Information pushing method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an information pushing method and device, electronic equipment and a storage medium, and relates to the fields of computer technology, artificial intelligence, machine learning and the like. The method comprises the steps of obtaining a target game and to-be-pushed information related to the target game; obtaining a plurality of candidate objects, and determining a plurality of candidate games corresponding to each candidate object; aiming at each candidate object, determining a first game associated with the target game and a second game except the first game from the plurality of candidate games; obtaining object features of each candidate object; through a classification model, determining an object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object; and selecting a plurality of target objects of which the object categories are associated categories from the plurality of candidate objects, and respectively pushing the to-be-pushed information to the plurality of target objects. The target object positioning accuracy is high, and the information pushing effect is good.

Description

Technical Field

[0001] The present application relates to the field of computer technology. Specifically, the present application relates to an information push method, apparatus, electronic device, and storage medium. Background Art

[0002] With the development of computer technology and Internet technology, electronic games have become an increasingly popular form of entertainment. In the early stage of a game manufacturer launching a new game, it is often necessary to promote the new game and push information related to the new game to specific target objects.

[0003] Currently, it is usually business personnel who select the target objects of the information to be pushed based on their own experience. In the existing information push methods, the selection of target objects is greatly affected by the subjective experience of business personnel, and the accuracy of determining target objects is relatively low, resulting in a relatively low matching degree between the information to be pushed of the new game and the target objects. Summary of the Invention

[0004] Embodiments of the present application provide an information push method, apparatus, electronic device, and storage medium, which can solve the problems of relatively low accuracy in positioning target objects and relatively low matching degree between the information to be pushed and the target objects in the existing information push methods.

[0005] The technical solutions are as follows:

[0006] According to one aspect of the embodiments of the present application, an information push method is provided, and the method includes:

[0007] Obtain a target game and information to be pushed related to the target game;

[0008] Obtain a plurality of candidate objects, and determine a plurality of candidate games respectively corresponding to each candidate object; for each candidate object, determine a first game associated with the target game from the plurality of candidate games, and use the games other than the first game among the plurality of candidate games as second games; the candidate games include the games that the candidate object has obtained;

[0009] For each candidate object, obtain the object feature of the candidate object; the object feature includes a first object feature for the first game and a second object feature for the second game;

[0010] The object feature is determined based on at least one of candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information that the candidate object respectively performs on the candidate games;

[0011] Through a classification model, determine the object category respectively corresponding to each candidate object based on the first object feature and the second object feature respectively corresponding to each candidate object;

[0012] The object categories include an associated category associated with the target game and a non-associated category not associated with the target game;

[0013] Select multiple target objects whose object categories are the associated category from the multiple candidate objects, and push the information to be pushed to the multiple target objects respectively.

[0014] Optionally, the determining the object category corresponding to each candidate object based on the first object feature and the second object feature respectively corresponding to each candidate object through the classification model includes:

[0015] For each candidate object, determine the fusion feature corresponding to the candidate object based on the first object feature and its corresponding first weight, and the second object feature and its corresponding second weight;

[0016] Input the fusion feature into the classification model to obtain the object category corresponding to the candidate object output by the classification model.

[0017] Optionally, the determining the object category corresponding to each candidate object based on the first object feature and the second object feature respectively corresponding to each candidate object through the classification model includes:

[0018] For each candidate object, input the first object feature and the second object feature respectively corresponding to the candidate object into the classification model to obtain a first classification result and a second classification result output by the classification model;

[0019] Determine the final classification result based on the first classification result and its corresponding third weight, and the second classification result and its corresponding fourth weight;

[0020] Determine the object category corresponding to the candidate object based on the final classification result.

[0021] Optionally, for each candidate object, the obtaining the first object feature of the candidate object for the first game includes:

[0022] Obtain the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data of the candidate object for the first game;

[0023] Extract the first feature of the first game acquisition data, the second feature of the first game virtual resource transfer data, the third feature of the first game interaction data, and the fourth feature of the first game operation data respectively;

[0024] Fuse the first feature, the second feature, the third feature, and the fourth feature to obtain the first object feature.

[0025] Optionally, for each candidate object, determining the first game associated with the target game from multiple candidate games includes:

[0026] Determining at least one target attribute label of the target game;

[0027] For each candidate game, determining at least one attribute label of the candidate game;

[0028] If at least one attribute label of the candidate game includes any one of the target attribute labels, then regarding the candidate game as the first game.

[0029] Optionally, the method further includes:

[0030] For each first game, determining at least one identical attribute label between the first game and the target game;

[0031] Based on the importance degree indicators respectively corresponding to each identical attribute label, determining the association degree indicator corresponding to the first game;

[0032] The association degree indicator is used to characterize the association degree between the first game and the target game;

[0033] Based on the association degree indicators respectively corresponding to each first game, determining the first weight corresponding to the first object feature or the third weight corresponding to the first classification result.

[0034] Optionally, obtaining the multiple candidate objects includes:

[0035] Obtaining multiple game objects from a preset application; the application is used to provide various game services;

[0036] Based on the activity degree indicators of the multiple game objects for each game in the application, selecting the multiple candidate objects from the multiple game objects.

[0037] Optionally, the classification model is determined in the following manner:

[0038] Obtaining a training sample set, and training an initial classification model based on the training sample set to obtain a trained first classification model; the training sample set includes multiple training sample object features and their corresponding training sample object categories;

[0039] Obtaining a test sample set; the test sample set includes multiple test sample object features and their corresponding test sample object categories;

[0040] Input multiple test sample object features into the first classification model; obtain the test sample prediction categories respectively corresponding to the respective test sample object features output by the first classification model;

[0041] Based on each test sample object category and each test sample prediction category, determine at least one model performance metric corresponding to the first classification model;

[0042] Use the first classification model whose at least one model performance metric meets a preset condition as the classification model.

[0043] Optionally, the determining at least one model performance metric corresponding to the first classification model based on each test sample object category and each test sample prediction category includes:

[0044] Based on the number of first test samples and the number of all positive samples in the test sample set, determine a first model performance metric of the first classification model; the first test samples include test samples whose test sample object category and corresponding test sample prediction category are both associated categories;

[0045] Based on the number of second test samples and the number of all test samples in the second test sample set, determine a second model performance metric of the first classification model; the second test samples include test samples whose test sample object category and corresponding test sample prediction category are the same;

[0046] Use the first classification model whose at least one model performance metric meets a preset condition as the classification model;

[0047] If the first model performance metric is greater than a first preset threshold and the second model performance metric is greater than a second preset threshold, use the initial classification model as the classification model.

[0048] According to another aspect of the embodiments of the present application, there is provided an information push device, and the device includes:

[0049] A first acquisition module, configured to acquire a target game and push information related to the target game;

[0050] A second acquisition module, configured to acquire multiple candidate objects and determine multiple candidate games respectively corresponding to each candidate object; for each candidate object, determine a first game associated with the target game from the multiple candidate games, and use the games other than the first game among the multiple candidate games as second games; the candidate games include the games that the candidate object has acquired;

[0051] A feature extraction module, configured to obtain object features of each candidate object; the object features include first object features for a first game and second object features for a second game;

[0052] The object features are determined based on at least one of candidate game information acquisition, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information that the candidate object performs respectively for the candidate game;

[0053] A classification module, configured to determine the object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object respectively through a classification model;

[0054] The object category includes an associated category associated with the target game and a non-associated category not associated with the target game;

[0055] A push module, configured to select multiple target objects whose object category is the associated category from the multiple candidate objects, and push the information to be pushed to the multiple target objects respectively.

