A method, related apparatus and device for information pushing

Abstract

This application discloses an information push method, related apparatus, and device to improve the accuracy of pushed information. The method includes: receiving an advertising recommendation request sent by a terminal device, wherein the advertising recommendation request carries an object identifier, which is used to uniquely identify a target object; obtaining a first advertising feature of the advertising information, a first object feature of the target object, and a spatiotemporal feature of the advertising recommendation request based on the advertising recommendation request, the spatiotemporal feature describing the time and space information that triggered the advertising recommendation request; performing cross-feature analysis on the first object feature and the spatiotemporal feature to obtain an object cross-feature; performing cross-feature analysis on the first advertising feature and the spatiotemporal feature to obtain an advertising cross-feature; determining the matching degree between the target object and the advertising information based on the first object feature, the object cross-feature, the first advertising feature, the advertising cross-feature, and the spatiotemporal feature; and pushing the advertising information to the terminal device based on the matching degree.

Description

Technical Field

[0001] This application relates to the field of information management technology, and in particular to a method, related apparatus and equipment for information push. Background Technology

[0002] With the development of technology, more and more advertisers are placing ads for their products through advertising platforms and hope that the ads placed through the advertising platforms can reach a good targeted audience. Therefore, the advertising platforms need to find suitable target audiences for each ad.

[0003] Finding the right target audience for each ad can usually be achieved by recalling ads that match the user through conventional ad recall strategies. However, in the process of recalling ads using conventional ad recall strategies, the models used by conventional ad recall strategies are prone to high ad concentration, that is, a large number of users retrieve a relatively small subset of the same ads. This leads to a low match between the recalled ads and the user, as well as a high repetition rate of ads recalled by a large number of users. As a result, the accuracy of users obtaining ads is reduced, and the exposure of other ads is also reduced. Summary of the Invention

[0004] This application provides an information push method, related apparatus, and device, which, by combining spatiotemporal features with the simultaneous application of target objects and advertising information, can obtain more complex and targeted user cross-features and spatial cross-features. This enables the model to learn more complex cross-features to recall more targeted advertising information, thereby reducing the concentration of advertising information and improving the accuracy of obtaining advertising information.

[0005] In view of this, this application provides a method for information push, including:

[0006] Receive an advertising recommendation request sent by a terminal device, wherein the advertising recommendation request carries an object identifier, which is used to uniquely identify the target object;

[0007] The first advertising feature, the first object feature of the target object, and the spatiotemporal feature of the advertising recommendation request are obtained based on the advertising recommendation request. The spatiotemporal feature is used to describe the time and space information that triggered the advertising recommendation request.

[0008] Cross-features are obtained by combining the first object features with the spatiotemporal features;

[0009] By performing cross-feature analysis on the first advertising feature and the spatiotemporal feature, the advertising cross-feature is obtained;

[0010] The matching degree between the target object and the advertising information is determined based on the first object feature, the object cross feature, the first advertisement feature, the advertisement cross feature, and the spatiotemporal feature.

[0011] Advertising information is pushed to terminal devices based on the matching degree.

[0012] Another aspect of this application provides an apparatus for pushing advertising information, comprising:

[0013] The acquisition unit is used to receive an advertising recommendation request sent by the terminal device, wherein the advertising recommendation request carries an object identifier, which is used to uniquely identify the target object;

[0014] The acquisition unit is also used to acquire the first advertising feature of the advertising information, the first object feature of the target object, and the spatiotemporal feature of the advertising recommendation request according to the advertising recommendation request, wherein the spatiotemporal feature is used to describe the time information and space information that triggered the advertising recommendation request;

[0015] The processing unit is used to perform cross-feature analysis on the first object features and the spatiotemporal features to obtain object cross features;

[0016] The processing unit is also used to perform cross-feature analysis on the first advertising feature and the spatiotemporal feature to obtain advertising cross-features;

[0017] The determining unit is used to determine the matching degree between the target object and the advertising information based on the first object features, object cross features, first advertising features, advertising cross features, and spatiotemporal features;

[0018] The processing unit is also used to push advertising information to terminal devices based on the matching degree.

[0019] In one possible design, in another implementation of the embodiments of this application, the processing unit may specifically be used for:

[0020] Cross features are obtained by combining the first object features with the time features;

[0021] Cross features are obtained by combining the first object features with the spatial features.

[0022] In one possible design, in another implementation of the embodiments of this application, the processing unit may specifically be used for:

[0023] By performing cross-feature analysis on the first advertising feature and the time feature, the advertising time cross-feature is obtained;

[0024] By performing cross-feature analysis on the first advertising feature and the spatial feature, the advertising spatial cross-feature is obtained.

[0025] In one possible design, in another implementation of the embodiments of this application,

[0026] The acquisition unit is also used to acquire historical spatiotemporal features corresponding to the object identifier based on the advertising recommendation request;

[0027] The processing unit is also used to perform cross-feature analysis on the first object features and historical spatiotemporal features to obtain object historical cross-features;

[0028] The processing unit is also used to perform cross-feature analysis on the first advertising feature and the historical spatiotemporal feature to obtain the advertising historical cross-feature.

[0029] Specifically, the determining unit can be used to: determine the matching degree between the target object and the advertising information based on the first object feature, object cross feature, object historical cross feature, first advertising feature, advertising cross feature, advertising historical cross feature, spatiotemporal feature, and historical spatiotemporal feature.

[0030] In one possible design, in another implementation of the embodiments of this application, the processing unit may specifically be used for:

[0031] Cross features are obtained by combining the first object feature with the historical time feature;

[0032] Cross features are obtained by combining the first object features with the historical space features.

[0033] In one possible design, in another implementation of the embodiments of this application, the processing unit may specifically be used for:

[0034] By performing cross-feature analysis on the first advertising feature and the historical time feature, the advertising historical time cross-feature is obtained;

[0035] Cross-feature analysis is performed on the first advertising feature and the historical space feature to obtain the advertising historical space cross-feature.

[0036] In one possible design, in another implementation of the embodiments of this application, the determining unit may specifically be used for:

[0037] The first object features, object cross features, and spatiotemporal features are input into the first network, and the first embedding vector is output.

[0038] The first ad features, ad cross features, and spatiotemporal features are input into the second network, and the second embedding vector is output.

[0039] The matching degree is calculated based on the first embedding vector and the second embedding vector.

[0040] In one possible design, in another implementation of the embodiments of this application, the determining unit may specifically be used for:

[0041] The first object feature, object cross feature, and spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0042] The first object feature, object cross feature, and spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0043] The first ad features, ad cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0044] The first ad features, ad cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0045] Calculate the exposure rate matching degree based on the first exposure rate embedding vector and the second exposure rate embedding vector;

[0046] The conversion rate matching degree is calculated based on the first conversion rate embedding vector and the second conversion rate embedding vector.

[0047] In one possible design, in another implementation of the embodiments of this application, the determining unit may specifically be used for:

[0048] Input the first object features, object cross features, object history cross features, and spatiotemporal features into the first network, and output the first embedding vector.

[0049] The first ad feature, ad cross feature, ad history cross feature, and spatiotemporal feature are input into the first network, and the second embedding vector is output.

[0050] In one possible design, in another implementation of the embodiments of this application, the determining unit may specifically be used for:

[0051] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0052] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0053] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0054] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0055] In one possible design, in another implementation of the embodiments of this application, the processing unit may specifically be used for:

[0056] Calculate the target matching degree based on the exposure rate matching degree and the conversion rate matching degree;

[0057] If the target match is greater than the match threshold, then the advertising information is pushed to the terminal device.

[0058] In one possible design, in another implementation of the embodiments of this application,

[0059] The acquisition unit is also used to acquire the associated object features corresponding to the object identifier based on the advertising recommendation request;

[0060] The processing unit is also used to perform cross-feature analysis on the first object features and the associated object features to obtain object-associated cross features;

[0061] The processing unit is also used to perform cross-feature analysis on the first advertisement feature and the associated object feature to obtain the advertisement association cross-feature;

[0062] Specifically, the determining unit can be used to: determine the matching degree between the target object and the advertising information based on the first object feature, object cross feature, object association cross feature, first advertising feature, advertising cross feature, advertising association cross feature, spatiotemporal feature, and associated object feature.

[0063] Another aspect of this application provides a computer device, including: a memory, a transceiver, a processor, and a bus system;

[0064] The memory is used to store programs;

[0065] The processor implements the methods described above when executing a program in memory;

[0066] Bus systems are used to connect memory and processor to enable communication between them.

[0067] Another aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the methods described above.

[0068] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a network device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the network device to perform the methods provided in the above aspects.

