An information recommendation method

Abstract

The application provides an information recommendation method and device, electronic equipment and a computer readable storage medium. The method comprises the following steps: presenting an information stream in a human-computer interaction interface; determining the influence degree when inserting recommendation information in multiple insertion positions in the information stream; determining a target insertion position from the multiple insertion positions according to the influence degree corresponding to each insertion position; and displaying the recommendation information at the target insertion position in the information stream. Through the application, the recommendation information can be presented in a suitable insertion position in the information stream.

Description

Technical Field

[0001] This application relates to Internet technology, and more particularly to an information recommendation method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] With the widespread adoption of internet technology, electronic devices can provide richer information streams, such as text streams (news streams), image streams, and video streams (short video streams). More and more users are accessing various information streams through electronic devices, such as browsing news through news apps, greatly improving the convenience of their lives. Various recommended information, such as advertisements and mini-programs, can be inserted into these information streams.

[0003] In related technologies, recommending information is presented at random or fixed locations in the information stream. This recommendation method inevitably interferes with or interrupts the browsing of the information stream, affecting the user experience; it also leads to the invalid push of recommended information and causes unnecessary consumption of server resources (including communication resources and computing resources). Summary of the Invention

[0004] This invention provides an information recommendation method, apparatus, electronic device, and computer-readable storage medium that can present recommendation information at appropriate insertion points in an information stream.

[0005] The technical solution of this invention is implemented as follows:

[0006] This invention provides an information recommendation method, including:

[0007] Presenting the information flow in the human-computer interaction interface;

[0008] Determine the degree of impact when inserting recommendation information at multiple insertion positions in the information stream;

[0009] Based on the degree of influence corresponding to the plurality of insertion positions, the target insertion position is determined from the plurality of insertion positions;

[0010] The recommendation information is displayed at the target insertion position in the information stream.

[0011] This invention provides an information recommendation device, comprising:

[0012] The presentation module is used to display the information flow on the human-computer interaction interface;

[0013] The processing module determines the degree of impact when inserting recommendation information at multiple insertion positions in the information stream;

[0014] The determining module is used to determine the target insertion position used when presenting the recommendation information from the plurality of insertion positions based on the degree of influence corresponding to each of the plurality of insertion positions;

[0015] A display module is used to display the recommendation information at the target insertion position in the information stream.

[0016] In the above technical solution, the processing module is further configured to determine the group to which the target recommended object belongs based on the historical recommendation information of the target recommended object;

[0017] Obtain multiple insertion position features corresponding to the group to which the target recommended object belongs;

[0018] Based on the multiple insertion position features corresponding to the group to which the target recommendation object belongs, the degree of influence when inserting the recommendation information at multiple insertion positions in the information flow is determined.

[0019] In the above technical solution, the processing module is also used to determine the tolerance and acceptance level of the target recommendation object to historical recommendation information;

[0020] The group to which the target recommendation object belongs is determined based on the target recommendation object's tolerance and acceptance of the historical recommendation information.

[0021] In the above technical solution, the processing module is used to determine parameters corresponding to the tolerance level of the target recommendation object to historical recommendation information, the parameters including:

[0022] The exposure of the target recommendation object's historical recommendation information, wherein the historical recommendation information is presented during the target recommendation object's browsing of the historical information stream;

[0023] The duration of the target recommended object's browsing history information stream, wherein the historical information stream contains the historical recommendation information;

[0024] The processing module is used to determine parameters corresponding to the degree to which the target recommendation object accepts historical recommendation information, the parameters including:

[0025] The click-through rate of the historical recommendation information of the target recommendation object;

[0026] The number of clicks on the historical recommendation information of the target recommendation object;

[0027] The processing module is also used to perform a weighted summation of the exposure volume of the target recommendation object's historical recommendation information and the duration of the target recommendation object's browsing history information stream, and use the weighted summation result as the target recommendation object's tolerance level for the historical recommendation information;

[0028] The click-through rate of the target recommendation object's historical recommendation information and the click volume of the target recommendation object's historical recommendation information are weighted and summed, and the weighted sum is used as the degree to which the target recommendation object accepts the historical recommendation information.

[0029] In the above technical solution, the processing module is further configured to determine, from multiple tolerance ranges, the target tolerance range in which the tolerance of the target recommendation object to the historical recommendation information is located, and to determine the group corresponding to the target tolerance range as the target group to which the target recommendation object belongs, wherein each tolerance range corresponds to a group;

[0030] The multiple tolerance ranges are obtained by dividing the range of tolerance values; the range of tolerance values ​​is composed of the tolerance values ​​of multiple historical recommended objects to the historical recommended information, and the range of tolerance values ​​ends with the maximum and minimum values ​​of the tolerance.

[0031] From multiple acceptance degree intervals, determine the target acceptance degree interval in which the target recommended object accepts the historical recommendation information, and determine the subgroup corresponding to the target acceptance degree interval as the subgroup to which the target recommended object belongs in the target group, wherein each acceptance degree interval corresponds to a subgroup;

[0032] The multiple acceptance degree intervals are obtained by dividing the value range of acceptance degree; the value range of acceptance degree is composed of the acceptance degree values ​​of the multiple historical recommendation objects to the historical recommendation information, and the value range of acceptance degree is terminated by the maximum and minimum values ​​of the acceptance degree.

[0033] In the above technical solution, the device further includes: a preprocessing module, used to divide the multiple historical recommendation objects into multiple groups according to the tolerance and acceptance of the multiple historical recommendation objects to historical recommendation information;

[0034] Determine the characteristics of each insertion position in the historical information stream of each group, and use them as the insertion position characteristics of the group.

[0035] In the above technical solution, the features of the insertion position include the exposure rate of the insertion position; the preprocessing module is further configured to traverse the historical information stream sent to the historical recommendation objects in each group, and perform the following processing for each insertion position in the traversed historical information stream:

[0036] Determine a first number of historical recommended objects in the group that have been viewed to the insertion position;

[0037] Determine a second number of historical recommended objects in the group that have been viewed up to the first insertion position;

[0038] The ratio of the first quantity to the second quantity is determined as the exposure rate of the insertion position.

[0039] In the above technical solution, the preprocessing module is further used to divide the range of tolerance levels of the multiple historical recommendation objects to the historical recommendation information into multiple tolerance ranges;

[0040] Each tolerance level interval corresponds to a group, and the range of tolerance level values ​​is defined by the maximum and minimum values ​​of the tolerance level.

[0041] Determine the tolerance range of each historical recommendation object to the historical recommendation information, and determine the group corresponding to the tolerance range as the target group to which the historical recommendation object belongs;

[0042] The range of values ​​for the degree of acceptance of the historical recommendation information by the historical recommendation object is divided into multiple acceptance degree ranges;

[0043] Each of the acceptance degree intervals corresponds to a subgroup in the target group, and the value interval of the acceptance degree is set to the maximum and minimum values ​​of the acceptance degree.

[0044] Determine the acceptance level range of the historical recommendation object to which it accepts the historical recommendation information, and determine the subgroup corresponding to the determined acceptance level range as the subgroup to which the historical recommendation object belongs in the target group.