[0056] According to another aspect of the embodiments of the present application, an electronic device is provided. The electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of any of the above information push methods are implemented.

[0057] According to still another aspect of the embodiments of the present application, a computer-readable storage medium is provided. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the steps of any of the above information push methods are implemented.

[0058] The beneficial effects brought by the technical solutions provided by the embodiments of the present application are as follows:

[0059] According to the association relationship between the candidate game and the target game, the first game associated with the target game and the second game not associated with the target game are distinguished from multiple candidate games. Based on at least one of candidate game information acquisition, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information corresponding to the first game and the second game respectively, the first object feature and the second object feature are determined respectively. In the process of constructing the object features, the corresponding object features are constructed from multiple aspects such as game acquisition, game virtual resource transfer, game interaction, and game operation. The information utilization is more comprehensive, and the obtained object features can fully reflect the activity degree of the candidate object in various aspects for the candidate game, so that the constructed object features have better feature expression ability, can better reflect the characteristics of the candidate object, and are beneficial to improving the accuracy of determining the target object based on the object features subsequently.

[0060] Based on the first object feature and the second object feature, the classification model determines the object category of the candidate object, and then determines multiple target objects to be pushed based on the object categories corresponding to each candidate object, realizing the automatic acquisition of multiple target objects to be pushed, avoiding the influence of subjective factors brought by manual screening, being able to quickly and accurately locate the target object from multiple candidate objects, improving the accuracy and efficiency of target object selection; improving the matching degree between the target object and the target game, making the probability that the target object receiving the push information becomes a user of the target game greater, and having a better information push effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] To more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description in the embodiments of the present application.

[0062] Figure 1 It is a diagram of the application scenario of the information push method provided by the embodiment of the present application;

[0063] Figure 2 It is a flowchart of an information push method provided by the embodiment of the present application;

[0064] Figure 3 It is a schematic diagram of a multi-level attribute label provided by the embodiment of the present application;

[0065] Figure 4 It is a flowchart of a classification model training method provided by the embodiment of the present application;

[0066] Figure 5 It is a schematic diagram of a processing flow based on the random forest algorithm provided by the embodiment of the present application;

[0067] Figure 6 It is a graphical schematic diagram of the sigmoid function provided by the embodiment of the present application;

[0068] Figure 7 It is a schematic diagram of the structure of an information push device provided by the embodiment of the present application;

[0069] Figure 8 It is a schematic diagram of the structure of an electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] The following describes the embodiments of the present application with reference to the drawings in the present application. It should be understood that the embodiments described below in conjunction with the drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application, and do not constitute limitations on the technical solutions of the embodiments of the present application.

[0071] Those skilled in the art can understand that, unless specifically stated, the singular forms "a", "an" and "the" used herein may also include the plural forms. It should be further understood that the terms "comprising" and "including" used in the embodiments of the present application mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude the implementation of other features, information, data, steps, operations, elements, components and / or their combinations supported by the technical field of the present invention. It should be understood that when we say an element is "connected" or "coupled" to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and the other element establish a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein indicates at least one of the items defined by the term. For example, "A and / or B" can be implemented as "A", or implemented as "B", or implemented as "A and B".

[0072] To make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0073] First, several terms related to the present application are introduced and explained:

[0074] Positive sample: The sample object features with the object category being the associated category in model training.

[0075] Negative sample: The sample object features with the object category being the non-associated category in model training

[0076] Learner: The basic model of training data. Commonly used ones include decision trees, support vector machines, neural networks, etc.

[0077] Decision tree: A decision tree is a tree structure, where each internal node represents a test on an attribute, each branch represents a test output, and each leaf node represents a category.

[0078] Random forest algorithm: It belongs to one of the ensemble models (a mixture of multiple decision trees) in machine learning. It refers to combining multiple weak decision trees and making the results of multiple weak decision trees vote or take the mean, so that the results of the overall model have high accuracy and generalization performance. The core of the random forest algorithm is the implementation of "randomness". The "randomness" here includes two meanings. One is sampling randomness, which means that the sample sampling for constructing each decision tree is random; the other is feature selection randomness, which means only considering a subset of features to split the nodes in each decision tree.

[0079] Random number seed: When the algorithm performs multiple rounds of iteration, it is necessary to specify the starting point of the iteration, and this starting point is the random number seed.

[0080] Purity: Purity indicates the likelihood that a randomly selected sample is correctly classified in the model. Common methods for representing purity include the Gini coefficient and entropy.

[0081] Gini coefficient: Characterizes the reasonable degree of distribution of the two positive and negative labels in a binary classification problem.

[0082] Maximum tree depth: The depth of a decision tree represents the distance between the leaf node and the root node. The maximum tree depth is the critical point for stopping the iteration of the decision tree. When the decision tree depth reaches the maximum tree depth, the decision tree will stop splitting.

[0083] Number of feature bins: The process of converting continuous features into discrete features.

[0084] Proportion of validation set: When constructing a model, the dataset is split into a training set and a validation set. The training set data is used to construct the model, and the validation set data is used to test the accuracy of the model.

[0085] ROC (Receiver Operating Characteristic Curve) curve: A curve plotted with the true positive rate (sensitivity) as the ordinate and the false positive rate (1 - specificity) as the abscissa according to a series of different binary classification methods.

[0086] True Positive (TP): The true class of the sample is positive, and the result predicted by the model is also positive.

[0087] True Negative (TN): The true class of the sample is negative, and the model predicts it as negative.

[0088] False Positive (FP): The true class of the sample is negative, but the model predicts it as positive.

[0089] False Negative (FN): The true class of the sample is positive, but the model predicts it as negative.

[0090] True positive rate: The proportion of samples predicted as positive by the classifier among the actual positive sample quantity. Also known as recall.

[0091] False positive rate: The proportion of samples predicted as positive by the classifier among the actual negative sample quantity.

[0092] Precision: The proportion of actual positive cases among all the results judged as positive.

[0093] AUC (Area Under Curve) value: The area enclosed by the ROC curve and the X-axis. The closer the value is to 1, the better the optimization effect of the model.

[0094] Accuracy: The number of samples with correct predictions / The total number of samples.

[0095] Recall rate: Also known as the recall ratio, it refers to the proportion of correctly predicted samples among all positive samples.

[0096] F1 value: The harmonic mean of precision and recall rate.

[0097] With the development of computer technology and Internet technology, electronic games have become an increasingly popular form of entertainment. To meet the needs and interests of different players, game manufacturers will continuously launch new games.

[0098] In the early stage of a game manufacturer launching a new game, it is often necessary to promote the new game and push information related to the new game to specific target objects. Currently, usually, business personnel select the target objects of the information to be pushed based on their own experience.

[0099] In the existing information push methods, the selection of target objects is greatly affected by the subjective experience of business personnel, and the accuracy of determining target objects is relatively low. As a result, the matching degree between the information to be pushed of the new game and the target objects is relatively low. Even if the target objects receive information related to the new game, they may not necessarily obtain and participate in the new game, and it is impossible to effectively convert the target objects who receive the information into users of the new game.

[0100] The information push method, device, electronic device and storage medium provided by this application aim to solve the above technical problems in the prior art.

[0101] Optionally, the method provided in the embodiments of this application can be implemented based on artificial intelligence (AI) technology. For example, the classification model can be obtained based on machine learning algorithms.