[0069] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0070] By receiving advertising recommendation requests from terminal devices, and obtaining the first advertising features of the advertising information, the first object features of the target object, and the spatiotemporal features of the advertising recommendation request based on the request, the model can then perform cross-feature analysis on the first object features and spatiotemporal features to obtain object cross-features, and on the first advertising features and spatiotemporal features to obtain advertising cross-features. Then, based on the first object features, object cross-features, first advertising features, advertising cross-features, and spatiotemporal features, the matching degree between the target object and the advertising information can be determined, and advertising information can be pushed to the terminal device based on the matching degree. Through this method, by performing feature cross-feature analysis on the first object features and the first advertising features with spatiotemporal features respectively, the model can simultaneously apply spatiotemporal features to the target object and advertising information, obtaining more complex and targeted object cross-features and spatial cross-features. This allows the model to learn more complex cross-features to recall more targeted advertising information, thereby reducing the concentration of advertising information and improving the accuracy of advertising information retrieval. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of the architecture of the data control system in an embodiment of this application;

[0072] Figure 2 This is a flowchart of one embodiment of the information push method in this application;

[0073] Figure 3 This is a schematic diagram illustrating the principle of the information push method in the embodiments of this application;

[0074] Figure 4 This is another schematic diagram illustrating the principle of the information push method in the embodiments of this application;

[0075] Figure 5 This is a schematic diagram of one embodiment of the information push device in this application;

[0076] Figure 6 This is a schematic diagram of one embodiment of the computer device described in this application. Detailed Implementation

[0077] This application provides an information push method, related apparatus, and device, which, by combining spatiotemporal features with the simultaneous application of target objects and advertising information, can obtain more complex and targeted object cross features and spatial cross features, enabling the model to learn more complex cross features to recall more targeted advertising information, thereby reducing the concentration of advertising information and improving the accuracy of obtaining advertising information.

[0078] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0079] It is understood that this application proposes an information push method, which is applied to... Figure 1 Please refer to the data control system shown. Figure 1 , Figure 1 This is a schematic diagram of the architecture of the data control system in an embodiment of this application, such as... Figure 1 As shown, the server receives ad recommendation requests from terminal devices and obtains the first ad feature of the ad information, the first object feature of the target object, and the spatiotemporal features that triggered the ad recommendation request based on the ad recommendation request. Then, it can perform cross-feature analysis on the first object feature and the spatiotemporal features to obtain object cross-features, and perform cross-feature analysis on the first ad feature and the spatiotemporal features to obtain ad cross-features. Then, based on the first object feature, object cross-features, first ad feature, ad cross-features, and spatiotemporal features, the matching degree between the target object and the ad information is determined, and the ad information is pushed to the terminal device according to the matching degree. Through the above method, by performing feature cross-feature analysis on the first object feature and the first ad feature with the spatiotemporal features respectively, the spatiotemporal features can be combined to simultaneously apply to the target object and the ad information, obtaining more complex and targeted object cross-features and spatial cross-features. This allows the model to learn more complex cross-features to recall more targeted ad information, thereby reducing the concentration of ad information and improving the accuracy of ad information retrieval.

[0080] It should be noted that the relevant embodiments can be applied to various scenarios such as cloud technology, artificial intelligence, or smart transportation.

[0081] With the rapid development of information technology, cloud technology is gradually permeating all aspects of people's lives. Cloud technology is a general term encompassing network technology, information technology, integration technology, management platform technology, and application technology based on the cloud computing business model. It can form resource pools, providing flexible and convenient on-demand access. Cloud computing technology will become a crucial support. Backend services of technical network systems require substantial computing and storage resources, such as video websites, image websites, and many portal websites. With the rapid development and application of the internet industry, every item may have its own identification mark in the future, requiring data to be transmitted to backend systems for logical processing. Data at different levels will be processed separately, and various industry data will require robust system support, which can only be achieved through cloud computing.

[0082] Cloud security refers to the collective term for security software, hardware, users, organizations, and security cloud platforms based on cloud computing business models. It integrates emerging technologies and concepts such as parallel processing, grid computing, and the identification of unknown virus behavior. Through a large network of clients, it monitors abnormal software behavior on the network, obtains the latest information on Trojans and malware on the internet, sends it to the server for automatic analysis and processing, and then distributes solutions for viruses and Trojans to each client.

[0083] Understandable, Figure 1 Only one type of terminal device is shown in the diagram. In real-world scenarios, many more types of terminal devices can participate in the data processing. These include, but are not limited to, mobile phones, computers, smart voice interaction devices, smart home appliances, and in-vehicle terminals. The specific number and types depend on the actual scenario and are not limited here. Furthermore, Figure 1 The diagram shows one server, but in real-world scenarios, multiple servers can be involved, especially in scenarios involving multi-model training and interaction. The number of servers depends on the specific scenario and is not limited here.

[0084] It should be noted that in this embodiment, the server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal devices and servers can be directly or indirectly connected via wired or wireless communication, and terminal devices and servers can be connected to form a blockchain network; this application does not impose any limitations on this.

[0085] To address the aforementioned issues, this application proposes an information push method, which is generally executed by a server or terminal device. Accordingly, the device used for information push is generally located in the server or terminal device.

[0086] It is understood that, as disclosed in this application, the information push method, related apparatus, and equipment can comprise a blockchain, with multiple servers or terminal devices forming nodes on the blockchain. In practical applications, data sharing between nodes is required within the blockchain, and each node can store advertising data, object data, etc.

[0087] The information push method used in this application will be described below. Please refer to [link / reference]. Figure 2 One embodiment of the information push method in this application includes:

[0088] In step S101, an advertising recommendation request sent by a terminal device is received, wherein the advertising recommendation request carries an object identifier, which is used to uniquely identify the target object;

[0089] In this embodiment, since the target object can generate a corresponding advertising recommendation request through the search terms entered in the search interface of the terminal device, or automatically generate a corresponding advertising recommendation request through the terminal device based on the user's login operation, the terminal device can then send the advertising recommendation request to the server so that the service can retrieve advertising information suitable for the target object based on the object identifier carried in the advertising recommendation request.

[0090] Among them, the object identifier is used to uniquely identify the target object. The object identifier includes user identifier, account identifier, or device identifier, etc. The object identifier can be specifically represented as an integer (int) string or a string, etc.

[0091] Specifically, when a target user logs into a short video application, video website, news website, or virtual goods trading platform through a terminal device, the terminal device can generate a corresponding advertising recommendation request based on the target user's browsing behavior of advertised products on the platform and send it to the server. This allows the server to retrieve targeted advertising information that the target user is highly interested in or pays close attention to based on the advertising recommendation request carrying the target's identifier, thus enabling better delivery to the target user. Furthermore, it should be understood that in this embodiment and subsequent embodiments, "user" refers to the user of the terminal device.

[0092] In step S102, the first advertising feature of the advertising information, the first object feature of the target object, and the spatiotemporal feature of the advertising recommendation request are obtained according to the advertising recommendation request. The spatiotemporal feature is used to describe the time and space information that triggered the advertising recommendation request.

[0093] In this embodiment, after receiving an advertising recommendation request from the terminal device, the system can obtain the first object feature of the target object corresponding to the object identifier, the spatiotemporal feature that triggered the advertising recommendation request, and the first advertising feature of the obtained advertising information. Specifically, the first object feature can be user gender, age, place of residence, user interests or eating habits, or other object features; no specific limitations are imposed here. The spatiotemporal feature describes the time and space information that triggered the advertising recommendation request, and can specifically be the time, location, or orientation of the advertising recommendation request. The first advertising feature can specifically be advertising attributes such as breakfast, snacks, or clothing, or other features; no specific limitations are imposed here.

[0094] Specifically, since different target objects correspond to different advertising recommendation requests, in order to more accurately distinguish the differences between the received advertising recommendation requests and recall more targeted advertising information based on the differences between the advertising recommendation requests, this embodiment can, when receiving an advertising recommendation request sent by a terminal device, obtain the first object feature of the target object corresponding to the object identifier and the spatiotemporal feature that triggered the advertising recommendation request from the object data, time data and behavior data recorded in the target object's log, based on the object identifier carried in the advertising recommendation request, and obtain the first advertising feature of the advertising information from the advertising data stored in the advertising database. There can be multiple advertising information, and each advertising information corresponds to an advertising feature. It can be understood that if there are N advertising information, there can be a first advertising feature to an Nth advertising feature, etc.

[0095] In step S103, the first object features and spatiotemporal features are cross-featured to obtain object cross features;

[0096] In this embodiment, after obtaining the first object feature and spatiotemporal feature of the target object, the object cross feature can be obtained by cross-feature analysis of the first object feature and the spatiotemporal feature. The interaction between the first object feature and the spatiotemporal feature can be taken into account so that the subsequent model can learn the hidden relationship between the first object feature and the spatiotemporal feature through the object cross feature.

[0097] Specifically, after obtaining the first object feature and spatiotemporal feature of the target object, the object cross feature can be obtained by cross-feature analysis of the first object feature and the spatiotemporal feature. Specifically, the first object feature and the spatiotemporal feature can be cross-feature analysis by means of feature multiplication or Cartesian product, or cross-feature analysis can be performed by algorithms such as Factorization Machines (FM) or Field-aware Factorization Machine (FFM). No specific restrictions are imposed here.