[0045] In the above technical solution, the processing module is further configured to perform the following processing for each of the multiple insertion positions in the information stream:

[0046] Based on the insertion position characteristics and the recommendation characteristics of the recommendation information, determine the information promotion benefit when inserting the recommendation information at the insertion position;

[0047] Wherein, the insertion position is any one of a plurality of insertion positions for inserting the recommendation information in the information stream for the target recommendation object;

[0048] Based on the insertion position characteristics, determine the information interference loss when inserting the recommendation information at the insertion position;

[0049] The remaining portion after offsetting the information promotion benefits by the information interference losses is taken as the degree of influence when inserting the recommendation information at the insertion position.

[0050] In the above technical solution, the processing module is further configured to use the difference between the information promotion benefit and the information interference loss when inserting the recommendation information at the insertion position as the degree of influence when inserting the recommendation information at the insertion position; or,

[0051] Multiply the loss quantization coefficient by the information interference loss, and use the result of the multiplication as the information interference loss after loss quantization.

[0052] The difference between the information promotion benefit when inserting the recommendation information at the insertion position and the information interference loss after loss quantification is taken as the degree of influence when inserting the recommendation information at the insertion position.

[0053] In the above technical solution, the insertion position features include the exposure rate of the insertion position and the click-through rate of the historical recommended information inserted at the insertion position; the recommendation features include the exposure cost and the estimated click-through rate; the processing module is further used to determine the ratio of the click-through rate of the historical recommended information inserted at the insertion position to the estimated click-through rate of the recommended information as the information promotion revenue weight of the recommended information;

[0054] The information promotion revenue weight of the recommended information, the exposure cost of the recommended information, and the exposure rate of the insertion position are multiplied together, and the result of the multiplication is used as the information promotion revenue corresponding to inserting the recommended information at the insertion position.

[0055] In the above technical solution, the information interference loss includes interference loss and information loss; wherein, the interference loss is used to quantify the following information: the number of new recommended information that can be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream continues;

[0056] The information loss is used to quantify the following information: the number of new recommended information that cannot be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream stops.

[0057] In the above technical solution, the insertion position features include the probability of continuing to browse the information stream after inserting the recommendation information at the insertion position, and the expected exposure of the insertion position; the processing module is further configured to multiply the probability of continuing to browse the information stream after inserting the recommendation information at the insertion position by the expected exposure of the insertion position, and use the product result as the interference loss when inserting the recommendation information at the insertion position.

[0058] In the above technical solution, the insertion position feature includes the probability of stopping browsing the information stream after inserting the recommended information at the insertion position, and the expected exposure of subsequent insertion positions at the insertion position; wherein, the priority of the information corresponding to the insertion position in the information stream is higher than the priority of the information corresponding to the subsequent insertion position in the information stream, and the priority sorting method is consistent with the priority sorting method of the information in the information stream; wherein, the priority sorting method of the information in the information stream includes at least one of the following: ascending or descending order based on publication time; ascending or descending order based on recent comment time; ascending or descending order based on popularity; ascending or descending order based on clicks; ascending or descending order based on reposts; and the order of information arrangement in the information stream; the processing module is further configured to multiply the probability of stopping browsing the information stream after inserting the recommended information at the insertion position by the expected exposure of subsequent insertion positions at the insertion position, and use the product result as the information loss when inserting the recommended information at the insertion position.

[0059] In the above technical solution, the determining module is further used to sort the degree of influence corresponding to the multiple insertion positions in descending order and filter out at least one degree of influence that is ranked first.

[0060] The insertion position corresponding to the selected degree of influence is determined as the target insertion position.

[0061] This invention provides an electronic device for information recommendation, the electronic device comprising:

[0062] Memory, used to store executable instructions;

[0063] The processor, when executing executable instructions stored in the memory, implements the information recommendation method provided in the embodiments of the present invention.

[0064] This invention provides a computer-readable storage medium storing executable instructions for inducing a processor to execute and implement the information recommendation method provided in this invention.

[0065] The embodiments of the present invention have the following beneficial effects:

[0066] By selecting the appropriate insertion position to present recommendation information based on the degree of influence, the integration of recommendation information and information flow is promoted. While ensuring the user experience of browsing information flow, a good information recommendation effect is achieved, without worrying that the browsing information flow will be interfered with or interrupted when presenting recommendation information, thus achieving the purpose of effective information push and saving server resources. Attached Figure Description

[0067] Figure 1This is a schematic diagram illustrating an application scenario of the information recommendation system provided in an embodiment of the present invention;

[0068] Figure 2 This is a schematic diagram of the structure of an electronic device 500 for information recommendation provided in an embodiment of the present invention;

[0069] Figure 3-7 This is a flowchart illustrating the information recommendation method provided in an embodiment of the present invention;

[0070] Figure 8 This is a schematic diagram of the recommendation system provided in the embodiments of the present invention;

[0071] Figure 9A This is a schematic diagram of the advertising recommendation interface provided in an embodiment of the present invention;

[0072] Figure 9B This is a schematic diagram of the advertising recommendation interface provided in an embodiment of the present invention;

[0073] Figure 10 This is a flowchart illustrating the advertising ranking algorithm provided in an embodiment of the present invention. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] In the following description, the terms "first" and "second" are used merely to distinguish similar recommended objects and do not represent a specific ordering of recommended objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of the invention described herein can be implemented in an order other than that illustrated or described herein.

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0077] Before providing a further detailed description of the embodiments of the present invention, the nouns and terms involved in the embodiments of the present invention will be explained, and the nouns and terms involved in the embodiments of the present invention shall be interpreted as follows.

[0078] 1) Click-Through-Rate (CTR): The click-through rate of recommended information (such as mini-programs, image ads, text ads, keyword ads, ranking ads, video ads, etc.), which is the ratio of the actual number of clicks (number of times the target page is reached) of the recommended information to the number of times the recommended information is displayed (Show Content).

[0079] 2) Cost Per Mile (CPM): The cost per thousand impressions is the unit of measurement for the cost of delivering a media outlet or media schedule to 1,000 people or "households." It is a method for measuring the actual utility of a recommendation's investment, i.e., the advertising cost required to reach each thousand people who hear or see a recommendation (e.g., an advertisement). CPM depends on the "impression" metric, which is the number of times a person's eyes focus on a recommendation within a fixed period of time. The formula for calculating CPM is as follows: Cost Per Mile = (Advertising Cost / Reach) × 1000, where Advertising Cost / Reach is usually expressed as a percentage. When estimating this percentage, it is important to consider whether the advertising investment is national or regional.

[0080] 3) Effective Cost Per Mile (eCPM): Revenue from referrals (e.g., ads) generated per thousand impressions. The unit of measurement can be a webpage, an ad unit, or even a single ad. By default, eCPM refers to revenue per thousand pageviews. eCPM is a parameter used to reflect a website's profitability and does not represent revenue itself. The formula for calculating eCPM is as follows: eCPM = Revenue / Pageviews × 1000, where Revenue = Ad Price × Page Click-Through Rate × Pageviews, i.e., eCPM = Ad Price × Page Click-Through Rate × 1000. eCPM is a metric unrelated to the number of pageviews.

[0081] 4) Conversion Rate (CVR): The ratio of the number of conversions to the total number of clicks on a recommended ad within a given statistical period. The formula is: Conversion Rate = (Number of Conversions / Number of Clicks) × 100%. For example: If 10 users see a recommended ad, 5 of them click on it and are redirected to the target page, and 2 of them subsequently convert their clicks, then the conversion rate for that ad is (2 / 5) × 100% = 40%.