[0102] Artificial intelligence is to use digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, and is a theory, method, technology and application system that can perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a way similar to human intelligence. Artificial intelligence also studies the design principles and implementation methods of various intelligent machines, enabling the machines to have the functions of perception, reasoning and decision-making.

[0103] Artificial intelligence technology is an interdisciplinary subject with a wide range of fields involved, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, autonomous driving, and intelligent transportation.

[0104] Machine Learning (ML) is an interdisciplinary subject that involves multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve their own performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent, and its applications cover all fields of artificial intelligence. Machine learning and deep learning usually include technologies such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and rote learning.

[0105] Optionally, the data processing involved in the method provided by the embodiments of the present application can also be implemented based on cloud technology. For example, various calculations involved in the classification model can be implemented using cloud computing technology, and data such as candidate game acquisition data and candidate game interaction data can be stored in the cloud. Among them, cloud computing is a computing model that distributes computing tasks on a resource pool composed of a large number of computing devices, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides resources is called the "cloud". The resources in the "cloud" seem to be infinitely expandable to users, and can be obtained at any time, used on demand, expanded at any time, and paid according to usage. Cloud storage is a new concept extended and developed from the concept of cloud computing. Cloud storage can integrate a large number of different types of storage devices (storage devices are also called storage nodes) in the network through application software or application interfaces to work together and jointly provide data storage and business access functions for the outside world.

[0106] It should be noted that, in the alternative embodiments of the present application, for relevant data such as object information (object features), when the embodiments in the present application are applied to specific products or technologies, object permission or consent is required, and the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards of relevant countries and regions. That is to say, if the embodiments in the present application involve data related to an object, it needs to be obtained under the authorization and consent of the object, the authorization and consent of relevant departments, and in compliance with relevant laws, regulations, and standards of the country and region. In the embodiments, if personal information is involved, the acquisition of all personal information requires the consent of the individual. If sensitive information is involved, the separate consent of the information subject needs to be obtained, and the embodiments also need to be implemented under the authorization and consent of the object.

[0107] The technical solutions of the embodiments of the present application and the technical effects produced by the technical solutions of the present application will be described below through the description of several exemplary embodiments. It should be noted that the following embodiments can refer to, draw on, or combine with each other. For the same terms, similar features, and similar implementation steps in different embodiments, they will not be described repeatedly.

[0108] Figure 1 The application scenario diagram of the information push method provided by the embodiments of the present application is as Figure 1 shown. In this application scenario, there are at least two terminals 101 and a server 102. One terminal can correspond to one object. For each terminal, the terminal can send multiple candidate games corresponding to it to the server. The server can obtain the target game to be pushed and the push information related to the target game, and based on the multiple candidate games sent by the terminal, determine a first game associated with the target game from the multiple candidate games, and use the games other than the first game in the multiple candidate games as the second games. Then, obtain the object features of the candidate objects; the object features include the first object features for the first game and the second object features for the second game; through a classification model, determine the object categories corresponding to the respective candidate objects based on the first object features and the second object features corresponding to the respective candidate objects; select multiple target objects with the object category being the associated category from the multiple candidate objects, and push the push information to the multiple target objects respectively.

[0109] In the above application scenario, a trained classification model is set in the server, and the server selects the target objects. In other application scenarios, the terminal can select the target objects.

[0110] Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server or server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms. The terminal can be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a laptop computer, a digital broadcast receiver, a MID (Mobile Internet Devices), a PDA (Personal Digital Assistant), a desktop computer, a smart home appliance, a vehicle-mounted terminal (such as a vehicle-mounted navigation terminal, a vehicle-mounted computer, etc.), a smart speaker, a smart watch, etc. The terminal and the server can be directly or indirectly connected through wired or wireless communication methods, but are not limited thereto.

[0111] Figure 2 It is a schematic flowchart of an information push method provided by an embodiment of this application. As Figure 2 shown, this method includes:

[0112] Step S110, obtain a target game and push information to be related to the target game.

[0113] Specifically, the target game can be a new game to be pushed. The push information to be related to the target game can include content for recommending the target game, such as text introducing the target game, interface screenshots of the target game, test links of the target game, and evaluation results of the target game. The content of the push information to be can be specifically set according to different requirements. The embodiments of this application do not make any restrictions on this.

[0114] Step S120, obtain multiple candidate objects, and determine multiple candidate games respectively corresponding to each candidate object; for each candidate object, determine a first game associated with the target game from the multiple candidate games, and use the games other than the first game among the multiple candidate games as second games; the candidate games include the games that the candidate objects have obtained.

[0115] Specifically, the candidate objects can be objects that have participated in games. For each candidate object, the multiple games that the candidate object has obtained can be used as the multiple candidate games corresponding to the candidate object. Among them, the multiple candidate games corresponding to the candidate object can include the games that the candidate object has participated in as a player, or can also include the games that the candidate object has participated in communication and discussion. The embodiments of this application do not make specific limitations on this.

[0116] For multiple candidate games corresponding to each candidate object, the first game associated with the target game and the second game other than the first game can be selected from the multiple candidate games according to the association relationship between the candidate games and the target game. Among them, the first game may include candidate games that are similar or identical to the target game in a certain aspect. For example, games with the same game category as the target game, games with the same game manufacturer as the target game, games belonging to the same game IP (intellectual property) as the target game, etc. The determination method of the first game will be elaborated in detail below.

[0117] Step S130: For each candidate object, obtain the object feature of the candidate object; the object feature includes the first object feature for the first game and the second object feature for the second game; the object feature is determined based on at least one of the candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information of the candidate object for the candidate game.

[0118] Specifically, for each candidate object, after determining the first game and the second game of the candidate object, the first object feature for the first game and the second object feature for the second game can be determined respectively. Among them, the first object feature can be used to describe the activity level of the candidate object participating in the first game, and the second object feature can be used to describe the activity level of the candidate object participating in the second game.

[0119] The first object feature can be determined based on one or a combination of the candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information for the first game; the second object feature can be determined based on one or a combination of the candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information for the second game. The specific construction process of the first object feature and the second object feature will be elaborated in detail below.

[0120] Among them, the candidate game acquisition information can be used to reflect the relevant information of the candidate object obtaining the candidate game, such as the time when the candidate object obtains the candidate game, the channel for obtaining the candidate game, etc.; the candidate game virtual resource transfer information can be used to reflect the payment situation of the candidate object for the candidate game; the candidate game interaction information can be used to reflect the activity of the candidate object participating in the communication and discussion of the candidate game; the candidate game operation information can be used to reflect the activity of the candidate object as a player participating in the candidate game.

[0121] Step S140: Based on the first object feature and the second object feature respectively corresponding to each candidate object, determine the object category corresponding to each candidate object through a classification model; the object category includes an associated category associated with the target game and a non-associated category not associated with the target game.

[0122] Specifically, after determining the first object feature and the second object feature respectively corresponding to each candidate object, the first object feature and the second object feature respectively corresponding to each candidate object can be input into a pre-trained classification model, and the classification model predicts the object category of each candidate object based on the first object feature and the second object feature respectively corresponding to each candidate object.

[0123] Among them, the object category can include an associated category associated with the target game and a non-associated category not associated with the target game. That is to say, candidate objects of the associated category are more interested in the target game and are more likely to become users of the target game; candidate objects of the non-associated category have a lower interest in the target game and are less likely to become users of the target game.

[0124] Step S150: Select multiple target objects with the object category of the associated category from multiple candidate objects, and push the information to be pushed to the multiple target objects respectively.