[0098] For example, in advertising and recommendation scenarios for virtual or physical products, such as predicting whether a target audience will purchase a beauty product, assuming the first characteristic of the target audience is that their gender is female, and the spatiotemporal characteristic is that it is a promotional day on a trading platform, combining these two characteristics, rather than the first characteristic or the spatiotemporal characteristic alone, can yield a combined characteristic that can further represent whether the target audience will purchase the beauty product.

[0099] Optionally, in another optional embodiment of the information push method provided in this application, the spatiotemporal features include time features and spatial features. Cross-feature analysis is performed on the first object features and the spatiotemporal features to obtain object cross-features, including:

[0100] Cross features are obtained by combining the first object features with the time features;

[0101] Cross features are obtained by combining the first object features with the spatial features.

[0102] In this embodiment, as shown in Table 1, the spatiotemporal features include temporal features and spatial features. The temporal feature refers to the current time when the terminal device issues the ad recommendation request. The spatial feature refers to the current space where the terminal device issues the ad recommendation request, and may include information such as the scene where the ad recommendation request is issued, the location, and the corresponding longitude and latitude. The object temporal cross-feature is used to represent the characteristics of the target object at the current time when the ad recommendation request is triggered. The object spatial cross-feature is used to represent the characteristics of the target object in the current space where the ad recommendation request is triggered, such as preferences or interests.

[0103] Table 1

[0104] Feature name Feature Description Time characteristics This refers to the current time that triggers the ad recommendation request. Different times when an ad recommendation request occurs can significantly alter its meaning, and this can be used to capture properties that dynamically change over time. Spatial features This refers to the current space that triggered the ad recommendation request. It can be used to indicate the scene information, location, and corresponding longitude and latitude of the location that triggered the ad recommendation request. Different locations can contain different scene information, which will make the meaning of the sent ad recommendation request different, and the target of user interest will also be different. Object time crossover features The first object feature can be cross-referenced with different time features to indicate and capture user characteristics at different times. For example, a user's diet differs between the morning and noon, or the items they purchase differ between the morning and evening. Object space intersection features The first object feature can be cross-referenced with different spatial features to indicate differences in user preferences across different locations, capturing the different preferences and interests exhibited by users in different locations. For example, a user in location A may prefer food products, while in location B they may prefer machinery products.

[0105] Specifically, such as Figure 3 As shown, after obtaining the first object feature and spatiotemporal feature, by performing feature cross-features of the first object feature with the temporal feature and the spatial feature respectively, we can obtain object-temporal cross-features that can further express the characteristics of the target object at the current time when the ad recommendation request is triggered, and object-space cross-features that can further express the characteristics of the target object at the current space when the ad recommendation request is triggered. Since the feature dimensions of the object-temporal cross-features and object-space cross-features obtained after feature cross-features are increased compared to the dimensions of the independent first object feature or spatiotemporal feature, the subsequent model can better and more accurately mine and learn the preferences or interests of the target object through object-temporal cross-features and object-space cross-features. It can also reflect feature segmentation through object-temporal cross-features and user-space cross-features, and can be used to reflect the characteristics and differences of ad recommendation requests under different times and contexts.

[0106] For example, assuming the first object feature is user identity such as a student, and the time feature is Saturday night, the feature cross-validation yields the time cross-feature "students on Saturday night," which can indicate that students prefer leisure and entertainment products or applications on Saturday nights. Or, assuming the first object feature is user identity such as an office worker, and the time feature is Monday morning rush hour, the feature cross-validation yields the time cross-feature "office workers on Monday morning rush hour," which can indicate that office workers pay more attention to ride-hailing apps or advertisements during Monday morning rush hour.

[0107] In step S104, the first advertising feature and the spatiotemporal feature are cross-featured to obtain the advertising cross-feature;

[0108] In this embodiment, since advertising information also changes significantly over time and is affected by spatial changes, after obtaining the first advertising feature and spatiotemporal feature of the target object, the advertising cross feature can be obtained by cross-feature analysis of the first advertising feature and the spatiotemporal feature. This fully considers the interaction between the first advertising feature and the spatiotemporal feature, so that the subsequent model can learn the hidden relationship between the first advertising feature and the spatiotemporal feature through the advertising cross feature.

[0109] Specifically, after obtaining the first object feature and spatiotemporal feature of the target object, the advertising cross feature can be obtained by performing cross feature analysis on the first advertising feature and the spatiotemporal feature. The method of performing cross feature analysis on the first advertising feature and the spatiotemporal feature is similar to the method of performing cross feature analysis on the first object feature and the spatiotemporal feature in step S103 to obtain the object cross feature, and will not be described again here.

[0110] For example, in a food advertising recommendation scenario, suppose the first advertising feature obtained is a breakfast bread advertisement, and the spatiotemporal feature is noon, etc. Compared with the first advertising feature or the spatiotemporal feature, combining these two features can yield a combined feature that can further represent whether the breakfast bread advertisement is suitable to be played at noon.

[0111] Optionally, in another optional embodiment of the information push method provided in this application, the spatiotemporal features include time features and spatial features. Cross-feature analysis is performed on the first advertisement features and the spatiotemporal features to obtain advertisement cross-features, including:

[0112] By performing cross-feature analysis on the first advertising feature and the time feature, the advertising time cross-feature is obtained;

[0113] By performing cross-feature analysis on the first advertising feature and the spatial feature, the advertising spatial cross-feature is obtained.

[0114] In this embodiment, as shown in Table 2, the ad time crossover feature is used to capture the characteristics of the ad information at the current time that triggers the ad recommendation request, and the ad space crossover feature is used to capture the characteristics of the ad information in the current space that triggers the ad recommendation request.

[0115] Table 2

[0116] Feature name Feature Description Advertising time crossover characteristics The first advertising feature can be cross-referenced with different time features to capture the activity level of advertising information in different time periods. For example, the number of plays for breakfast food ads differs between the morning and noon periods, and the play duration for car rental ads differs between the lunch break and the evening periods, and so on. Cross-features of advertising space The first advertising feature can be cross-referenced with different spatial features to capture the activity level of advertising content in different locations. For example, environmental street advertisements in city A may be played more frequently at location A, but less frequently at location B compared to location A.

[0117] Specifically, such as Figure 3 As shown, after obtaining the first spatial feature and spatiotemporal feature, the first spatial feature can be cross-referenced with the time feature and spatial feature respectively to obtain the advertising time cross-feature. The advertising time cross-feature can be used to capture or statistically analyze the activity level of advertising information at the current time when the advertising recommendation request is triggered. Similarly, the advertising space cross-feature can be obtained, which can also capture or statistically analyze the activity level of advertising information in the current space when the advertising recommendation request is triggered. Moreover, the feature dimensions of the advertising time cross-feature and advertising space cross-feature obtained after feature cross-reference are increased compared to the dimensions of the independent first advertising feature and spatiotemporal feature. This allows the subsequent model to better mine and learn the timeliness or activity of advertising information through the advertising time cross-feature and advertising space cross-feature. The activity level of advertising information can be specifically represented by the playback duration or number of playbacks of advertising information, or other forms of representation, which are not specifically limited here.

[0118] For example, suppose the first ad feature is an ad introducing a scenic spot in city A, and the time feature is a holiday evening. After feature cross-referencing, we get the ad time cross-feature as an ad introducing a scenic spot in city A that is played on a holiday evening. We can use this ad time cross-feature to count that the number of plays for the ad introducing a scenic spot in city A on a holiday evening is 10. Or, suppose the first ad feature is an ad introducing a scenic spot in city A, and the spatial feature is city B. After feature cross-referencing, we get the ad spatial cross-feature as an ad introducing a scenic spot in city A that is played in city B. We can use this ad spatial cross-feature to count that the number of plays for the ad introducing a scenic spot in city A in city B is 1000.

[0119] In step S105, the matching degree between the target object and the advertising information is determined based on the first object feature, the object cross feature, the first advertisement feature, the advertisement cross feature, and the spatiotemporal feature.

[0120] In this embodiment, after obtaining the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature, feature processing can be performed on the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature respectively, and the matching degree between the target object and the advertisement information can be accurately calculated through the feature vector obtained after feature processing. The matching degree can be specifically expressed as a score or probability, and no specific limitation is made here.

[0121] Specifically, such as Figure 3 As shown, when the first object feature, object cross feature, first advertisement feature, advertisement cross feature, and spatiotemporal feature are obtained, feature processing can be performed on the first object feature, object cross feature, first advertisement feature, advertisement cross feature, and spatiotemporal feature respectively to obtain the corresponding feature vector. Specifically, a normalization algorithm or a discretization algorithm can be used, such as the word2vec model, or the glove model, etc., for feature processing, or other algorithms can be used. No specific restrictions are made here.

[0122] Furthermore, the multiple feature vectors obtained after feature processing can be learned and encoded through a network model to obtain an embedding vector that can convert discrete variables into continuous vector representations. Then, the matching degree between the target user and the advertising information can be calculated through the embedding vector. Specifically, it can be calculated using a distance formula or cosine similarity, or other calculation formulas. No specific restrictions are imposed here.