[0082] 5) Feeds: A continuously updated information stream presented to users. A feed is a content aggregator that combines several message sources actively subscribed to by users, helping them continuously obtain the latest content from these sources. These sources are typically news websites and blogs. Feeds can be displayed in various formats, primarily timelines and ranks. A timeline displays content in the order it was updated, such as on Weibo or WeChat Moments. A rank calculates the weight of content based on certain factors, determining the order in which it is displayed. For example, the current Weibo homepage feed algorithm has abandoned the original timeline and adopted a new intelligent sorting system.

[0083] 6) Insertion position: The insertion position can be the position between any two pieces of information in the information flow. For example, if an information flow consists of 3 pieces of information (i.e., the first piece of information, the second piece of information, and the third piece of information), then the position between the first piece of information and the second piece of information is an insertion position, and the position between the second piece of information and the third piece of information is an insertion position. The insertion position can also be the position adjacent to the information in the information flow. For example, if an information flow consists of 2 pieces of information (i.e., the first piece of information and the second piece of information), then the position before the first piece of information is an insertion position, the position after the first piece of information and before the second piece of information is an insertion position, and the position after the second piece of information is an insertion position.

[0084] This invention provides an information recommendation method, apparatus, electronic device, and computer-readable storage medium that can present recommendation information at appropriate insertion points in an information stream.

[0085] The following describes an exemplary application of the electronic device for information recommendation provided in the embodiments of the present invention.

[0086] The electronic device for information recommendation provided in this invention can be various types of terminals or servers. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smartwatch, or any other intelligent device capable of browsing information. The terminal and server can be directly or indirectly connected via wired or wireless communication, and this invention does not impose any limitations on this connection.

[0087] Taking a server as an example, the server determines the target insertion position for displaying the recommended information based on the perceived impact of inserting it at multiple points in the information stream. The server then sends the recommended information and the target insertion position to the client, which displays the recommended information at the target insertion position in the information stream. For instance, in a news application, recommended advertisements are displayed at the target insertion position in a news stream composed of multiple news articles; in a short video application, recommended game mini-programs are displayed at the target insertion position in a short video stream composed of multiple short videos.

[0088] See Figure 1 , Figure 1 This is a schematic diagram of an application scenario of the information recommendation system 10 provided in an embodiment of the present invention. The terminal 200 is connected to the server 100 through the network 300. The network 300 can be a wide area network or a local area network, or a combination of the two.

[0089] The terminal where the client (such as a news client, short video client, etc.) is located can send information stream requests. For example, when a user scrolls through a news page through the terminal's human-computer interaction interface to browse news, the terminal automatically obtains news update requests.

[0090] In some embodiments, terminal 200 sends an information stream request to server and invokes the information recommendation function provided by server 100. Server 100, through the information recommendation method provided in this embodiment, determines the target insertion position for presenting the recommended information among multiple insertion positions based on the degree of influence corresponding to inserting recommended information at multiple insertion positions in the information stream, and sends the recommended information and the target insertion position to present the recommended information at the target insertion position in the information stream. For example, if a news client is installed on terminal 200, after the user scrolls through the news page, terminal 200 automatically generates a news update request and sends the news update request to server 100 through network 300. Server 100 determines the recall information corresponding to the user's profile information from the database based on the user's profile information, and filters the recall information to select recommended information that matches the user's interests. Based on the degree of influence corresponding to inserting recommended information at multiple insertion positions in the news stream, server 100 determines the target insertion position for presenting the recommended information among multiple insertion positions and sends the recommended information and the target insertion position to the news client. The recommended information is presented through the display interface 210 of terminal 200 at the target insertion position in the news stream.

[0091] The structure of the electronic device for information recommendation provided in the embodiments of the present invention is described below. The electronic device for information recommendation can be various terminals, such as mobile phones, computers, etc.

[0092] See Figure 2 , Figure 2This is a schematic diagram of the structure of an electronic device 500 for information recommendation provided in an embodiment of the present invention. Taking the electronic device 500 as a server as an example, the electronic device 500 includes: at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together through a bus system 540. It is understood that the bus system 540 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 2 The general labeled all buses as Bus System 540.

[0093] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0094] Memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this embodiment is intended to include any suitable type of memory. Memory 550 may optionally include one or more storage devices physically located away from processor 510.

[0095] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0096] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0097] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0098] In some embodiments, the information recommendation device provided in this invention can be implemented in software. Figure 2 An information recommendation device 555 stored in a memory 550 is shown. It may be software in the form of programs and plug-ins, and includes a series of modules, including a presentation module 5551, a processing module 5552, a determination module 5553, a display module 5554, and a preprocessing module 5555. The presentation module 5551, the processing module 5552, the determination module 5553, the display module 5554, and the preprocessing module 5555 are used to implement the information recommendation function provided in the embodiments of the present invention.

[0099] As can be understood from the above, the information recommendation method provided in the embodiments of the present invention can be implemented by various types of electronic devices used for information recommendation, such as smart terminals and servers.

[0100] The information recommendation method provided by the embodiments of the present invention will be described below with reference to exemplary applications and implementations of the server provided in the embodiments of the present invention. See also Figure 3 , Figure 3 This is a flowchart illustrating the information recommendation method provided in an embodiment of the present invention, combined with... Figure 3 The steps shown are explained.

[0101] In step 101, the client sends an information stream request to the server.

[0102] For example, the client terminal (such as a social networking client, news client, short video client, etc.) can send information stream requests to the server. For instance, after swiping through a news page on the terminal's human-computer interaction interface and browsing the current news page, the terminal automatically obtains a news update request to request more news streams.

[0103] In step 102, the server sends an information stream to the client.

[0104] For example, after receiving an information stream request, the server responds to the client's request by sending an information stream to the client. The client can then display the information stream in the human-computer interaction interface. This embodiment does not limit the order of sending the information stream and subsequently sending the target insertion position; that is, steps 102 and 105 do not have a clear order. For example, after determining the target insertion position, the server can send the information stream, the target insertion position, and the recommendation information to the client together.

[0105] In step 103, the server determines the degree of impact when inserting recommendation information at multiple insertion positions in the information stream.

[0106] As an example of obtaining recommendation information, see Figure 8 , Figure 8This is a schematic diagram of the principle of the recommendation system provided in this embodiment of the invention. The recommendation system is used to obtain recommendation information, such as advertisements, game mini-programs, etc. During the process of the target user using the client running on the terminal, the client reports the interaction behavior (user log) of the target user on the historical recommendation information to the server 200 as training sample data, as well as the user profile and user features corresponding to the target user. The training sample data comes from the behavior data of different users reported by various terminals. Based on these behavior data, the prediction model is trained. The user profile and user features are obtained from the terminal corresponding to a certain user. The prediction model predicts the candidate recommendation information based on the obtained user profile and user features corresponding to the target user, and performs mixed ranking processing and recommendation based on the prediction results. Therefore, it can recommend content that meets the user's interests. Under this premise, the diversity of recommended content can be achieved through mixed ranking processing.