[0125] Specifically, after determining the object category corresponding to each candidate object respectively, multiple candidate objects with the object category of the associated category can be used as multiple target objects, and the information to be pushed is pushed to the multiple target objects respectively. For example, the information to be pushed can be pushed through a pop-up window, a notification bar, a text message, etc., and the specific push method can be set according to different application scenarios, which is not limited in this embodiment of the present application.

[0126] In the embodiment of the present application, according to the association relationship between the candidate game and the target game, the first game associated with the target game and the second game not associated with the target game are distinguished from multiple candidate games, and at least one of the candidate game acquisition information, the candidate game virtual resource transfer information, the candidate game interaction information, and the candidate game operation information respectively corresponding to the first game and the second game is obtained, and the first object feature and the second object feature are respectively determined. In the construction process of the object feature, the corresponding object feature is constructed from multiple aspects such as game acquisition, game virtual resource transfer, game interaction, and game operation. The information utilization is more comprehensive, and the obtained object feature can fully reflect the activity degree of the candidate object in various aspects for the candidate game, so that the constructed object feature has better feature expression ability, can better reflect the characteristics of the candidate object, and is beneficial to improving the accuracy of determining the target object based on the object feature subsequently.

[0127] Based on the first object feature and the second object feature, the classification model determines the object category of the candidate object, and then based on the object categories corresponding to each candidate object respectively, determines multiple target objects to be pushed, realizing the automatic acquisition of multiple target objects to be pushed, avoiding the influence of subjective factors brought by manual screening, and being able to quickly and accurately locate the target object from multiple candidate objects, improving the accuracy and efficiency of target object selection; improving the matching degree between the target object and the target game, making the probability that the target object receiving the information to be pushed becomes a user of the target game greater, and having a better information push effect.

[0128] As an alternative embodiment, based on the first object feature and the second object feature corresponding to each candidate object respectively, the classification model determines the object category corresponding to each candidate object, including:

[0129] For each candidate object, based on the first object feature and its corresponding first weight, and the second object feature and its corresponding second weight, determine the fusion feature corresponding to the candidate object;

[0130] Input the fusion feature into the classification model to obtain the object category corresponding to the candidate object output by the classification model.

[0131] Specifically, different weights can be set for the first object feature and the second object feature respectively. The first object feature can be the object feature of the candidate object for the first game, and the first game is a game associated with the target game, and a higher weight can be set for the first object feature.

[0132] For each candidate object, the first weight corresponding to the first object feature and the second weight corresponding to the second object feature can be determined, and based on the first weight and the second weight, the first object feature and the second object feature are weighted and fused, and the fused feature is used as the fusion feature corresponding to the candidate object. Among them, the first weight and the second weight can be set in advance, and the first weight can be greater than the second weight.

[0133] Input the fusion feature into the pre-trained classification model, and the classification model predicts and outputs the object category of the candidate object based on the fusion feature.

[0134] In the embodiments of the present application, by setting corresponding first weights and second weights for the first object feature and the second object feature respectively, and performing weighted fusion on the first object feature and the second object feature based on the first weight and the second weight, a fused feature is obtained. Considering that the first game is a game associated with the target game, that is, the candidate objects who have participated in the first game are more likely to be interested in the target game. When the first weight is set to be greater than the second weight, the contribution degree of the first object feature corresponding to the first game to the fused feature can be enhanced, and at the same time, the contribution degree of the second object feature corresponding to the second game to the fused feature can be weakened, so that the constructed fused feature can accurately reflect the tendency of the candidate object towards the target game, highlighting the potential target objects interested in the target game, making the probability that the candidate object interested in the target game is determined as the target object greater, and further improving the accuracy and efficiency of target object selection.

[0135] As an alternative embodiment, based on the first object feature and the second object feature respectively corresponding to each candidate object, the object category corresponding to each candidate object is determined through a classification model, including:

[0136] For each candidate object, the first object feature and the second object feature respectively corresponding to the candidate object are input into the classification model, and the first classification result and the second classification result output by the classification model are obtained;

[0137] Based on the first classification result and its corresponding third weight, and the second classification result and its corresponding fourth weight, the final classification result is determined;

[0138] Based on the final classification result, the object category corresponding to the candidate object is determined.

[0139] Specifically, for each candidate object, the first object feature and the second object feature of the candidate object can also be input into a pre-trained classification model first, and the first classification result and the second classification result corresponding thereto are determined by the classification model respectively based on the first object feature and the second object feature.

[0140] The third weight corresponding to the first classification result and the fourth weight corresponding to the second classification result can be determined, and based on the third weight and the fourth weight, weighted fusion is performed on the first classification result and the second classification result to obtain the final classification result, and based on the final classification result, the object category is determined. The third weight and the fourth weight can be preset, and the third weight can be greater than the fourth weight.

[0141] Among them, the first classification result may include the probability that the object category of the candidate object determined based on the first object feature is the associated category; the second classification result may include the probability that the object category of the candidate object determined based on the second object feature is the associated category; the final classification result may include the probability that the object category of the candidate object is the associated category.

[0142] Optionally, a preset threshold may be set. If the value of the final classification result is greater than the associated threshold, it is determined that the corresponding object category is the associated category; otherwise, it is the non-associated category.

[0143] In the embodiments of the present application, by respectively setting corresponding third weights and fourth weights for the first classification result and the second classification result, and performing weighted fusion on the first classification result and the second classification result based on the third weights and the fourth weights, a final classification result is obtained. When the third weight is set to be greater than the fourth weight, the contribution degree of the first classification result corresponding to the first game to the final classification result can be enhanced, and at the same time, the contribution degree of the second classification result corresponding to the second game to the final classification result can be weakened, so that the obtained final classification result can accurately reflect the tendency of the candidate object for the target game, highlighting potential target objects interested in the target game, making the probability that a candidate object interested in the target game is determined as a target object greater, and further improving the accuracy and efficiency of target object selection.

[0144] As an alternative embodiment, for each candidate object, obtaining the first object feature of the candidate object for the first game includes:

[0145] Obtaining the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data of the candidate object for the first game;

[0146] Respectively extracting the first feature of the first game acquisition data, the second feature of the first game virtual resource transfer data, the third feature of the first game interaction data, and the fourth feature of the first game operation data;

[0147] Fusing the first feature, the second feature, the third feature, and the fourth feature to obtain the first object feature.

[0148] Specifically, to determine the first object feature, the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data of the candidate object for the first game may be obtained first.

[0149] Among them, the first game acquisition data may include data such as the date when the candidate object obtains a preset game platform, the number of days of use on the preset game platform, and the object level. The preset game platform may include the game platform corresponding to the first game, or may include a game service platform related to the first game, such as an application store for downloading the first game, etc.

[0150] The first game virtual resource transfer data may be data related to the candidate object's payment for the first game. Among them, the candidate object's payment for the first game includes the candidate object's recharge in the first game, or the candidate object's acquisition of game packages related to the first game, etc.

[0151] The first game virtual resource transfer data may specifically include: the total payment amount of the candidate object for the first game within a preset time period (such as one year, one quarter, or one month, etc.), the payment frequency of the candidate object (for example, the time interval between the last payment and the current payment), the highest payment amount of the candidate object, the payment change trend of the candidate object (for example, the change in the candidate object's weekly payment amount), etc.

[0152] The first game interaction data may be data related to the candidate object's discussion about the first game on a communication and discussion platform. Among them, the communication and discussion platform may include communities, forums, Tieba, etc.