[0123] Optionally, in another optional embodiment of the information push method provided in this application, determining the matching degree between the target object and the advertising information based on the first object feature, object cross-feature, first advertisement feature, advertisement cross-feature, and spatiotemporal feature includes:

[0124] The first object features, object cross features, and spatiotemporal features are input into the first network, and the first embedding vector is output.

[0125] The first ad features, ad cross features, and spatiotemporal features are input into the second network, and the second embedding vector is output.

[0126] The matching degree is calculated based on the first embedding vector and the second embedding vector.

[0127] In this embodiment, the first network can specifically be a Convolutional Neural Network (CNN), a Deep Neural Network (DNN), or other neural networks; no specific limitation is made here. The first embedding vector is used to indicate the characteristics of the target object in the time and space that trigger the ad recommendation request. The second embedding vector is used to indicate the characteristics of the ad information in the time and space that trigger the ad recommendation request.

[0128] Specifically, such as Figure 3 As shown, when the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature are obtained, feature processing can be performed on the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature respectively to obtain the feature vectors corresponding to the first object feature, the object intersection feature, the first advertisement feature, the advertisement intersection feature and the spatiotemporal feature.

[0129] Furthermore, by adopting, such as Figure 3 The dual-tower structure of object feature-spatiotemporal feature and advertising feature-spatiotemporal feature, as shown, can concatenate the feature vectors corresponding to the first object feature, the object cross feature, and the spatiotemporal feature, and input the concatenated long vector into the first network to obtain the first embedding vector. Simultaneously, the feature vectors corresponding to the first advertising feature, the advertising cross feature, and the spatiotemporal feature can be concatenated, and the concatenated long vector can be input into the second network to obtain the second embedding vector. Then, the similarity between the first embedding vector and the second embedding vector can be calculated by using algorithms such as distance formula or cosine similarity, and this similarity represents the matching degree between the target object and the advertising information.

[0130] It is understandable that, such as Figure 4As shown, since the advertising database stores at least one piece of advertising information, similarly, N advertising features corresponding to N pieces of advertising information can be obtained according to the advertising recommendation request. Then, the N advertising cross features can be obtained by performing cross features with the first advertising feature and the spatiotemporal feature in step S104. Then, the N matching degrees between the target object and the N pieces of advertising information can be calculated based on the first object feature, the object cross features, the first advertising feature, the N advertising cross features and the spatiotemporal feature.

[0131] Optionally, in another optional embodiment of the information push method provided in this application, the first object feature, object cross feature, and spatiotemporal feature are input into the first network, and the first embedding vector is output, including:

[0132] The first object feature, object cross feature, and spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0133] The first object feature, object cross feature, and spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0134] The first ad features, ad cross features, and spatiotemporal features are input into the second network, which outputs a second embedding vector, including:

[0135] The first ad features, ad cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0136] The first ad features, ad cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0137] The matching degree is calculated based on the first embedding vector and the second embedding vector, including:

[0138] Calculate the exposure rate matching degree based on the first exposure rate embedding vector and the second exposure rate embedding vector;

[0139] The conversion rate matching degree is calculated based on the first conversion rate embedding vector and the second conversion rate embedding vector.

[0140] In this embodiment, the first network can specifically be represented as a first exposure rate network (CTR) or a first conversion rate network (CVR), or it can be represented as other networks, such as a first exposure-to-conversion rate network (CTCVR), without specific limitations. The second network can specifically be represented as a second exposure rate network or a second conversion rate network, or it can be represented as other networks, such as a second exposure-to-conversion rate network, without specific limitations. The first exposure rate embedding vector is used to represent the characteristics of the target object's click-through rate in the time and space when the ad recommendation request is triggered. The first conversion rate embedding vector is used to represent the characteristics of the target object's click-through conversion rate in the time and space when the ad recommendation request is triggered. The second exposure rate embedding vector is used to indicate the characteristics of the ad information's click-through rate in the time and space when the ad recommendation request is triggered. The second conversion rate embedding vector is used to indicate the characteristics of the ad information's click-through conversion rate in the time and space when the ad recommendation request is triggered.

[0141] Specifically, such as Figure 3 As shown, when the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature are obtained, feature processing can be performed on the first object feature, object intersection feature, first advertisement feature, advertisement intersection feature and spatiotemporal feature respectively to obtain the feature vectors corresponding to the first object feature, the object intersection feature, the first advertisement feature, the advertisement intersection feature and the spatiotemporal feature.

[0142] Furthermore, since the degree of matching between the target object and the advertising information can also be measured by a comprehensive judgment of features such as the target object's click-through rate (CTR), click-to-conversion rate (CPC), and exposure-to-conversion rate (PCPC) of the advertising information, this embodiment can use a network model trained based on CTR, CPC, and PCPC metrics to perform feature learning and obtain the embedding vectors corresponding to CTR, CPC, and PCPC metrics. Specifically, this can be achieved by concatenating the feature vectors corresponding to the first object features, the object cross features, and the spatiotemporal features, and then inputting the concatenated long vectors into the first exposure rate network and the first conversion rate network in the first network to obtain the first exposure rate embedding vector and the first conversion rate embedding vector. At the same time, the feature vectors corresponding to the first advertising features, the advertising cross features, and the spatiotemporal features can be concatenated, and the concatenated long vectors can be input into the second exposure rate network and the second conversion rate network in the second network to obtain the second exposure rate embedding vector and the second conversion rate embedding vector.

[0143] Understandably, the higher the probability that a target audience clicks on a particular advertisement, the better the match between that target audience and the advertisement. Similarly, the higher the probability that a target audience converts to a particular advertisement, the better the match. Therefore, the matching degree between the target audience and the advertisement can be calculated using the embedding vectors corresponding to metrics such as click-through rate (CTR), click-to-conversion rate (CPC), and exposure-to-conversion rate (PCP). Specifically, the similarity between the first PCP embedding vector and the second PCP embedding vector can be calculated using algorithms such as distance formulas or cosine similarity. This similarity represents the matching degree between the target audience and the advertisement under the CTR metric. Similarly, the similarity between the first PCP embedding vector and the second PCP embedding vector can be calculated to represent the matching degree between the target audience and the advertisement under the PCP metric.

[0144] In step S106, advertising information is pushed to the terminal device based on the matching degree.

[0145] Specifically, after obtaining the matching degree between the target object and the advertising information, the matching degree between the target object and the advertising information can be compared with a preset matching degree threshold to further determine whether the target object is suitable for the advertising information. This allows the advertising information to be pushed to the target object more accurately. The preset matching degree threshold is set according to the actual application scenario and is not specifically limited here.

[0146] Furthermore, if the matching degree between the target object and the advertising information is greater than the preset matching degree threshold, it can be understood that the advertising information is suitable for the target object, and the advertising information can be pushed to the terminal device. Conversely, if the matching degree is less than the target object, the advertising information is not suitable for the target object, and the advertising information will not be pushed to the terminal device.

[0147] For example, assuming that the match between target A and the advertisement introducing mobile phone performance is 0.86, and the preset match threshold is 0.6, it can be seen that target A is a good match for the advertisement introducing mobile phone performance, and the advertisement can be pushed to the terminal device used by target A.

[0148] It is understandable that, such as Figure 4As shown, after obtaining N matching scores between the target object and N advertising information, the matching scores that meet the preset matching score threshold can be obtained first. These matching scores can then be sorted in ascending or descending order to obtain the advertising information with the highest matching score to the target object. Alternatively, after obtaining N matching scores between the target object and N advertising information, the N matching scores can be compared pairwise to obtain the matching score with the largest value. If the matching score with the largest value is greater than the preset matching score threshold, it can be understood that the advertising information corresponding to the matching score with the largest value is suitable for the target object, and the advertising information can be pushed to the terminal device.

[0149] Furthermore, after testing, the following methods were adopted: Figure 3 as well as Figure 4 The dual-tower structure of object features-spatiotemporal features and advertising features-spatiotemporal features for advertising information processing can reduce advertising concentration by about 10% and increase advertising information consumption by about 4.73%. This can effectively reduce advertising concentration and increase advertising consumption, thereby increasing advertising exposure and reducing advertising costs to a certain extent. In turn, it can improve the performance of the advertising system and enhance the user experience.

[0150] Optionally, in another optional embodiment of the information push method provided in this application, pushing advertising information to the terminal device based on the matching degree includes:

[0151] Calculate the target matching degree based on the exposure rate matching degree and the conversion rate matching degree;

[0152] If the target match is greater than the match threshold, then the advertising information is pushed to the terminal device.