[0107] See Figure 8 Terminal 200 connects to server 100 via network 300. Server 100 includes an information generation module, a prediction module, and a ranking module. The information generation module uses various recommendation algorithms to quickly filter out candidate recommendation information relevant to the user from the candidate recommendation information database 400, and submits the recalled candidate recommendation information to the prediction module. The prediction module predicts and ranks the recalled candidate recommendation information based on the target user's user profile and user characteristics, obtaining ranked recommendation information. After obtaining the recommendation information, in response to the recommendation information request, the server can determine the degree of impact corresponding to inserting recommendation information at any insertion position in the information stream. For example, if there are 3 insertion positions in the information stream, the server can determine the degree of impact corresponding to inserting recommendation information at each of the 3 insertion positions.

[0108] See Figure 4 , Figure 4 This is an optional flowchart illustrating the information recommendation method provided in this embodiment of the invention. To accurately determine the degree of impact when recommendation information is inserted at any insertion point in the information stream, Figure 4 Show Figure 3 Step 103 can be achieved through Figure 4 Steps 1031 to 1033 shown implement the following: In step 1031, the group to which the target recommended object belongs is determined based on the historical recommendation information of the target recommended object; in step 1032, multiple insertion position features corresponding to the group to which the target recommended object belongs are obtained; in step 1033, the degree of influence when inserting recommendation information at multiple insertion positions in the information flow is determined based on the multiple insertion position features corresponding to the group to which the target recommended object belongs.

[0109] The target recommendation object is an object that meets the targeting conditions of the recommendation information. The server can first determine the group to which the target recommendation object belongs by using the historical recommendation information of the target recommendation object, such as obtaining the historical recommendation information of the target user from the target user's log file, and determine multiple insertion position features corresponding to the group to which the target recommendation object belongs. Finally, by referring to the multiple insertion position features corresponding to the group to which the target recommendation object belongs, the server can determine the degree of influence corresponding to inserting recommendation information at multiple insertion positions in the information flow.

[0110] See Figure 5 , Figure 5 This is an optional flowchart illustrating the information recommendation method provided in this embodiment of the invention. To accurately determine the group to which the target recommendation object belongs, Figure 5 Show Figure 4 Step 1031 can be achieved through Figure 5 Steps 311 to 312 shown implement the following: In step 311, the tolerance and acceptance of the target recommendation object to historical recommendation information are determined; in step 312, the group to which the target recommendation object belongs is determined based on the tolerance and acceptance of the target recommendation object to historical recommendation information.

[0111] Following the example above, in order to assign the target recommended object to the accurate group, we can obtain the target recommended object's tolerance and acceptance of historical recommended information based on its historical recommended objects. Since the tolerance and acceptance of recommended information vary among different groups, we can determine the group to which the target recommended object belongs by measuring its tolerance and acceptance of historical recommended information.

[0112] In some embodiments, determining the tolerance and acceptance level of the target recommendation object to historical recommendation information includes: weighting and summing the exposure of the target recommendation object's historical recommendation information and the duration of the target recommendation object's browsing of the historical information stream, and using the weighted sum as the target recommendation object's tolerance level to historical recommendation information; and weighting and summing the click pass rate of the target recommendation object's historical recommendation information and the number of clicks on the target recommendation object's historical recommendation information, and using the weighted sum as the target recommendation object's acceptance level to historical recommendation information.

[0113] Following the above example, the parameters used to determine the target user's tolerance for historical recommendation information include: the exposure volume of the target user's historical recommendation information, wherein the historical recommendation information is presented while the target user is browsing the historical information stream; and the duration of the target user browsing the historical information stream, wherein historical recommendation information is inserted into the historical information stream. The parameters used to determine the target user's acceptance of historical recommendation information include: the click-through rate of the target user's historical recommendation information; and the number of clicks on the target user's historical recommendation information. The tolerance level in this embodiment is not limited to exposure volume and duration; any indicator that reflects the passive nature of the target user's response to recommendation information is acceptable. Similarly, the acceptance level in this embodiment is not limited to click-through rate and clicks; any indicator that reflects the active nature of the target user's response to recommendation information is acceptable.

[0114] In addition, the exposure of the target recommendation object's historical recommendation information or the duration of the target recommendation object's browsing of the historical information stream can be directly used as the target recommendation object's tolerance for historical recommendation information; the click-through rate of the target recommendation object's historical recommendation information or the number of clicks on the target recommendation object's historical recommendation information can be used as the target recommendation object's acceptance of historical recommendation information.

[0115] In some embodiments, determining the group to which the target recommendation object belongs based on the target recommendation object's tolerance and acceptance of historical recommendation information includes: determining the target tolerance interval from multiple tolerance intervals, and determining the group corresponding to the target tolerance interval as the target group to which the target recommendation object belongs; determining the target acceptance interval from multiple acceptance intervals, and determining the subgroup corresponding to the target acceptance interval as the subgroup to which the target recommendation object belongs in the target group.

[0116] Each tolerance interval corresponds to a group, and the multiple tolerance intervals are obtained by dividing the tolerance value range. The tolerance value range is composed of the tolerance values ​​of multiple historical recommended objects towards the historical recommended information, with the maximum and minimum tolerance values ​​as endpoints. Similarly, each acceptance interval corresponds to a subgroup, and the multiple acceptance intervals are obtained by dividing the acceptance value range. The acceptance value range is composed of the acceptance values ​​of multiple historical recommended objects towards the historical recommended information, with the maximum and minimum acceptance values ​​as endpoints.

[0117] For example, metrics used to characterize tolerance include exposure; metrics used to characterize acceptance include click-through rate; target groups include: low-level exposure group, medium-level exposure group, and high-level exposure group; when the exposure of a target recommendation object is less than or equal to the low exposure threshold, the target recommendation object is classified into the low-level exposure group; when the exposure of a target recommendation object is greater than or equal to the high exposure threshold, the recommendation object is classified into the high-level exposure group; when the exposure of a target recommendation object is greater than the low exposure threshold but less than the high exposure threshold, the target recommendation object is classified into the medium-level exposure group; where the low exposure threshold is less than the high exposure threshold. Subgroups are further subdivided into groups under the target group, including low-level click group, medium-level click group, and high-level click group. When a target recommendation object is classified into the medium-level exposure group, it is further classified into its subgroup within the target group based on the click-through rate of the recommendation information. For example, when a target recommendation object is classified into the medium-level exposure group, it is further classified into the medium-level click group under the medium-level exposure group based on the click-through rate of the recommendation information.

[0118] For example, if there are multiple tolerance ranges, namely tolerance range 1 (value range 0-10) and tolerance range 2 (value range 10-20), and the current target recommendation object's tolerance for historical recommendation information is 5, then the target recommendation object's tolerance for historical recommendation information belongs to tolerance range 1. If there are multiple acceptance ranges, namely acceptance range 1 (value range 0-10) and tolerance range 2 (value range 10-20), and the current target recommendation object's acceptance for historical recommendation information is 15, then the target recommendation object's acceptance for historical recommendation information belongs to acceptance range 2 within tolerance range 1.

[0119] In some embodiments, before sending the information stream, the method further includes: dividing multiple historical recommendation objects into multiple groups based on the tolerance and acceptance of historical recommendation information by multiple historical recommendation objects; determining the features of each insertion position in the historical information stream of each group, and using them as the insertion position features of the group.