[0153] The first game interaction data may specifically include the date when the candidate object joins the relevant communication and discussion platform, the monthly active times of the platform, the weekly active times, the number of platforms participated in, the scale of the platform, whether there is recharge within the platform, whether there is a chat operation with other players, the number of players who have a chat operation within the platform, the total payment amount of the players who have a chat operation, the average payment amount per player who has a chat operation, the payment increase rate of the players who have a chat operation before and after the chat, etc.

[0154] The first game operation data may include data related to the candidate object's participation in the first game as a player, such as the login date, login frequency, and online duration of the candidate object in the first game, etc.

[0155] After obtaining the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data, the features corresponding to each type of data can be extracted respectively. For example, the text of each data can be subjected to feature extraction to obtain the corresponding first feature, second feature, third feature, and fourth feature respectively.

[0156] Subsequently, the first feature, the second feature, the third feature, and the fourth feature can be fused, and the fused feature is used as the first object feature. Among them, each feature can be spliced, or vector operations can be performed on each feature. The specific manner of fusing features is not limited in the embodiments of the present application.

[0157] It should be noted that the same method as described above can be used to determine the second object feature corresponding to the second game, which will not be elaborated here.

[0158] In the embodiments of the present application, by obtaining the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data of the candidate object for the first game, and extracting the corresponding first feature, second feature, third feature, and fourth feature, and fusing each feature, the first object feature is obtained. The object feature is constructed from multiple aspects such as game acquisition, game payment, game interaction, and game operation, so that the information is more comprehensively utilized. The obtained object feature can fully reflect the activity degree of the candidate object in various aspects for the candidate game, making the constructed object feature have better feature expression ability and being able to better reflect the characteristics of the candidate object.

[0159] As an alternative embodiment, for each candidate object, determining a first game associated with the target game from multiple candidate games includes:

[0160] Determining at least one target attribute label of the target game;

[0161] For each candidate game, determining at least one attribute label of the candidate game;

[0162] If at least one attribute label of the candidate game contains any one of the target attribute labels, then the candidate game is used as the first game.

[0163] Specifically, to screen out the first game, at least one target attribute label of the target game and at least one attribute label corresponding to each candidate game can be determined. Among them, the target attribute label can be used to describe the characteristics of the target game, and the attribute label can be used to describe the characteristics of the corresponding candidate game.

[0164] Optionally, multi-level attribute labels can be set. First, at least one first-level attribute label can be set, and at least one corresponding second-level attribute label can be set for some or all of the first-level attribute labels. At least one corresponding third-level attribute label can be set for some or all of the second-level attribute labels, and so on. The attribute labels can be refined at multiple levels to achieve a finer-grained distinction of each game.

[0165] After determining the attribute labels of each candidate game and the target attribute labels of the target game, for each candidate game, the attribute labels of the candidate game can be compared with the target attribute labels of the target game. If the attribute labels of the candidate game contain any one of the target attribute labels, that is, there is at least one overlapping attribute label between the candidate game and the target game, it indicates that the candidate game is similar or identical to the target game in a certain aspect. The players of the candidate game can be potential target objects of the target game, and then the candidate game is used as the first game. By performing the above judgment operations on each candidate game respectively, multiple first games can be obtained.

[0166] As an alternative embodiment, the method further includes:

[0167] For each first game, determine at least one identical attribute label between the first game and the target game;

[0168] Based on the importance degree indicators respectively corresponding to each identical attribute label, determine the association degree indicator corresponding to the first game; the association degree indicator is used to represent the association degree between the first game and the target game;

[0169] Based on the association degree indicators respectively corresponding to each first game, determine the first weight corresponding to the first object feature or the third weight corresponding to the first classification result.

[0170] Specifically, although the first game is a game associated with the target game, the association degrees of different first games with the target game are different. Therefore, after determining multiple first games, the association degrees between each first game and the target game can be further determined.

[0171] For each first game, at least one identical attribute label between the first game and the target game can be determined, where the identical attribute label can be an attribute label shared by the first game and the target game.

[0172] Each attribute label is used to describe the characteristics of the game in a certain aspect, and the importance of each attribute label may vary. The association degree indicator between the first game and the target game can be determined based on the importance degree indicators respectively corresponding to each identical attribute label. For example, the importance degree indicator can be used as the weight corresponding to the identical attribute label, and each identical attribute label can be weighted to obtain the association degree indicator.

[0173] Among them, the importance degree indicator is used to represent the importance of the attribute label for determining the target object. The larger the value of the importance degree indicator, the more important the attribute label; the association degree indicator is used to represent the association degree between the first game and the target game. The larger the value of the association degree indicator, the closer the association with the target game.

[0174] Optionally, the importance degree index may be determined based on the hierarchy of the attribute tags. The lower the hierarchy where the attribute tag is located, the larger the corresponding importance degree index. Figure 3 FIG. Figure 3 is a schematic diagram of a multi-level attribute tag provided by an embodiment of the present application. As Figure 3 shown, the attribute tags include multiple first-level attribute tags such as game category, game manufacturer, game IP type, etc.; for the first-level attribute tag of the game category, the game category may further include multiple second-level attribute tags such as strategy, role-playing, competitive, etc.; further, for the second-level attribute tag of the competitive category, multiple third-level attribute tags such as ball games, music, dance, etc. may be set. On this basis, if the importance degrees corresponding to the first-level attribute tag, the second-level attribute tag, and the third-level attribute tag are w1, w2, and w3 respectively, then w1 < w2 < w3 may be set.

[0175] By determining the importance degree index of the attribute tag according to the hierarchy of the attribute tag, the importance degree index of the finer-grained attribute tag is larger, and the correlation degree between the first game and the target game can be more accurately reflected.

[0176] Optionally, different importance degree indexes may be set for multiple attribute tags at the same level. For example, when Figure 3 the attribute tags include multiple first-level attribute tags such as game category, game manufacturer, game IP type, etc., considering that the object stickiness of the same game IP is stronger, a higher importance degree index may be set for the game IP type.

[0177] After obtaining the correlation degree indexes corresponding to each first game, based on the correlation degree indexes corresponding to each first game, the first weight corresponding to the first object feature or the third weight corresponding to the first classification result may be determined.

[0178] In the embodiment of the present application, by adaptively determining the weight of the feature related to the first game based on the correlation degree between each first game and the target game, the contribution of the first game that is more closely related to the target game to the feature representing the first game is greater, and the probability of overlap of the corresponding object groups is greater, which is beneficial to subsequent screening of the target object of the target game.

[0179] As an alternative embodiment, obtaining multiple candidate objects includes:

[0180] Obtaining multiple game objects from a preset application; the application is used to provide various game services;

[0181] Based on the activity degree indexes of the multiple game objects for each game in the application, multiple candidate objects are selected from the multiple game objects.

[0182] Specifically, to screen out multiple candidate objects, multiple game objects can be obtained from a preset application first. The application can be used to provide various game services, such as services for downloading games, services for communicating and discussing games, etc. The multiple game objects can include objects that have obtained the preset application, objects that have downloaded the preset application, and objects that have browsed the preset application.

[0183] Obtain the activity level indicators of each game for the multiple game objects in the application, and screen out multiple candidate objects according to the activity level indicators respectively corresponding to each game object.