[0153] Specifically, after obtaining the exposure rate matching degree and conversion rate matching degree of the target object and the advertising information under different indicators, the exposure rate matching degree and conversion rate matching degree can be jointly calculated according to the principle of multiple branches working together when the ANN algorithm triggers branches. Specifically, it can be calculated by summing according to the preset indicator weights, or by calculating the median or standard deviation, etc., to obtain the target matching degree of the target object and the advertising information under the comprehensive indicators. Then, by comparing the target matching degree of the target object and the advertising information with the preset matching degree threshold, it can be further determined whether the target object is suitable for the advertising information, so as to push the advertising information to the target object more accurately.

[0154] Furthermore, if the target audience and the advertising information match a preset matching threshold, it can be understood that the advertising information is suitable for the target audience, and the advertising information can be pushed to the terminal device. Conversely, if the target audience and the advertising information do not match, the advertising information will not be pushed to the terminal device.

[0155] For example, assuming that the target audience A has an exposure rate match of 0.6 and a conversion rate match of 0.8 with the advertisement introducing mobile phone performance, with an exposure rate weight of 0.64 and a conversion rate weight of 0.36, and the preset match threshold is 0.56, it can be seen that the target match between target audience A and the advertisement introducing mobile phone performance is 0.672, which is greater than the preset match threshold. That is, target audience A and the advertisement introducing mobile phone performance are compatible, and the advertisement introducing mobile phone performance can be pushed to the terminal device used by target audience A.

[0156] It is understandable that, such as Figure 4 As shown, after obtaining N exposure rate matching degrees and N conversion rate matching degrees between the target object and N advertising information, we can first obtain N target matching degrees between the target object and N advertising information. Then, we can compare the N target matching degrees pairwise to obtain the target matching degree with the largest value. If the target matching degree with the largest value is greater than the preset matching degree threshold, it can be understood that the advertising information corresponding to the target object is suitable for the target object, and the advertising information can be pushed to the terminal device.

[0157] In this application embodiment, an information push method is provided. By performing feature cross-referencing on the first object feature and the first advertisement feature with spatiotemporal features, the spatiotemporal features can be combined to simultaneously apply to the target object and advertisement information, thereby obtaining more complex and targeted object cross-features and spatial cross-features. This allows the model to learn more complex cross-features to recall more targeted advertisement information, thereby reducing the concentration of advertisement information and improving the accuracy of obtaining advertisement information.

[0158] Optionally, in the above Figure 3 Based on the corresponding embodiments, in another optional embodiment of the advertising information push method provided in this application, before determining the matching degree between the target object and the advertising information based on the first object features, object cross features, first advertising features, advertising cross features, and spatiotemporal features, the method further includes:

[0159] Based on the advertising recommendation request, obtain the historical spatiotemporal features corresponding to the object identifier;

[0160] Cross-feature analysis is performed on the first object features and historical spatiotemporal features to obtain the object's historical cross-features;

[0161] By performing cross-feature analysis on the first advertising feature and the historical spatiotemporal feature, the advertising historical cross-feature is obtained;

[0162] The matching degree between the target object and the advertising information is determined based on the first object feature, object cross feature, object historical cross feature, first advertising feature, advertising cross feature, advertising historical cross feature, spatiotemporal feature, and historical spatiotemporal feature.

[0163] Specifically, since the target object may exhibit different interests or preferences in different time periods and spaces in the past, in order to further reflect the differences in advertising recommendation requests under different times and spaces, and thus obtain advertising information that is more suitable for the advertising recommendation request, and achieve more accurate push of advertising information, after obtaining the first object feature and spatiotemporal feature of the target object, this embodiment can also obtain the historical spatiotemporal feature corresponding to the object identifier according to the advertising recommendation request.

[0164] Furthermore, historical cross features of an object can be obtained by cross-feature analysis of the first object features and historical spatiotemporal features. This takes into account the interaction between the first object features and historical spatiotemporal features, so that subsequent models can learn the hidden interaction relationship between the first object features and historical spatiotemporal features through the historical cross features of the object. Similarly, since advertising information can exhibit different activity levels in different time periods and spaces in the past, historical cross features of advertising can be obtained by cross-feature analysis of the first advertising features and historical spatiotemporal features. This takes into account the interaction between the first advertising features and historical spatiotemporal features, so that subsequent models can learn the hidden interaction relationship between the first advertising features and historical spatiotemporal features through the advertising cross features.

[0165] Optionally, in another optional embodiment of the information push method provided in this application, the historical spatiotemporal features include historical time features and historical space features. Cross-feature analysis is performed on the first object features and the historical spatiotemporal features to obtain object historical cross-features, including:

[0166] Cross features are obtained by combining the first object feature with the historical time feature;

[0167] Cross features are obtained by combining the first object features with the historical space features.

[0168] In this embodiment, as shown in Table 3, historical spatiotemporal features include historical time features and historical spatial features. Historical time features refer to past time periods before the current time that triggers the ad recommendation request, such as the previous half hour, one hour, two hours, or half a day. Historical spatial features refer to the user's location during the past time periods before the current time that triggers the ad recommendation request. Object historical time intersection features represent the behavioral characteristics of the target object during the past time periods before the current time that triggers the ad recommendation request. That is, within different time periods before the current time that triggers the ad recommendation request, such as the previous half hour, one hour, two hours, or half a day, the user performs certain actions through the terminal device, such as browsing, clicking, converting, or saving. Object historical spatial intersection features represent the behavioral characteristics of the user in the same space during the past time periods before the current time that triggers the ad recommendation request. That is, within the past time periods before the current time that triggers the ad recommendation request, the user records certain behaviors performed by the user at the current location through the terminal device, such as browsing records, click records, and conversion records.

[0169] Table 3

[0170] Feature name Feature Description Historical time characteristics This refers to a period of time preceding the current time that triggers the ad recommendation request. The meaning of the ad recommendation request can vary significantly depending on the user's actions at different times. Historical spatial characteristics This refers to the user's location within a certain period prior to the time that triggers the ad recommendation request. Within the same location, different user actions at different times will result in different meanings expressed in the ad recommendation request, and the target of the user's interest will also be different. Object historical time intersection features This refers to the actions a user took in the period leading up to the time when the ad recommendation request was triggered. For example, actions taken by the user in different time periods such as half an hour, an hour, two hours, or half a day before a specific point in time, including browsing, clicking, converting, or saving. Object historical space intersection features This refers to a user's historical actions at a specific location. For example, a user's browsing history, click history, and conversion history at location A (last half hour, last two hours, etc.). It not only reflects the spatial nature of a location but also its temporal relationship.

[0171] Specifically, such as Figure 3 As shown, after obtaining the first object feature and historical spatiotemporal features, by performing feature cross-interaction between the first object feature and the historical time feature and historical space feature respectively, we can obtain object historical time cross-features that can further express the behavioral characteristics of the target object in the past time period before the current time that triggers the ad recommendation request, and object space cross-features that can further express the behavioral characteristics of the target object in the current space in the past time period before the current time that triggers the ad recommendation request. This allows the subsequent model to better mine and learn the target object's preferences or interests in the historical time period through object historical time cross-features and object historical space cross-features, and to reflect the characteristics and differences of ad recommendation requests in different historical time periods and contexts through object historical time cross-features and object historical space cross-features, thereby enabling the recall of more targeted ad information for ad recommendation requests.

[0172] For example, assuming the first object feature is user identity such as a student, and the historical time feature is from 9:00 AM to 11:00 AM on Saturday, the cross-feature result is that the object's historical time cross-feature is "students from 9:00 AM to 11:00 AM on Saturday," which can be used to statistically analyze the browsing, clicking, conversion, or favorites performed by students through terminal devices from 9:00 AM to 11:00 AM on Saturday. Alternatively, assuming the first object feature is user identity such as a student, and the historical spatial feature is "at home from 9:00 AM to 11:00 AM on Saturday," the cross-feature result is that the object's historical spatial cross-feature is "students at home from 9:00 AM to 11:00 AM on Saturday," which can be used to statistically analyze the browsing, clicking, and conversion records generated by students at home on terminal devices from 9:00 AM to 11:00 AM on Saturday.

[0173] Optionally, in another optional embodiment of the information push method provided in this application, the historical spatiotemporal features include historical time features and historical space features. Cross-feature analysis is performed on the first advertising features and the historical spatiotemporal features to obtain advertising historical cross-features, including:

[0174] By performing cross-feature analysis on the first advertising feature and the historical time feature, the advertising historical time cross-feature is obtained;

[0175] Cross-feature analysis is performed on the first advertising feature and the historical space feature to obtain the advertising historical space cross-feature.

[0176] In this embodiment, as shown in Table 4, the advertising history time intersection feature is used to capture the characteristics of advertising information during a period of time before the current time that triggers the advertising recommendation request. The advertising history spatial intersection feature is used to capture the characteristics of advertising information in a certain space during a period of time before the current time that triggers the advertising recommendation request.

[0177] Table 4

[0178] Feature name Feature Description Advertising historical time crossover characteristics The first advertising feature can be cross-referenced with different historical time features to capture the activity level of advertising information in the period before the current time that triggers the advertising recommendation request. Advertising historical spatial cross features The first advertising feature can be cross-referenced with different historical spatial features to capture the activity level of ad content at the same location over a period of time prior to the current time that triggered the ad recommendation request. For example, street environmental ads in city A are played more frequently at location A between 4 PM and 6 PM, and less frequently between 2 PM and 3 PM.