[0120] To avoid the need for real-time calculation of multiple insertion position features corresponding to the groups to which the target recommended object belongs, multiple insertion position features corresponding to groups can be calculated offline. This involves first dividing multiple historical recommended objects into multiple groups based on their tolerance and acceptance of historical recommendation information. Then, the features of each insertion position in the historical information stream of each group are determined and pre-stored in the database as the group's insertion position features. When it is necessary to obtain multiple insertion position features corresponding to a group, such as after determining the group to which the target recommended object belongs, the multiple insertion position features corresponding to the target recommended object's group are retrieved from the database. This saves the time required to calculate the multiple insertion position features corresponding to the target recommended object's group, enabling real-time acquisition of these features.

[0121] In some embodiments, determining the characteristics of each insertion position in the historical information stream of each group includes: traversing the historical information stream sent to historical recommended objects in each group, and performing the following processing for each insertion position in the traversed historical information stream: determining a first number of historical recommended objects in the group that have been viewed to the insertion position; determining a second number of historical recommended objects in the group that have been viewed to the first insertion position; and determining the ratio of the first number to the second number as the exposure rate of the insertion position.

[0122] The characteristics of the insertion position include the exposure rate at that position. The exposure rate pb at insertion position t. t The calculation method is: the number of users exposed at insertion position t divided by the number of users exposed at the first insertion position.

[0123] The features of the insertion position may also include: 1) the click-through rate of historical recommended information inserted at the insertion position; 2) the probability of continuing to browse the information stream after the recommended information is inserted at the insertion position; 3) the expected exposure of the insertion position; 4) the probability of stopping browsing the information stream after the recommended information is inserted at the insertion position; and 5) the expected exposure of subsequent insertion positions of the insertion position.

[0124] Among them, 1) the click pass rate of historical recommendation information inserted at insertion position t (ectr) t The calculation method is: the number of clicks of the historical recommendation information inserted at position t divided by the number of exposures of the historical recommendation information inserted at position t;

[0125] 2) The probability of continuing to browse the information stream after inserting recommendation information at insertion position t (prob1) t , which is the probability that a user continues to browse after a recommendation is inserted at position t in the information stream. It is calculated as the ratio of the number of users who browse to the insertion position t+1 to the number of users at the insertion position t.

[0126] 3) Expected exposure at insertion position t: expo1 ≥t Exponential exposure (expo1) refers to the expected exposure at insertion position t and subsequent insertion positions in the information stream. It is calculated as follows: the expected exposure at each position equals the average exposure per person at that position multiplied by the exposure rate at insertion position t. The accumulated expected exposure from insertion position t forward is then expressed as expo1. ≥t The value of ;

[0127] 4) The probability of stopping browsing the information stream after inserting recommendation information at insertion position t (prob2) t That is, the probability that a user stops browsing after inserting recommended information at position t in the information stream, 1 minus probability1. t That is, the probability of exiting the browser;

[0128] 5) The expected exposure expo2 at the subsequent insertion position after insertion position t. >t Exponential exposure (expo2) refers to the expected exposure of subsequent insertion positions after insertion position t in the information stream. It is calculated as follows: the expected exposure of each position equals the average exposure per person at that position multiplied by the current position's exposure rate. The expected exposure is accumulated starting from insertion position t+1. >t The value of .

[0129] In some embodiments, based on the tolerance and acceptance levels of multiple historical recommendation objects towards historical recommendation information, multiple historical recommendation objects are divided into multiple groups, including: dividing the range of tolerance levels of multiple historical recommendation objects towards historical recommendation information into multiple tolerance ranges; determining the tolerance range to which each historical recommendation object's tolerance level falls, and determining the group corresponding to the tolerance range as the target group to which the historical recommendation object belongs; dividing the range of acceptance levels of historical recommendation objects towards historical recommendation information into multiple acceptance ranges; determining the acceptance range to which the historical recommendation object's acceptance level falls, and determining the subgroup corresponding to the determined acceptance range as the subgroup to which the historical recommendation object belongs within the target group.

[0130] Each tolerance interval corresponds to a group, and the tolerance range is defined by the maximum and minimum values ​​of the tolerance. Each acceptance interval corresponds to a subgroup within the target group, and the acceptance range is defined by the maximum and minimum values ​​of the acceptance.

[0131] For example, metrics used to characterize tolerance include exposure; metrics used to characterize acceptance include click-through rate; target groups include: low-level exposure group, medium-level exposure group, and high-level exposure group; when the exposure of a target recommendation object is less than or equal to the low exposure threshold, the target recommendation object is classified into the low-level exposure group; when the exposure of a target recommendation object is greater than or equal to the high exposure threshold, the recommendation object is classified into the high-level exposure group; when the exposure of a target recommendation object is greater than the low exposure threshold but less than the high exposure threshold, the target recommendation object is classified into the medium-level exposure group; wherein, the low exposure threshold is less than the high exposure threshold. Subgroups include low-level click group, medium-level click group, and high-level click group. When a target recommendation object is classified into the medium-level exposure group, it is further classified into its subgroup within the target group based on the click-through rate of the recommendation information.

[0132] See Figure 6 , Figure 6 This is an optional flowchart illustrating the information recommendation method provided in an embodiment of the present invention. Figure 6 Show Figure 4 Step 1033 can be achieved through Figure 6 Steps 331 to 333 shown implement the following processing for each of the multiple insertion positions in the information flow: In step 331, the information promotion benefit when inserting recommendation information at the insertion position is determined based on the insertion position characteristics of the insertion position and the recommendation characteristics of the recommendation information; In step 332, the information interference loss when inserting recommendation information at the insertion position is determined based on the insertion position characteristics of the insertion position; In step 333, the remaining part after the information promotion benefit is offset by the information interference loss is taken as the degree of influence when inserting recommendation information at the insertion position.

[0133] The insertion position is any one of multiple insertion positions for inserting the recommendation information into the information stream for the target recommendation object. The group to which the target recommendation object belongs corresponds to multiple insertion positions, and each insertion position corresponds to an insertion position feature. The insertion position features of the same insertion position corresponding to different groups are different. For example, if the group to which the target recommendation object 1 belongs is group 1 and the group to which the target recommendation object 2 belongs is group 2, then the insertion position feature of insertion position 1 corresponding to group 1 is different from the insertion position feature of insertion position 1 corresponding to group 2.

[0134] For example, by using the insertion position features corresponding to the group to which the target recommended object belongs, and the recommendation features of the recommended information, the degree of influence corresponding to inserting recommended information at the insertion position can be characterized. Based on the insertion position features corresponding to the group to which the target recommended object belongs, and the recommendation features of the recommended information, the information promotion benefit corresponding to inserting recommended information at the insertion position is determined. Then, based on the insertion position features corresponding to the group to which the target recommended object belongs, the information interference loss corresponding to inserting recommended information at the insertion position is determined. Finally, the information interference loss is offset from the information promotion benefit, and the remaining portion of the information promotion benefit after offsetting the information interference loss is taken as the degree of influence corresponding to inserting recommended information at the insertion position. The larger the remaining portion, the greater the degree of influence corresponding to inserting recommended information at the insertion position.

[0135] In some embodiments, the remaining portion after the information promotion gains are offset by information interference losses is used as the degree of influence when inserting recommendation information at the insertion position. This includes: using the difference between the information promotion gains and the information interference losses corresponding to inserting recommendation information at the insertion position as the degree of influence when inserting recommendation information at the insertion position; or, multiplying the loss quantification coefficient by the information interference loss and using the product as the information interference loss after loss quantification; and using the difference between the information promotion gains when inserting recommendation information at the insertion position and the corresponding information interference loss after loss quantification as the degree of influence when inserting recommendation information at the insertion position.