[0184] For example, in the preset application, the object level of the game object can be determined according to the activity level of the game object, such as the activity level indicators in forums, Tieba, communities, etc., or the activity level indicators of playing games. For example, the object level can include a low-level ordinary level or a high-level commander level. Considering that the higher the object level, the higher the degree of love of the object for the game, the game objects with the object level of commander level can be used as candidate objects.

[0185] As an alternative embodiment, Figure 4 is a schematic flowchart of a classification model training method provided by an embodiment of the present application. As Figure 4 shown, the classification model is determined based on the following method:

[0186] Obtain a training sample set, and train an initial classification model based on the training sample set to obtain a trained first classification model; the training sample set includes multiple training sample object features and their corresponding training sample object categories;

[0187] Obtain a test sample set; the test sample set includes multiple test sample object features and their corresponding test sample object categories;

[0188] Input the multiple test sample object features into the first classification model to obtain the test sample prediction categories respectively corresponding to each of the multiple test sample object features output by the first classification model;

[0189] Based on each test sample object category and each test sample prediction category, determine at least one model performance index corresponding to the first classification model;

[0190] Use the first classification model whose at least one model performance index meets the preset conditions as the classification model.

[0191] Specifically, a training sample set can be obtained first. The training sample set includes multiple training sample object features and their corresponding training sample object categories, and the training sample object category can be the true category of the training sample object feature.

[0192] Perform at least one training operation on the initial classification model based on the training sample set, and use the initial classification model that meets the preset training end condition as the trained first classification model;

[0193] Among them, the training operation includes:

[0194] (1) Input multiple training sample object features and their corresponding training sample object categories into the initial classification model, and obtain the training sample prediction categories corresponding to the multiple training sample object features output by the initial classification model;

[0195] (2) Determine the loss function based on each training sample object category and each training sample prediction category;

[0196] (3) Adjust the model parameters of the initial classification model based on the loss function, and use the initial classification model with adjusted parameters as the initial classification model corresponding to the next training operation.

[0197] Specifically, input multiple training sample object features and their corresponding training sample object categories into the initial classification model, and through the initial classification model, the prediction category corresponding to each training sample object feature can be obtained. Based on the difference between the object category corresponding to each training sample object feature and the prediction category, the loss function of the initial classification model is determined.

[0198] Based on the loss function, the parameters of the initial classification model corresponding to the current training operation can be adjusted, and the initial classification model with adjusted parameters can participate in the next training operation. By continuously performing the above training operations and constraining the training of the model based on the loss function, the prediction category of the initial classification model is getting closer and closer to the object category of the training sample object feature until it meets the preset training end condition, and the initial classification model that meets the preset training end condition is used as the trained classification model.

[0199] Among them, the training end condition can be the convergence of the loss function. For example, the loss function is less than the set value or the loss function calculated continuously for the set number of times is less than the set value; the training end condition can also be that the number of training times reaches the preset number of times, and the embodiments of the present application do not limit this.

[0200] After obtaining the first classification model initially trained based on the training sample set, a test sample set can be obtained. Among them, the test sample set includes multiple test sample object features and their corresponding test sample object categories, and the test sample object category can be the true category of the test sample object feature.

[0201] Input multiple test sample object features into the first classification model, and obtain the test sample prediction categories corresponding to each test sample object feature output by the first classification model.

[0202] Determine at least one model performance metric corresponding to the first classification model based on each test sample object category and each test sample predicted category, where the model performance metric can be used to reflect the model performance of the first classification model.

[0203] After obtaining at least one model performance metric, determine whether the at least one model performance metric meets a preset condition. Use the first classification model for which the at least one model performance metric meets the preset condition as the classification model; otherwise, retrain the first classification model until the at least one model performance metric of the first classification model meets the preset condition.

[0204] Among them, the at least one model performance metric can include true positive rate, false positive rate, precision rate, AUC value, accuracy rate, recall rate, and F1 value, etc.

[0205] It should be noted that the construction process of the sample object features during the training process is the same as the object feature construction process in the prediction process described above, and will not be elaborated here.

[0206] In the embodiment of the present application, the initial classification model is trained through a training sample set to obtain a preliminarily trained first classification model, and the first classification model is tested through a test sample set to obtain at least one model performance metric of the first classification model. Use the first classification model for which the at least one model performance metric meets the preset condition as the classification model, thereby ensuring the model performance of the classification model for online prediction, and further ensuring the accuracy and reliability of screening out the target object.

[0207] As an alternative embodiment, determining at least one model performance metric corresponding to the first classification model based on each test sample object category and each test sample predicted category includes:

[0208] Determine the first model performance metric of the first classification model based on the number of first test samples and the number of all positive samples in the test sample set; the first test samples include test samples for which both the test sample object category and the corresponding test sample predicted category are associated categories;

[0209] Determine the second model performance metric of the first classification model based on the number of second test samples and the number of all test samples in the second test sample set; the second test samples include test samples for which the test sample object category and the corresponding test sample predicted category are the same;

[0210] Use the first classification model for which the at least one model performance metric meets the preset condition as the classification model;

[0211] If the first model performance metric is greater than the first preset threshold and the second model performance metric is greater than the second preset threshold, use the initial classification model as the classification model.

[0212] Specifically, for each test sample object, the corresponding test sample object category and the test sample prediction category can be compared. The test samples in which both the test sample object category and the corresponding test sample prediction category are associated categories are used as the first test samples; the test samples in which the test sample object category is the same as the corresponding test sample prediction category are used as the second test samples.

[0213] Based on the ratio between the number of the first test samples and the number of all positive samples in the test sample set, it is used as the first model performance index of the first classification model; based on the ratio between the number of the second test samples and the number of all test samples in the second test sample set, it is used as the second model performance index of the first classification model.

[0214] Among them, the first model performance index can be used to reflect the classification ability of the model, and the second model performance index can be used to reflect the accuracy of the model prediction.

[0215] On this basis, the preset condition is set as that the first model performance index is greater than the first preset threshold and the second model performance index is greater than the second preset threshold.

[0216] If the calculated first model performance index and second model performance index meet the above preset conditions, it indicates that the first classification model has good performance, and the first classification model is used as the classification model for online prediction.

[0217] As an alternative embodiment, a classification model can be trained using the random forest algorithm.

[0218] Figure 5 It is a schematic diagram of a processing flow based on the random forest algorithm provided by an embodiment of the present application. As Figure 5 shown, the processing process of the random forest algorithm includes: constructing multiple sample object features and dividing them into a training sample set and a test sample set according to a preset ratio (such as 7:3). Two - layer random sampling is performed on the training sample set: sample randomization and feature randomization. The initial classification model can include multiple decision trees. Figure 5 where Di represents the i - th training sample subset, and Ci represents the i - th decision tree (i.e., classifier). The sampled training samples are put into the decision tree for training to obtain the prediction results of each decision tree. The prediction result of each classifier is obtained by a sigmoid function (an activation function) to convert real numbers in any range into probability values between 0 and 1.

[0219] Figure 6 It is a graphical schematic diagram of the sigmoid function provided by an embodiment of the present application. As Figure 6As shown, it can be seen that the sigmoid function is an S-shaped curve, and its value ranges from [0, 1]. The value of the function will quickly approach 0 or 1 far from 0. This characteristic of it is very important for solving binary classification problems.

[0220] The formula of the sigmoid function is as follows:

[0221]

[0222] Among them, x represents the input data, e represents the natural constant, and θ represents the parameter to be obtained. The meaning of this function is to find the probability that the function value is equal to 1 under the conditions of given x and θ.