[0179] Specifically, such as Figure 3 As shown, after obtaining the first spatial feature and historical spatiotemporal feature, the first spatial feature can be cross-referenced with the historical time feature and historical spatial feature respectively to obtain the advertising historical time cross-feature. The advertising historical time cross-feature can be used to capture or statistically analyze the activity level of advertising information in the period before the current time that triggers the advertising recommendation request. Similarly, the advertising historical spatial cross-feature can be obtained, and the advertising spatial cross-feature can be used to capture or statistically analyze the activity level of advertising information in the same location in the current space that triggers the advertising recommendation request. This allows the subsequent model to better mine and learn the timeliness or activity of advertising information through the advertising time cross-feature and advertising spatial cross-feature.

[0180] For example, suppose the first ad feature is an ad introducing scenic spots in city A, and the historical time feature is 7:20 PM on holidays. After feature cross-referencing, the ad's historical time cross-feature is that the ad introducing scenic spots in city A was played at 7:20 PM on holidays. This historical time cross-feature can be used to count 500 plays of the ad introducing scenic spots in city A at 7:20 PM on holidays. Or, suppose the first ad feature is an ad introducing scenic spots in city A, and the historical spatial feature is that the ad was played at 7:20 PM on holidays in city B. After feature cross-referencing, the ad's historical spatial cross-feature is that the ad introducing scenic spots in city A was played at 7:20 PM on holidays in city B. This historical spatial cross-feature can be used to count 2000 plays of the ad introducing scenic spots in city A at 7:20 PM on holidays in city B.

[0181] Optionally, in another optional embodiment of the information push method provided in this application, the first object feature, object cross feature, and spatiotemporal feature are input into the first network, and the first embedding vector is output, including:

[0182] Input the first object features, object cross features, object history cross features, and spatiotemporal features into the first network, and output the first embedding vector.

[0183] The first ad features, ad cross features, and spatiotemporal features are input into the second network, which outputs a second embedding vector, including:

[0184] The first ad feature, ad cross feature, ad history cross feature, and spatiotemporal feature are input into the second network, and the second embedding vector is output.

[0185] Specifically, such as Figure 3 As shown, when the first object feature, object cross feature, object historical cross feature, first advertisement feature, advertisement cross feature, advertisement historical cross feature, spatiotemporal feature, and historical spatiotemporal feature are obtained, feature processing can be performed on the first object feature, object cross feature, object historical cross feature, first advertisement feature, advertisement cross feature, advertisement historical cross feature, spatiotemporal feature, and historical spatiotemporal feature respectively. This will yield the feature vectors corresponding to the first object feature, the object cross feature, the object historical cross feature, the first advertisement feature, the advertisement cross feature, the advertisement historical cross feature, the spatiotemporal feature, and the historical spatiotemporal feature.

[0186] Furthermore, a long vector can be obtained by concatenating the feature vectors corresponding to the first object features, the object cross features, the object historical cross features, the spatiotemporal features, and the historical spatiotemporal features. Compared to concatenating the feature vectors corresponding to the first object features, the object cross features, and the spatiotemporal features, this long vector has increased dimensionality and yields more complex features. Then, the concatenated long vector is input into the first network to obtain a first embedding vector with richer information content. Simultaneously, the feature vectors corresponding to the first advertisement features, the advertisement cross features, and the historical spatiotemporal features can be used to obtain a long vector. The feature vectors corresponding to the historical cross features of the advertisement, the feature vectors corresponding to the spatiotemporal features, and the feature vectors corresponding to the historical spatiotemporal features are concatenated, and the concatenated long vector is input into the second network to obtain a second embedding vector with richer information content. Then, the similarity between the first embedding vector and the second embedding vector with richer information content can be calculated by using algorithms such as distance formula or cosine similarity, so as to obtain a more accurate similarity. This more accurate similarity can then represent a more precise matching degree between the target object and the advertisement information, so that the advertisement information adapted to the target object can be accurately obtained based on the matching degree between the target object and the advertisement information, thereby achieving precise delivery of advertisement information.

[0187] Optionally, in another optional embodiment of the information push method provided in this application, the first object feature, object cross feature, and spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output, including:

[0188] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0189] The first object feature, object cross feature, and spatiotemporal feature are input into the first transformation rate network in the first network, and the first transformation rate embedding vector is output, including:

[0190] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0191] The first ad features, ad cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output, including:

[0192] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0193] The first ad features, ad cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the output is the second conversion rate embedding vector, including:

[0194] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0195] Specifically, such as Figure 3 As shown, when the first object feature, object cross feature, object historical cross feature, first advertisement feature, advertisement cross feature, advertisement historical cross feature, spatiotemporal feature, and historical spatiotemporal feature are obtained, feature processing can be performed on the first object feature, object cross feature, object historical cross feature, first advertisement feature, advertisement cross feature, advertisement historical cross feature, spatiotemporal feature, and historical spatiotemporal feature respectively. This will yield the feature vectors corresponding to the first object feature, the object cross feature, the object historical cross feature, the first advertisement feature, the advertisement cross feature, the advertisement historical cross feature, the spatiotemporal feature, and the historical spatiotemporal feature.

[0196] Furthermore, the long vector obtained by concatenating the feature vectors corresponding to the first object features, the object cross features, the object historical cross features, the spatiotemporal features, and the historical spatiotemporal features has an increased vector dimension compared to the long vector obtained by concatenating the feature vectors corresponding to the first object features, the object cross features, and the spatiotemporal features. This concatenated long vector is then input into the first exposure rate network and the first conversion rate network within the first network to obtain first exposure rate embedding vectors and first conversion rate embedding vectors with richer information content. Simultaneously, the feature vectors corresponding to the first advertisement features, the advertisement cross features, the advertisement historical cross features, the spatiotemporal features, and the historical spatiotemporal features are concatenated, and the resulting long vector is input into the second network to obtain second exposure rate embedding vectors and second conversion rate embedding vectors with richer information content. This allows for a more accurate similarity score to be obtained by comparing the similarity scores of the first and second embedding vectors with richer information content, thus representing a more precise match between the target object and the advertisement information.

[0197] Optionally, in the above Figure 3Based on the corresponding embodiments, in another optional embodiment of the information push method provided in this application, before determining the matching degree between the target object and the advertising information based on the first object features, object cross features, first advertising features, advertising cross features, and spatiotemporal features, the method further includes:

[0198] Based on the ad recommendation request, obtain the associated object characteristics corresponding to the object identifier;

[0199] Cross features are obtained by combining the features of the first object with the features of the associated objects;

[0200] By performing cross-feature analysis on the first advertisement feature and the associated object feature, the advertisement association cross-feature is obtained;

[0201] The matching degree between the target object and the advertising information is determined based on the first object feature, object cross feature, object association cross feature, first advertisement feature, advertisement cross feature, advertisement association cross feature, spatiotemporal feature, and associated object feature.

[0202] Specifically, since the target object contacts different objects or conducts product transactions with merchants, it can exhibit different personalities or preferences, thereby further reflecting the differences in advertising recommendation requests under different associated objects. This allows for the acquisition of advertising information that is more suitable for the advertising recommendation request, achieving more accurate delivery of advertising information. Therefore, after obtaining the first object characteristics and spatiotemporal characteristics of the target object, this embodiment can also obtain the associated object characteristics corresponding to the object identifier according to the advertising recommendation request. The associated object characteristics can specifically be the associated object's occupation, gender, or place of residence, or other object characteristics, without specific limitations here.

[0203] Furthermore, object-related cross features can be obtained by cross-feature analysis of the first object features and the associated object features. This takes into account the interaction between the first object features and the associated object features, so that subsequent models can learn the hidden interaction relationship between the first object features and the associated object features through object-related cross features. Similarly, since advertising information can exhibit different levels of activity when targeting different associated objects, advertising-related cross features can be obtained by cross-feature analysis of the first advertising features and the associated object features. This takes into account the interaction between the first advertising features and the associated object features, so that subsequent models can learn the hidden interaction relationship between the first advertising features and the object-related features through advertising-related features.

[0204] The information push device in this application is described in detail below. Please refer to [link / reference]. Figure 5 , Figure 5 This is a schematic diagram of one embodiment of the advertising push device in this application. The information push device 20 includes:

[0205] The acquisition unit 201 is used to receive an advertising recommendation request sent by the terminal device, wherein the advertising recommendation request carries an object identifier, and the object identifier is used to uniquely identify the target object;

[0206] The acquisition unit 201 is also used to acquire the first advertising feature of the advertising information, the first object feature of the target object, and the spatiotemporal feature of the advertising recommendation request according to the advertising recommendation request. The spatiotemporal feature is used to describe the time information and space information that triggered the advertising recommendation request.