[0136] Among them, the loss quantification coefficient can put loss and gain on the same dimension. It assigns a scaling factor to the loss so that loss and gain are on the same dimension. The loss quantification coefficient can be set based on empirical data.

[0137] See Figure 7 , Figure 7 This is an optional flowchart illustrating the information recommendation method provided in an embodiment of the present invention. Figure 7 Show Figure 6 Step 331 in the process can be achieved through Figure 7 Steps 3311 to 3312 shown are implemented as follows: In step 3311, the ratio of the click pass rate of the historical recommended information inserted at the insertion position to the estimated click pass rate of the recommended information is determined as the information promotion revenue weight of the recommended information; In step 3312, the information promotion revenue weight of the recommended information, the exposure cost of the recommended information, and the exposure rate of the insertion position are multiplied together, and the result of the multiplication is used as the information promotion revenue when inserting recommended information at the insertion position.

[0138] The insertion location features include the exposure rate of the insertion location and the click-through rate of historical recommendation information inserted at the insertion location; the recommendation features include the exposure cost and the estimated click-through rate. This represents the information promotion revenue corresponding to inserting recommendation information at insertion position t, where ecpm ad pb represents the cost of exposure for recommended information. t This indicates that the exposure rate at insertion position t is multiplied. The ectr represents the information promotion revenue weight of the recommended information. t ectr represents the click-through rate of historical recommendations inserted at insertion position t. ad This indicates the estimated click-through rate of the recommended information.

[0139] In some embodiments, the information interference loss includes interference loss and information loss, wherein the interference loss is used to quantify the number of new recommendations that can be presented when the recommendation is displayed at the insertion position and the browsing of the information stream continues; wherein the information loss is used to quantify the number of new recommendations that cannot be presented when the recommendation is displayed at the insertion position and the browsing of the information stream stops. For example, the sum of the interference loss and the information loss is taken as the information interference loss corresponding to inserting recommendation information at the insertion position.

[0140] In some embodiments, the insertion location features include the probability of continuing to browse the news feed after inserting recommendation information at the insertion location, and the expected exposure of the insertion location. t ×expo1 t This represents the interference loss corresponding to inserting recommendation information at insertion position t, prob1 t Expo1 represents the probability of continuing to browse the information stream after inserting recommendation information at insertion position t. t This represents the expected exposure at insertion position t. The probability of continuing to browse the feed after inserting recommendation information at the insertion position is multiplied by the expected exposure at the insertion position, and the product is used as the interference loss corresponding to inserting recommendation information at the insertion position.

[0141] In some embodiments, the insertion position features include the probability of stopping browsing the information stream after inserting recommendation information at the insertion position, and the expected exposure of subsequent insertion positions at the insertion position. The information at the corresponding insertion position in the information stream has a higher priority than the information at subsequent insertion positions, and the priority ranking method is consistent with the priority ranking method of information in the information stream. The priority ranking method of information in the information stream includes at least one of the following: ascending or descending order based on publication time; ascending or descending order based on recent comment time; ascending or descending order based on popularity; ascending or descending order based on clicks; ascending or descending order based on reposts; and the order in which information is arranged in the information stream. t ×expo2 >t This represents the information loss when inserting recommendation information at insertion position t, prob2 t Expo2 represents the probability of continuing to browse the information stream after inserting recommendation information. >t This represents the expected exposure at insertion position t. The probability of stopping browsing the feed after inserting recommendation information at the insertion position is multiplied by the expected exposure at subsequent insertion positions. The product is used as the information loss corresponding to inserting recommendation information at the insertion position.

[0142] The priority sorting method is consistent with the priority sorting method of information in the information flow. For example, the insertion position t corresponds to the first information, and the subsequent insertion position is the insertion position t+1, which corresponds to the second information. The priority of the first information in the information flow is higher than that of the second information, that is, the first information is before the second information. Therefore, the insertion position t is before the insertion position t+1. The insertion positions t and t+1 can be inserted in a way that is before the information in the information flow; or they can be inserted in a way that is after the information in the information flow; or the insertion position t can be inserted in a way that is before the information in the information flow, and the insertion position t+1 can be inserted in a way that is after the information in the information flow.

[0143] For example, if the insertion position is insertion position 1 and the subsequent insertion position is insertion position 2, and a certain information stream consists of 3 pieces of information (i.e., information 1 and information 2), then the insertion position before information 1 is insertion position 1, and the insertion position before information 2 is insertion position 2; or, the insertion position after information 1 and before information 2 is insertion position 1, and the insertion position after information 2 is insertion position 2; or, the insertion position before information 1 is insertion position 1, and the insertion position after information 2 is insertion position 2.

[0144] Information in the feed can be arranged in ascending or descending order based on publication time; in ascending or descending order based on recent comment time; in ascending or descending order based on popularity; in ascending or descending order based on click count; in ascending or descending order based on share count; or in the order in which information is arranged within the feed.

[0145] In step 104, the server determines the target insertion position from among the multiple insertion positions based on the degree of influence corresponding to each of the multiple insertion positions.

[0146] For example, the influence levels corresponding to multiple insertion positions are sorted in descending order, and at least one influence level at the top of the sort is selected. The insertion position corresponding to the selected influence level is then determined as the target insertion position for presenting recommendation information, so that recommendation information is presented at the target insertion position.

[0147] In step 105, the server sends the recommendation information and the target insertion position to the client.

[0148] In step 106, the client presents recommended information at the target insertion position in the information stream.

[0149] For example, after the client receives the information stream, recommendation information, and target insertion position, it presents the recommendation information at the target insertion position in the information stream, thus achieving dynamic presentation of recommendation information.

[0150] As an example, when target user 1 sends a data stream request to the server through a client running on terminal 1, and target user 2 sends a data stream request to the server through a client running on terminal 2, for example, when target user 1 and target user 2 are browsing news simultaneously, such as... Figure 9A As shown, the server recommends car advertisements (recommendation information) to target user 1 based on the user profile and user characteristics of target user 1. Based on the influence levels corresponding to multiple insertion positions, the server determines the target insertion position for displaying the car advertisement as position 901 between news 1 and news 2. Figure 9B As shown, based on the user profile and characteristics of target user 2, the server also recommends the same car advertisement to target user 2, and determines the target insertion position for displaying the car advertisement as position 902 between news 3 and news 4 based on the influence corresponding to multiple insertion positions. Figure 9A and Figure 9BIt can be seen that the server decides the insertion position of the recommended information differently for different users. That is, for each user, it will choose an appropriate insertion position to present the recommended information, which promotes the integration of recommended information with the information flow and ensures the user experience of browsing the information flow.

[0151] The following will describe an exemplary application of the embodiments of the present invention in a practical application scenario.

[0152] The embodiments of the present invention can be applied to information recommendation scenarios, such as advertising recommendations, etc. Figure 1 As shown, the terminal connects to the server 100 deployed in the cloud via network 300. A news client application is installed on the terminal. After the user scrolls to the bottom of the news page, the terminal 200 automatically generates a news update request and sends it to the server 100 via network 300. The server 100 determines the target insertion position for presenting the recommendation information based on the degree of impact (total benefit) of inserting recommendation information at multiple insertion positions in the news stream, and sends the recommendation information and the target insertion position to the news client. The recommendation information is then presented at the target insertion position in the news stream.