[0223] Finally, the prediction results of multiple decision trees are used for voting. Adopting the principle of the minority obeying the majority, the prediction result with more votes is the final output prediction result.

[0224] During the training process, when the specified number of iterations (such as 50) is reached or the Gini coefficient threshold (such as 0.01) is satisfied, the iterative training is stopped to obtain the first classification model. The optimized first classification model is a strong classifier, and the parameters of the first classification model may include random number seed: 24; number of decision trees: 50; purity: Gini coefficient; maximum tree depth: 6; maximum number of feature bins: 32; test set ratio: 20%.

[0225] Use the first classification model to predict the test sample data set. Based on the predicted categories corresponding to the features of multiple test sample objects in the test sample set and the labeled object categories, calculate the AUC value and accuracy of the first classification model, and use the first classification model with both the AUC value and the accuracy greater than 0.5 as the classification model for online prediction.

[0226] Figure 7 It is a schematic structural diagram of an information push device provided by an embodiment of the present application. As Figure 7 shown, the device includes:

[0227] The first acquisition module 210 is used to acquire the target game and the information to be pushed related to the target game;

[0228] The second acquisition module 220 is used to acquire multiple candidate objects and determine multiple candidate games respectively corresponding to each candidate object; for each candidate object, determine the first game associated with the target game from the multiple candidate games, and use the games other than the first game in the multiple candidate games as the second games; the candidate games include the games already acquired by the candidate object;

[0229] A feature extraction module 230, configured to obtain object features of each candidate object; the object features include first object features for a first game and second object features for a second game;

[0230] The object features are determined based on at least one of candidate game information acquisition, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information for the candidate object with respect to a candidate game;

[0231] A classification module 240, configured to determine object categories corresponding to respective candidate objects based on the first object features and the second object features corresponding to the respective candidate objects through a classification model;

[0232] The object categories include an associated category associated with a target game and a non-associated category not associated with the target game;

[0233] A push module 250, configured to select multiple target objects with the object category being the associated category from the multiple candidate objects, and push the information to be pushed to the multiple target objects respectively.

[0234] As an alternative embodiment, the classification module is specifically configured to:

[0235] For each candidate object, determine a fusion feature corresponding to the candidate object based on the first object feature and its corresponding first weight, and the second object feature and its corresponding second weight;

[0236] Input the fusion feature into the classification model to obtain the object category corresponding to the candidate object output by the classification model.

[0237] As an alternative embodiment, the classification module is specifically configured to:

[0238] For each candidate object, input the first object feature and the second object feature corresponding to the candidate object into the classification model to obtain a first classification result and a second classification result output by the classification model;

[0239] Determine a final classification result based on the first classification result and its corresponding third weight, and the second classification result and its corresponding fourth weight;

[0240] Determine the object category corresponding to the candidate object based on the final classification result.

[0241] As an alternative embodiment, the feature extraction module is specifically configured to:

[0242] Obtain the first game acquisition data, the first game virtual resource transfer data, the first game interaction data, and the first game operation data of the candidate object for the first game;

[0243] Extract the first feature of the first game acquisition data, the second feature of the first game virtual resource transfer data, the third feature of the first game interaction data, and the fourth feature of the first game operation data respectively;

[0244] Fuse the first feature, the second feature, the third feature, and the fourth feature to obtain the first object feature.

[0245] As an alternative embodiment, the second acquisition module is specifically configured to:

[0246] Determine at least one target attribute label of the target game;

[0247] For each candidate game, determine at least one attribute label of the candidate game;

[0248] If at least one attribute label of the candidate game contains any one of the target attribute labels, then use the candidate game as the first game.

[0249] As an alternative embodiment, the device further includes a weight determination module, which is used to:

[0250] For each first game, determine at least one same attribute label between the first game and the target game;

[0251] Based on the importance degree indicators respectively corresponding to each same attribute label, determine the correlation degree indicator corresponding to the first game;

[0252] The correlation degree indicator is used to characterize the correlation degree between the first game and the target game;

[0253] Based on the correlation degree indicators respectively corresponding to each first game, determine the first weight corresponding to the first object feature or the third weight corresponding to the first classification result.

[0254] As an alternative embodiment, the first acquisition module is specifically configured to:

[0255] Obtain multiple game objects from a preset application; the application is used to provide various game services;

[0256] Based on the activity degree indicators of the multiple game objects for each game in the application, select the multiple candidate objects from the multiple game objects.

[0257] As an alternative embodiment, the device further includes a classification model determination module, and the classification model determination module includes:

[0258] A training sub-module, configured to obtain a training sample set and train an initial classification model based on the training sample set to obtain a trained first classification model; the training sample set includes multiple training sample object features and their corresponding training sample object categories.

[0259] A testing sub-module, configured to obtain a testing sample set; the testing sample set includes multiple testing sample object features and their corresponding testing sample object categories.

[0260] Input the multiple testing sample object features into the first classification model; obtain the testing sample prediction categories respectively corresponding to the respective testing sample object features output by the first classification model.

[0261] A judging sub-module, configured to determine at least one model performance metric corresponding to the first classification model based on the respective testing sample object categories and the respective testing sample prediction categories.

[0262] Use the first classification model for which the at least one model performance metric meets a preset condition as the classification model.

[0263] As an optional embodiment, the judging sub-module is specifically configured to:

[0264] Determine a first model performance metric of the first classification model based on the number of first testing samples and the number of all positive samples in the testing sample set; the first testing samples include testing samples for which the testing sample object category and the corresponding testing sample prediction category are both associated categories.

[0265] Determine a second model performance metric of the first classification model based on the number of second testing samples and the number of all testing samples in the second testing sample set; the second testing samples include testing samples for which the testing sample object category and the corresponding testing sample prediction category are the same.

[0266] If the first model performance metric is greater than a first preset threshold and the second model performance metric is greater than a second preset threshold, use the initial classification model as the classification model.

[0267] The device according to the embodiments of the present application can execute the method provided by the embodiments of the present application, and the implementation principle is similar. The actions performed by each module in the device of each embodiment of the present application correspond to the steps in the method of each embodiment of the present application. For the detailed function descriptions of each module of the device, reference can specifically be made to the descriptions in the corresponding methods shown above, and details are not described herein again.

[0268] An electronic device is provided in an embodiment of the present application, which includes a memory, a processor, and a computer program stored in the memory. The processor executes the above computer program to implement the steps of the above information push method. Compared with the related art, it can be achieved that: in the process of constructing object features, corresponding object features are constructed from multiple aspects such as game acquisition, game virtual resource transfer, game interaction, and game operation. The information utilization is more comprehensive, and the obtained object features can fully reflect the activity degree of the candidate object in various aspects for the candidate game, so that the constructed object features have better feature expression ability, can better reflect the characteristics of the candidate object, and are beneficial to improving the accuracy of determining the target object based on the object features subsequently; it realizes the automatic acquisition of multiple target objects to be pushed, avoids the influence of subjective factors brought by manual screening, can quickly and accurately locate the target object from multiple candidate objects, and improves the accuracy and efficiency of target object selection; it improves the matching degree between the target object and the target game, makes the probability that the target object receiving the to-be-pushed information becomes a user of the target game greater, and the information push effect is better.

[0269] In an alternative embodiment, an electronic device is provided, such as Figure 8 shown. Figure 8 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. Among them, the processor 4001 and the memory 4003 are connected, such as connected through a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, and the transceiver 4004 may be used for data interaction between this electronic device and other electronic devices, such as data sending and / or data receiving, etc. It should be noted that in practical applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation to the embodiments of the present application.