[0207] Processing unit 202 is used to perform cross-feature analysis on the first object features and spatiotemporal features to obtain object cross-features;

[0208] The processing unit 202 is also used to perform cross-feature analysis on the first advertising feature and the spatiotemporal feature to obtain advertising cross-features;

[0209] The determining unit 203 is used to determine the matching degree between the target object and the advertising information based on the first object features, object cross features, first advertising features, advertising cross features and spatiotemporal features;

[0210] The processing unit 202 is also used to push advertising information to the terminal device based on the matching degree.

[0211] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the processing unit 202 may specifically be used for:

[0212] Cross features are obtained by combining the first object features with the time features;

[0213] Cross features are obtained by combining the first object features with the spatial features.

[0214] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the processing unit 202 may specifically be used for:

[0215] By performing cross-feature analysis on the first advertising feature and the time feature, the advertising time cross-feature is obtained;

[0216] By performing cross-feature analysis on the first advertising feature and the spatial feature, the advertising spatial cross-feature is obtained.

[0217] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application,

[0218] The acquisition unit 201 is also used to acquire the historical spatiotemporal features corresponding to the object identifier based on the advertising recommendation request;

[0219] The processing unit 202 is also used to perform cross-feature analysis on the first object features and historical spatiotemporal features to obtain object historical cross-features;

[0220] The processing unit 202 is also used to perform cross-feature analysis on the first advertising feature and the historical spatiotemporal feature to obtain advertising historical cross-feature.

[0221] Specifically, the determining unit 203 can be used to: determine the matching degree between the target object and the advertising information based on the first object feature, object cross feature, object historical cross feature, first advertising feature, advertising cross feature, advertising historical cross feature, spatiotemporal feature, and historical spatiotemporal feature.

[0222] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the processing unit 202 may specifically be used for:

[0223] Cross features are obtained by combining the first object feature with the historical time feature;

[0224] Cross features are obtained by combining the first object features with the historical space features.

[0225] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the processing unit 202 may specifically be used for:

[0226] By performing cross-feature analysis on the first advertising feature and the historical time feature, the advertising historical time cross-feature is obtained;

[0227] Cross-feature analysis is performed on the first advertising feature and the historical space feature to obtain the advertising historical space cross-feature.

[0228] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the advertising information push device provided in this application, the determining unit 203 may specifically be used for:

[0229] The first object features, object cross features, and spatiotemporal features are input into the first network, and the first embedding vector is output.

[0230] The first ad features, ad cross features, and spatiotemporal features are input into the second network, and the second embedding vector is output.

[0231] The matching degree is calculated based on the first embedding vector and the second embedding vector.

[0232] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the determining unit 203 may specifically be used for:

[0233] The first object feature, object cross feature, and spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0234] The first object feature, object cross feature, and spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0235] The first ad features, ad cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0236] The first ad features, ad cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0237] The exposure rate matching degree is calculated based on the first exposure rate embedding vector and the obtained second exposure rate embedding vector.

[0238] The conversion rate matching degree is calculated based on the first conversion rate embedding vector and the obtained second conversion rate embedding vector.

[0239] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the determining unit 203 may specifically be used for:

[0240] Input the first object features, object cross features, object history cross features, and spatiotemporal features into the first network, and output the first embedding vector.

[0241] The first ad feature, ad cross feature, ad history cross feature, and spatiotemporal feature are input into the second network, and the second embedding vector is output.

[0242] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the advertising information push device provided in this application, the determining unit 203 may specifically be used for:

[0243] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output.

[0244] The first object features, object cross features, object historical cross features, and spatiotemporal features are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output.

[0245] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output.

[0246] The first ad features, ad cross features, ad history cross features, and spatiotemporal features are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

[0247] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application, the processing unit 202 may specifically be used for:

[0248] Calculate the target matching degree based on the exposure rate matching degree and the conversion rate matching degree;

[0249] If the target match is greater than the match threshold, then the advertising information is pushed to the terminal device.

[0250] Optionally, in the above Figure 5 Based on the corresponding embodiments, in another embodiment of the information push device provided in this application,

[0251] The acquisition unit 201 is also used to acquire the associated object features corresponding to the object identifier based on the advertising recommendation request;

[0252] The processing unit 202 is also used to perform cross-feature analysis on the first object features and the associated object features to obtain object association cross features;

[0253] The processing unit 202 is also used to perform cross-feature analysis on the first advertisement feature and the associated object feature to obtain the advertisement association cross-feature;

[0254] Specifically, the determining unit 203 can be used to: determine the matching degree between the target object and the advertising information based on the first object features, object cross features, object association cross features, first advertising features, advertising cross features, advertising association cross features, spatiotemporal features, and associated object features.

[0255] This application also provides a schematic diagram of another computer device, such as... Figure 6 As shown, Figure 6This is a schematic diagram of a computer device structure provided in an embodiment of this application. The computer device 300 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 331 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the computer device 300. Furthermore, the CPU 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the computer device 300.

[0256] The computer device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 333, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0257] The aforementioned computer device 300 is also used to perform, for example Figure 2 The steps in the corresponding embodiments.

[0258] Another aspect of this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform actions such as... Figure 2 The steps in the method described in the illustrated embodiment.

[0259] Another aspect of this application provides a computer program product containing instructions that, when run on a computer or processor, cause the computer or processor to perform actions such as Figure 2 The steps in the method described in the illustrated embodiment.

[0260] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0261] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0262] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0263] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0264] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method of information push, characterized by, include: The device receives an advertising recommendation request sent by a terminal device, wherein the advertising recommendation request carries an object identifier, which is used to uniquely identify the target object; The first advertising feature of the advertising information, the first object feature of the target object, and the spatiotemporal feature of the advertising recommendation request are obtained according to the advertising recommendation request, wherein the spatiotemporal feature is used to describe the time information and spatial information that trigger the advertising recommendation request; The first object feature and the spatiotemporal feature are cross-featured to obtain the object cross feature; The first advertising feature and the spatiotemporal feature are cross-featured to obtain the advertising cross-feature; Based on the first object features, the object cross features, the first advertisement features, the advertisement cross features, and the spatiotemporal features, the matching degree between the target object and the advertisement information is determined; The advertising information is pushed to the terminal device based on the matching degree.

2. The method of claim 1, wherein, The spatiotemporal features include temporal features and spatial features; The step of performing cross-feature analysis on the first object feature and the spatiotemporal feature to obtain object cross-features includes: The first object feature and the time feature are cross-featured to obtain the object time cross feature; The first object feature and the spatial feature are cross-featured to obtain the object spatial cross feature.

3. The method of claim 1, wherein, The spatiotemporal features include temporal features and spatial features; The step of performing cross-feature analysis on the first advertising feature and the spatiotemporal feature to obtain advertising cross-features includes: By performing cross-feature analysis on the first advertising feature and the time feature, an advertising time cross-feature is obtained; The advertising spatial cross feature is obtained by performing cross feature analysis on the first advertising feature and the spatial feature.

4. The method of claim 1, wherein, Before determining the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertisement features, the advertisement cross features, and the spatiotemporal features, the method further includes: Based on the advertising recommendation request, obtain the historical spatiotemporal features corresponding to the object identifier; The first object feature and the historical spatiotemporal feature are cross-featured to obtain the object historical cross feature; The first advertising feature and the historical spatiotemporal feature are cross-featured to obtain the advertising history cross feature; Determining the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertising features, the advertising cross features, and the spatiotemporal features includes: The matching degree between the target object and the advertising information is determined based on the first object feature, the object cross feature, the object historical cross feature, the first advertisement feature, the advertisement cross feature, the advertisement historical cross feature, the spatiotemporal feature, and the historical spatiotemporal feature.

5. The method of claim 4, wherein, The historical spatiotemporal characteristics include historical temporal characteristics and historical spatial characteristics; The step of performing cross-feature analysis on the first object features and the historical spatiotemporal features to obtain object historical cross-features includes: By performing cross-feature analysis on the first object feature and the historical time feature, the object historical time cross-feature is obtained; Cross features are obtained by combining the first object features with the historical space features.

6. The method according to claim 4, characterized in that, The historical spatiotemporal characteristics include historical temporal characteristics and historical spatial characteristics; The step of performing cross-feature analysis on the first advertising feature and the historical spatiotemporal feature to obtain advertising historical cross-features includes: The first advertising feature and the historical time feature are cross-featured to obtain the advertising historical time cross feature; The advertising historical space cross feature is obtained by performing cross feature analysis on the first advertising feature and the historical space feature.

7. The method according to any one of claims 4 to 6, characterized in that, Determining the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertising features, the advertising cross features, and the spatiotemporal features includes: The first object feature, the object intersection feature, and the spatiotemporal feature are input into the first network, and the first embedding vector is output. The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second network, and the second embedding vector is output. The matching degree is calculated based on the first embedding vector and the second embedding vector.

