical period is any one of the historical periods in the second period group. The third probability and the fourth probability are obtained as unbiased estimates of state transition probabilities of the second target historical period.

8. An information processing apparatus, characterized by comprising: The device comprises: A first feature acquisition module is configured to acquire object features of each object in a target period and information features of target information in the target period; the target information comprises a coupon; the object features comprise object click behavior feature data on a travel service applet page, and the information features comprise click rates, coupon download rates, exposure amounts, and click amounts of each function point of the travel service; A first label acquisition module is configured to acquire object classification labels of each object in each historical period in a first period group; the first period group comprises m historical periods before the target period; the object classification labels are used to indicate whether the object generates a target action on the target information; m is an integer greater than or equal to 2; the target action comprises downloading a coupon; A prediction module is configured to acquire unbiased estimates of state transition probabilities of m historical periods in the first period group based on the object classification labels of each object in each historical period in the first period group; construct a first state transition matrix based on the unbiased estimates of state transition probabilities of the m historical periods in the first period group; the first state transition matrix is used to indicate a change trend of a probability that the object generates the target action on the target information in the first period group; and predict classification results of the objects based on the object features of each object in the target period, the information features of the target information in the target period, and the first state transition matrix; the classification results are used to indicate the probability that the object generates the target action on the target information in the target period. A processing module is configured to perform first processing on the target information corresponding to a target object in the objects in the target period based on the classification results of the objects.

9. An information processing apparatus, characterized by comprising: The device comprises: A second feature acquisition module is configured to acquire object feature samples of each object in a first historical period and information feature samples of target information in the first historical period; the target information comprises a coupon; the object feature samples comprise object click behavior feature data on a travel service applet page, and the information feature samples comprise click rates, coupon download rates, exposure amounts, and click amounts of each function point of the travel service; A second label acquisition module is configured to acquire object classification labels of each object in each historical period in a second period group; the second period group comprises m historical periods before the first historical period; the object classification labels are used to indicate whether the object generates a target action on the target information; m is an integer greater than or equal to 2; and the target action comprises downloading a coupon. The matrix obtaining module is configured to: obtain unbiased estimates of state transition probabilities of m historical periods in the second period group based on the object classification labels of the respective objects in each of the m historical periods in the second period group; and construct a second state transition matrix based on the unbiased estimates of the state transition probabilities of the m historical periods in the second period group. The second state transition matrix is configured to indicate a trend of change in the probability of the respective objects producing the target action on the target information within the second period group. The model training module is configured to train an object classification model based on the object features of the respective objects in the first historical period, the information features of the target information in the first historical period, and the second state transition matrix. The object classification model is configured to predict a classification result of the respective objects, and the classification result is configured to indicate the probability of the objects producing the target action on the target information in a target period.

10. A computer device, comprising: The computer device comprises a processor and a memory. The memory stores at least one computer program. The at least one computer program is loaded and executed by the processor to implement the information processing method according to any one of claims 1 to 7.

11. A computer readable storage medium, characterized in that, The computer readable storage medium stores at least one computer program. The computer program is loaded and executed by the processor to implement the information processing method according to any one of claims 1 to 7.

12. A computer program product, characterised in that, The computer program product comprises at least one computer program. The computer program is loaded and executed by the processor to implement the information processing method according to any one of claims 1 to 7.

Images