[0270] The processor 4001 may be a CPU (Central Processing Unit, central processor), a general-purpose processor, a DSP (Digital Signal Processor, data signal processor), an ASIC (Application Specific Integrated Circuit, application-specific integrated circuit), an FPGA (Field Programmable Gate Array, field programmable gate array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in conjunction with the disclosure of the present application. The processor 4001 may also be a combination for implementing computing functions, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0271] The bus 4002 may include a path for transmitting information among the above components. The bus 4002 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 4002 can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 8 it is only represented by a thick line in the figure, but it does not mean that there is only one bus or one type of bus.

[0272] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or it can also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, which is not limited herein.

[0273] The memory 4003 is used to store the computer program for implementing the embodiments of the present application and is controlled by the processor 4001 to execute. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0274] The embodiments of the present application provide a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps and corresponding contents of the foregoing method embodiments can be implemented.

[0275] The terms "first", "second", "third", "fourth", "1", "2", etc. (if any) in the specification, claims and drawings of the present application are used to distinguish similar objects and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than the order shown or described in words.

[0276] It should be understood that although the flowcharts in the embodiments of the present application indicate various operation steps by arrows, the execution order of these steps is not limited to the order indicated by the arrows. Unless there is a clear description in this article, in some implementation scenarios of the embodiments of the present application, the implementation steps in each flowchart can be executed in other orders according to requirements. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage among these sub-steps or stages can also be executed at different times respectively. In the scenario where the execution times are different, the execution order of these sub-steps or stages can be flexibly configured according to requirements, and the embodiments of the present application do not limit this.

[0277] The above are only optional implementation manners of some implementation scenarios of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical concept of the solution of the present application, using other similar implementation means based on the technical idea of the present application also belongs to the protection scope of the embodiments of the present application.

Claims

1. An information push method, characterized in that: include: Obtaining a target game and information to be pushed related to the target game; Acquire multiple candidate objects, and determine multiple candidate games corresponding to each candidate object; For each candidate object, a first game associated with the target game is determined from a plurality of candidate games, and games other than the first game from the plurality of candidate games are used as second games; the candidate games include games that the candidate object has acquired; For each candidate object, obtaining an object feature of the candidate object; the object feature includes a first object feature for the first game and a second object feature for the second game; The object feature is determined based on at least one of candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information of the candidate object for the candidate game respectively; Determine, by means of a classification model, the object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object; The object categories include an associated category associated with a target game and a non-associated category not associated with the target game; A plurality of target objects whose object category is an associated category are selected from the plurality of candidate objects, and the information to be pushed is pushed to the plurality of target objects respectively.

2. The information push method according to claim 1, characterized in that: The determining, by the classification model, the object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object, comprises: For each candidate object, based on a first object feature and a first weight corresponding to the candidate object, and a second object feature and a second weight corresponding to the candidate object, determine a fusion feature corresponding to the candidate object; The fused features are input into the classification model to obtain the object category corresponding to the candidate object output by the classification model.

3. The information push method according to claim 1, characterized in that: The determining, by the classification model, the object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object, comprises: For each candidate object, inputting the first object feature and the second object feature respectively corresponding to the candidate object into the classification model to obtain a first classification result and a second classification result output by the classification model; Determining a final classification result based on the first classification result and its corresponding third weight, and the second classification result and its corresponding fourth weight; Based on the final classification result, the object category corresponding to the candidate object is determined.

4. The information push method according to claim 1, characterized in that: For each candidate object, obtaining a first object feature of the candidate object for the first game includes: Acquire first game acquisition data, first game virtual resource transfer data, first game interaction data, and first game operation data of the candidate object for the first game; Respectively extracting a first feature of the first game acquisition data, a second feature of the first game virtual resource transfer data, a third feature of the first game interaction data, and a fourth feature of the first game operation data; The first feature, the second feature, the third feature and the fourth feature are fused to obtain the first object feature.

5. The information push method according to claim 2 or 3, characterized in that: For each candidate object, determining a first game associated with the target game from a plurality of candidate games comprises: Determining at least one target attribute tag of the target game; For each candidate game, determining at least one attribute tag of the candidate game; If at least one attribute tag of the candidate game includes any target attribute tag, the candidate game is used as the first game.

6. The information push method according to claim 5, characterized in that: The method further comprises: For each first game, determining at least one identical attribute tag between the first game and the target game; Determining a correlation index corresponding to the first game based on importance indexes corresponding to the same attribute tags; The correlation index is used to characterize the correlation between the first game and the target game; Based on the association degree indicators respectively corresponding to the first games, a first weight corresponding to the first object feature or a third weight corresponding to the first classification result is determined.

7. The information push method according to claim 1, characterized in that: The obtaining of multiple candidate objects includes: Acquire multiple game objects from a preset application; the application is used to provide multiple game services; The plurality of candidate objects are selected from the plurality of game objects based on activity level indicators of the plurality of game objects for each game in the application.

8. The information push method according to claim 1, characterized in that: The classification model is determined based on the following method: Acquire a training sample set, and train an initial classification model based on the training sample set to obtain a trained first classification model; the training sample set includes a plurality of training sample object features and their corresponding training sample object categories; Acquire a test sample set; the test sample set includes a plurality of test sample object features and their corresponding test sample object categories; Inputting a plurality of test sample object features into the first classification model; Obtaining the test sample prediction categories corresponding to the test sample object features output by the first classification model; Determining at least one model performance indicator corresponding to the first classification model based on each test sample object category and each test sample prediction category; The first classification model whose at least one model performance indicator meets a preset condition is used as the classification model.

9. The information push method according to claim 8, characterized in that: The determining, based on each test sample object category and each test sample prediction category, at least one model performance indicator corresponding to the first classification model comprises: Determining a first model performance indicator of the first classification model based on the number of first test samples and the number of all positive samples in the test sample set; the first test samples include test samples whose test sample object categories and corresponding test sample prediction categories are both associated categories; Determining a second model performance indicator of the first classification model based on the number of second test samples and the number of all test samples in the second test sample set; the second test samples include test samples whose test sample object categories are the same as the corresponding test sample prediction categories; The first classification model where the at least one model performance indicator meets a preset condition is used as the classification model; If the first model performance indicator is greater than a first preset threshold and the second model performance indicator is greater than a second preset threshold, the initial classification model is used as the classification model.

10. An information push device, characterized in that: include: A first acquisition module, used to acquire a target game and information to be pushed related to the target game; A second acquisition module is used to acquire multiple candidate objects and determine multiple candidate games corresponding to each candidate object; For each candidate object, a first game associated with the target game is determined from a plurality of candidate games, and games other than the first game from the plurality of candidate games are used as second games; the candidate games include games that the candidate object has acquired; A feature extraction module, for obtaining, for each candidate object, an object feature of the candidate object; the object feature comprises a first object feature for the first game and a second object feature for the second game; The object feature is determined based on at least one of candidate game acquisition information, candidate game virtual resource transfer information, candidate game interaction information, and candidate game operation information of the candidate object for the candidate game respectively; A classification module, configured to determine the object category corresponding to each candidate object based on the first object feature and the second object feature corresponding to each candidate object through a classification model; The object categories include an associated category associated with a target game and a non-associated category not associated with the target game; The push module is used to select a plurality of target objects whose object category is an associated category from the plurality of candidate objects, and push the information to be pushed to the plurality of target objects respectively.

11. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images