8. The method according to claim 7, characterized in that, The step of inputting the first object feature, the object cross feature, and the spatiotemporal feature into the first network and outputting the first embedding vector includes: The first object feature, the object cross feature, and the spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output. The first object feature, the object cross feature, and the spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output. The step of inputting the first advertisement feature, the advertisement cross feature, and the spatiotemporal feature into the second network and outputting the second embedding vector includes: The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output. The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output. The step of calculating the matching degree based on the first embedding vector and the second embedding vector includes: Calculate the exposure rate matching degree based on the first exposure rate embedding vector and the second exposure rate embedding vector; The conversion rate matching degree is calculated based on the first conversion rate embedding vector and the second conversion rate embedding vector.

9. The method according to claim 7, characterized in that, The step of inputting the first object feature, the object cross feature, and the spatiotemporal feature into the first network and outputting the first embedding vector includes: The first object feature, the object cross feature, the object history cross feature, and the spatiotemporal feature are input into the first network, and the first embedding vector is output. The step of inputting the first advertisement feature, the advertisement cross feature, and the spatiotemporal feature into the second network and outputting the second embedding vector includes: The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second network, and the second embedding vector is output.

10. The method according to claim 8, characterized in that, The step of inputting the first object feature, the object cross feature, and the spatiotemporal feature into the first exposure rate network in the first network, and outputting the first exposure rate embedding vector, includes: The first object feature, the object cross feature, the object historical cross feature, and the spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output. The step of inputting the first object feature, the object cross feature, and the spatiotemporal feature into the first conversion rate network in the first network, and outputting the first conversion rate embedding vector, includes: The first object feature, the object cross feature, the object historical cross feature, and the spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output. The step of inputting the first ad feature, the ad cross feature, and the spatiotemporal feature into the second exposure rate network in the second network, and outputting the second exposure rate embedding vector, includes: The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output. The step of inputting the first ad feature, the ad cross feature, and the spatiotemporal feature into the second conversion rate network in the second network, and outputting the second conversion rate embedding vector, includes: The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

11. The method according to claim 8, characterized in that, The step of pushing the advertising information to the terminal device based on the matching degree includes: Calculate the target matching degree based on the exposure rate matching degree and the conversion rate matching degree; If the target matching degree is greater than the matching degree threshold, the advertising information is pushed to the terminal device.

12. The method according to claim 1, characterized in that, Before determining the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertisement features, the advertisement cross features, and the spatiotemporal features, the method further includes: Based on the advertising recommendation request, obtain the associated object features corresponding to the object identifier; Cross features are obtained by combining the features of the first object with the features of the associated object; The first advertisement feature and the associated object feature are cross-featured to obtain the advertisement association cross feature; Determining the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertising features, the advertising cross features, and the spatiotemporal features includes: The matching degree between the target object and the advertising information is determined based on the first object feature, the object cross feature, the object association cross feature, the first advertisement feature, the advertisement cross feature, the advertisement association cross feature, the spatiotemporal feature, and the associated object feature.

13. An information push device, characterized in that, include: The acquisition unit is used to receive an advertising recommendation request sent by a terminal device, wherein the advertising recommendation request carries an object identifier, and the object identifier is used to uniquely identify a target object; The acquisition unit is further configured to acquire, based on the ad recommendation request, a first ad feature of the ad information, a first object feature of the target object, and a spatiotemporal feature of the ad recommendation request, wherein the spatiotemporal feature is used to describe the time and space information that triggered the ad recommendation request; The processing unit is used to perform cross-feature analysis on the first object feature and the spatiotemporal feature to obtain object cross-features; The processing unit is further configured to perform cross-feature analysis on the first advertising feature and the spatiotemporal feature to obtain advertising cross-features; The determining unit is configured to determine the matching degree between the target object and the advertising information based on the first object features, the object cross features, the first advertising features, the advertising cross features, and the spatiotemporal features; The processing unit is also configured to push the advertising information to the terminal device based on the matching degree.

14. The apparatus according to claim 13, characterized in that, The spatiotemporal features include temporal features and spatial features; the processing unit is specifically used for: The first object feature and the time feature are cross-featured to obtain the object time cross feature; The first object feature and the spatial feature are cross-featured to obtain the object spatial cross feature.

15. The apparatus according to claim 13, characterized in that, The spatiotemporal features include temporal features and spatial features; the processing unit is specifically used for: By performing cross-feature analysis on the first advertising feature and the time feature, an advertising time cross-feature is obtained; The advertising spatial cross feature is obtained by performing cross feature analysis on the first advertising feature and the spatial feature.

16. The apparatus according to claim 13, characterized in that, Before determining the matching degree between the target object and the advertising information based on the first object feature, the object cross feature, the first advertising feature, the advertising cross feature, and the spatiotemporal feature, the acquisition unit is further configured to acquire the historical spatiotemporal feature corresponding to the object identifier based on the advertising recommendation request; The processing unit is further configured to: The first object feature and the historical spatiotemporal feature are cross-featured to obtain the object historical cross feature; The first advertising feature and the historical spatiotemporal feature are cross-featured to obtain the advertising history cross feature; The determining unit is specifically used to determine the matching degree between the target object and the advertising information based on the first object feature, the object cross feature, the object historical cross feature, the first advertisement feature, the advertisement cross feature, the advertisement historical cross feature, the spatiotemporal feature, and the historical spatiotemporal feature.

17. The apparatus according to claim 16, characterized in that, The historical spatiotemporal features include historical temporal features and historical spatial features; the processing unit is specifically used for: By performing cross-feature analysis on the first object feature and the historical time feature, the object historical time cross-feature is obtained; Cross features are obtained by combining the first object features with the historical space features.

18. The apparatus according to claim 16, characterized in that, The historical spatiotemporal features include historical temporal features and historical spatial features; the processing unit is specifically used for: The first advertising feature and the historical time feature are cross-featured to obtain the advertising historical time cross feature; The advertising historical space cross feature is obtained by performing cross feature analysis on the first advertising feature and the historical space feature.

19. The apparatus according to any one of claims 16 to 18, characterized in that, The determining unit is specifically used for: The first object feature, the object intersection feature, and the spatiotemporal feature are input into the first network, and the first embedding vector is output. The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second network, and the second embedding vector is output. The matching degree is calculated based on the first embedding vector and the second embedding vector.

20. The apparatus according to claim 19, characterized in that, The determining unit is specifically used for: The first object feature, the object cross feature, and the spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output. The first object feature, the object cross feature, and the spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output. The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output. The first ad feature, the ad cross feature, and the spatiotemporal feature are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output. Calculate the exposure rate matching degree based on the first exposure rate embedding vector and the second exposure rate embedding vector; The conversion rate matching degree is calculated based on the first conversion rate embedding vector and the second conversion rate embedding vector.

21. The apparatus according to claim 19, characterized in that, The determining unit is specifically used for: The first object feature, the object cross feature, the object history cross feature, and the spatiotemporal feature are input into the first network, and the first embedding vector is output. The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second network, and the second embedding vector is output.

22. The apparatus according to claim 20, characterized in that, The determining unit is specifically used for: The first object feature, the object cross feature, the object historical cross feature, and the spatiotemporal feature are input into the first exposure rate network in the first network, and the first exposure rate embedding vector is output. The first object feature, the object cross feature, the object historical cross feature, and the spatiotemporal feature are input into the first conversion rate network in the first network, and the first conversion rate embedding vector is output. The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second exposure rate network in the second network, and the second exposure rate embedding vector is output. The first ad feature, the ad cross feature, the ad history cross feature, and the spatiotemporal feature are input into the second conversion rate network in the second network, and the second conversion rate embedding vector is output.

23. The apparatus according to claim 20, characterized in that, The processing unit is specifically used for: Calculate the target matching degree based on the exposure rate matching degree and the conversion rate matching degree; If the target matching degree is greater than the matching degree threshold, the advertising information is pushed to the terminal device.

24. The apparatus according to claim 13, characterized in that, Before determining the matching degree between the target object and the advertising information based on the first object feature, the object cross feature, the first advertisement feature, the advertisement cross feature, and the spatiotemporal feature, the acquisition unit is further configured to acquire the associated object feature corresponding to the object identifier based on the advertisement recommendation request; The processing unit is further configured to perform cross-feature analysis on the first object feature and the associated object feature to obtain object association cross-features; The processing unit is further configured to perform cross-feature analysis on the first advertisement feature and the associated object feature to obtain advertisement association cross-features; The determining unit is specifically used to determine the matching degree between the target object and the advertising information based on the first object feature, the object cross feature, the object association cross feature, the first advertisement feature, the advertisement cross feature, the advertisement association cross feature, the spatiotemporal feature, and the associated object feature.

25. A computer device, characterized in that, include: Memory, transceiver, processor, and bus system; The memory is used to store programs; When the processor executes a program in the memory, it implements the method as described in any one of claims 1 to 12; The bus system is used to connect the memory and the processor to enable communication between the memory and the processor.

26. A computer-readable storage medium comprising instructions that, when executed on a computer, cause the computer to perform the method as claimed in any one of claims 1 to 12.

27. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method described in any one of claims 1 to 12.

Images