[0153] In the recommendation field, common ranking scenarios involve sorting content of a single category, such as ads, information, or products, by predicting the target's CTR, CVR, or CPM, and ranking them based on the predicted values. However, in real-world business scenarios, multiple types of content are delivered within the same context. For example, a certain number of ads are inserted into a spatial feed. In this scenario, the feed and ads cannot be equated, and a conventional CTR / CPM prediction method cannot be used for ranking. This ranking requirement is generally called mixed ranking, which involves ranking multiple different categories of content together. In related technologies, spatial feed stream mixed ranking solutions use a fixed-position approach, inserting ads into fixed positions within the feed stream, with higher ad eCPM values ​​resulting in earlier insertion positions. This mixed ranking method primarily focuses on product form and user experience, ensuring that inserted ads maintain an eCPM order from high to low without disrupting the user experience.

[0154] However, the mixed-layout schemes in related technologies have the following problems in practice: 1) Using the same mixed-layout method for all users does not take into account the tolerance of different groups for advertisements; 2) Using a mixed-layout scheme with fixed positions does not quantitatively measure the benefits and losses brought by inserting advertisements, and the mixed-layout scheme may not be optimal.

[0155] To address the aforementioned issues, this invention, considering the actual business model of the product, proposes a practical and comprehensive advertising ranking algorithm (an information recommendation method). The algorithm calculates the optimal insertion position of an advertisement using a specific formula, specifically calculating the impact of inserting an advertisement at different positions. The insertion position with the greatest impact is determined as the optimal insertion position. For example, if inserting an advertisement at position 1 has an impact of 2, inserting it at position 2 has an impact of 5, and inserting it at position 3 has an impact of 3, then the optimal insertion position is position 2. The algorithm scheme designs a method for calculating the impact of inserting advertisements at different positions (including the benefits of information promotion from pushing advertisements (information promotion benefits) and the losses caused by interference with the user's browsing information flow (information interference losses)). By comparing the magnitude of the benefits and losses, the optimal insertion position (target insertion position) is obtained, achieving dynamic ranking of advertisements.

[0156] The embodiments of the present invention can be used in recommendation scenarios where there is a need for mixed placement, such as mixed placement of advertisements in the feed stream of a space, mixed placement of advertisements in the information stream of news applications / microblog applications, and mixed placement of advertisements in short video streams. While minimizing disruption to the user experience, advertisements are inserted in the optimal position in the information stream (including image / text message streams, video streams (e.g., feed streams, information streams, or short video streams)) that the user is browsing, thereby maximizing the benefits.

[0157] In the mixed arrangement of feeds in space, this embodiment of the invention divides the audience based on the degree of ad tolerance and measures the probability of interruption of user browsing after ad insertion based on prior knowledge, so as to quantitatively calculate the expected loss brought by ad insertion. At the same time, combined with the potential benefits brought by the ad, under the premise of a fixed number of ads, the magnitude of loss and benefit is compared to find the optimal position of ad insertion, thereby taking into account both the goal of maximizing user experience and ad revenue.

[0158] like Figure 10 As shown, the advertising ranking algorithm of this embodiment mainly includes three parts when running online: The first part is the offline processing (preprocessing) part, which divides the audience and calculates the features of the required insertion position for different audiences offline and writes them into a memory, such as Kafka; the second part is the online processing part, which requests advertising information and related features (such as the e CPM and estimated CTR of the advertisement) in real time when users browse the Feeds stream, and pulls the features of the insertion position from Kafka; the third part is the online processing part, which calculates the optimal position (target insertion position) of the advertisement based on the requested features. The first part (audience segmentation and feature calculation) and the third part are described in detail below:

[0159] Part 1: Offline Processing

[0160] A) Population segmentation

[0161] The purpose of segmenting users (groups) is to achieve more accurate ranking and sorting. Different user groups exhibit significant differences in exposure and click-through rate (CTR). If these characteristics are calculated using a uniform set of data, the feature values ​​will be forcibly averaged, introducing unnecessary bias for user groups whose activity levels deviate from the average. Therefore, statistical analysis is used to segment users, employing metrics strongly correlated with user activity, such as exposure and ad CTR, to stratify users and use different strata as the classification results. Exposure and ad CTR are chosen as the basis for user segmentation because: exposure reflects the depth of a user's browsing and their tolerance for advertising (exposure is a passive indicator, meaning the number of times recommended information appears in the information feed, i.e., the number of passive views); higher exposure indicates deeper browsing and higher tolerance for advertising. Ad CTR reflects the user's acceptance of advertising (ad CTR is an active indicator, meaning the probability of actively viewing recommended information during the information feed, i.e., the probability of actively clicking); higher CTR indicates higher user acceptance, and vice versa.

[0162] In this invention, a user's browsing depth can conform to a certain discrete probability distribution, and the user himself conforms to a certain population distribution. Therefore, there can be a reasonable mapping relationship between the user and the browsing depth. This embodiment of the invention can use a joint probability distribution of "population-browsing depth". The loss can be calculated based on this joint probability distribution, making the loss calculation more reasonable.

[0163] The embodiments of the present invention are not limited to impressions / CTR, as long as they reflect passive viewing / active clicking. For example, impressions can be the time a user stays on the information feed page with embedded ads, and CTR can be the number of clicks.

[0164] Furthermore, when segmenting audiences, it's unnecessary to differentiate between ad CTR when exposure is too low or too high. This is because: firstly, with excessively low exposure, ad CTR becomes overly sensitive, its value fluctuates too much, and its randomness is too great to be statistically significant; secondly, with excessively high exposure, ad CTR becomes negligible in differentiating users, meaning the ad has little impact on users with high exposure, thus eliminating the need to further segment audiences using ad CTR.

[0165] Therefore, audience segmentation should be based on a comprehensive approach, considering both impressions and ad CTR. However, for excessively low or high impressions, the ad CTR can be disregarded; the specific thresholds for audience segmentation should be determined based on actual business conditions.

[0166] B) Calculate the characteristics of the insertion positions for different population groups.

[0167] After the population segmentation is completed, the following location features are calculated for each population group:

[0168] 1) ecpm ad Ad score, the estimated CPM (cost per mille) for an ad;

[0169] 2)ectr ad : Estimated CTR for advertisements;

[0170] 3)ectr t The estimated CTR of the ad inserted at position t in the feed (the click-through rate of historical recommendations inserted at position t) is calculated using data from the past month. The calculation method is: the number of ad clicks at position t divided by the number of ad impressions at position t.

[0171] 4)pb t The exposure rate at insertion position t in the feed stream is calculated using data from the past month. The calculation method is: the number of users exposed at insertion position t divided by the number of users exposed at the first insertion position.

[0172] 5)expo1 ≥t The expected exposure of the feed at insertion position t and subsequent insertion positions is calculated using data from the past month. The calculation method is as follows: the expected exposure at each position equals the average exposure per person at that position multiplied by the exposure rate at insertion position t. The accumulated expected exposure from insertion position t forward is given by expo1. ≥t The value of ;

[0173] 6)expo2 >t The expected exposure of positions following insertion position t in the feed stream, using data from the past month. The calculation method is as follows: the expected exposure of each position equals the average exposure per person at that position multiplied by the current position's exposure rate. The expected exposure is accumulated starting from insertion position t+1, i.e., expo2. >t The value of ;

[0174] 7)prob1 t The probability that a user will continue browsing after an ad is inserted at position t in the feed stream is calculated using data from the past month. Based on the meaning of conditional probability, the calculation method is: the proportion of users who browse to the insertion position t+1 to the number of users who browse to the insertion position t.

[0175] 8)prob2 tThe probability that a user will leave the browser (stop browsing) after an ad is inserted at position t in the feed stream, calculated using data from the past month, minus probability 1. t That is, the probability of exiting the browser;

[0176] 9) w: Loss quantification coefficient, which scales the loss so that the loss and the gain are on the same dimension; based on the total gain and total loss of each group of people inserting ads, a scaling coefficient is assigned to the loss so that the total loss and the total gain are on the same dimension. w is set based on empirical data.

[0177] Part 3: Calculating the Optimal Solution for the Insertion Position

[0178] Based on the above positional characteristics, the mixed arrangement formula provided in this embodiment of the invention is as shown in formula (1):

[0179]

[0180] in, This represents the weighted revenue (information promotion revenue) generated by inserting an advertisement at insertion position t. This represents the weight, which means: when ectr t Greater than ectr ad When the time is right, the returns are amplified; conversely, they are diminished. t ×(expo1 ≥t -expo2 >t This indicates the loss (interference loss) incurred when a user continues browsing after an ad is inserted at insertion position t; prob2 t ×expo2 >t This represents the loss (information loss) caused by a user leaving the browser after an ad is inserted at insertion position t. (prob1) t ×(expo1 ≥t -expo2 >t ) and prob2 t ×expo2 >t The summation, followed by scaling by a scaling factor w, yields the total loss (information interference loss).

[0181] The range of the formula is the range of insertion positions in the feed stream requested by the user. The position number is counted from the user's first request. The maximum value t of the formula can be obtained by using the above formula (1) within the position range of each request. This value is the optimal insertion position for the advertisement.

[0182] Therefore, from a product perspective, the mixed-layout scheme takes into account the acceptance and tolerance levels of different user groups for advertising, achieving real-time mixed-layout of ads in the feed stream while maximizing both user experience and advertising revenue. From a technical perspective, a method for calculating the loss caused by ad insertion is proposed: the probability calculation method of users continuing to browse and exiting the browsing after ad insertion is quantified, thus comprehensively and reasonably calculating the expected loss caused by ad insertion. By comparing the revenue and loss of different insertion positions, the optimal ad insertion position can be obtained. This invention overcomes the problems of fixed-rule ad mixed-layout schemes, maximizing advertising revenue and taking into account user browsing experience without increasing the number of ads; at the same time, by designing a loss calculation method, the inserted ads and recommended content are compared within the same category to determine the optimal ad insertion position.

[0183] The exemplary application and implementation of the server provided in the embodiments of the present invention have been used to illustrate the information recommendation method provided in the embodiments of the present invention. The following describes the scheme for each module in the information recommendation device 555 provided in the embodiments of the present invention to cooperate in implementing information recommendation.

[0184] Presentation module 5551 is used to present the information stream on the human-computer interaction interface; processing module 5552 is used to determine the degree of influence when inserting recommendation information at multiple insertion positions in the information stream; determination module 5553 is used to determine the target insertion position used to present the recommendation information from the multiple insertion positions according to the degree of influence corresponding to each of the multiple insertion positions; display module 5554 is used to display the recommendation information at the target insertion position in the information stream.

[0185] In some embodiments, the processing module 5552 is further configured to determine the group to which the target recommended object belongs based on the historical recommendation information of the target recommended object; obtain multiple insertion position features corresponding to the group to which the target recommended object belongs; and determine the degree of influence when inserting the recommendation information at multiple insertion positions in the information stream based on the multiple insertion position features corresponding to the group to which the target recommended object belongs.

[0186] In some embodiments, the processing module 5552 is further configured to determine the tolerance and acceptance level of the target recommendation object to historical recommendation information; and to determine the group to which the target recommendation object belongs based on the tolerance and acceptance level of the target recommendation object to the historical recommendation information.

[0187] In some embodiments, the processing module 5552 is further configured to determine parameters corresponding to the tolerance level of the target recommendation object to historical recommendation information, the parameters including: the exposure volume of the target recommendation object's historical recommendation information, wherein the historical recommendation information is presented during the target recommendation object's browsing of the historical information stream; the duration of the target recommendation object's browsing of the historical information stream, wherein the historical recommendation information is inserted into the historical information stream; the processing module 5552 is further configured to determine parameters corresponding to the acceptance level of the target recommendation object to historical recommendation information, the parameters including: the click-through rate of the target recommendation object's historical recommendation information; the number of clicks on the target recommendation object's historical recommendation information; the processing module 5552 is further configured to perform a weighted summation of the exposure volume of the target recommendation object's historical recommendation information and the duration of the target recommendation object's browsing of the historical information stream, and use the weighted summation result as the target recommendation object's tolerance level to the historical recommendation information; and to perform a weighted summation of the click-through rate of the target recommendation object's historical recommendation information and the number of clicks on the target recommendation object's historical recommendation information, and use the weighted summation result as the target recommendation object's acceptance level to the historical recommendation information.

[0188] In some embodiments, the processing module 5552 is further configured to determine, from multiple tolerance ranges, the target tolerance range in which the tolerance of the target recommended object to the historical recommendation information lies, and to determine the group corresponding to the target tolerance range as the target group to which the target recommended object belongs, wherein each tolerance range corresponds to a group; wherein the multiple tolerance ranges are obtained by dividing the range of tolerance values; the range of tolerance values ​​is composed of the tolerance values ​​of multiple historical recommended objects to the historical recommendation information, and the range of tolerance values ​​is defined by the maximum and minimum values ​​of the tolerance values. The value is an end value; from multiple acceptance degree intervals, the target acceptance degree interval in which the target recommended object accepts the historical recommendation information is located is determined, and the subgroup corresponding to the target acceptance degree interval is determined as the subgroup to which the target recommended object belongs in the target group, wherein each acceptance degree interval corresponds to a subgroup; wherein the multiple acceptance degree intervals are obtained by dividing the value interval of acceptance degree; the value interval of acceptance degree is composed of the values ​​of the acceptance degree of the multiple historical recommended objects to the historical recommendation information, and the value interval of acceptance degree has the maximum and minimum values ​​of acceptance degree as end values.

[0189] In some embodiments, the apparatus further includes: a preprocessing module 5555, configured to divide the multiple historical recommendation objects into multiple groups based on the tolerance and acceptance of the multiple historical recommendation objects to historical recommendation information; determine the features of each insertion position in the historical information stream of each group, and use them as the insertion position features of the group.

[0190] In some embodiments, the features of the insertion position include the exposure rate of the insertion position; the preprocessing module 5555 is further configured to traverse the historical information stream sent to the historical recommended objects in each group, and perform the following processing for each insertion position in the traversed historical information stream: determine a first number of historical recommended objects in the group that have been viewed to the insertion position; determine a second number of historical recommended objects in the group that have been viewed to the first insertion position; and determine the ratio of the first number to the second number as the exposure rate of the insertion position.

[0191] In some embodiments, the preprocessing module 5555 is further configured to divide the tolerance range of the multiple historical recommendation objects to the historical recommendation information into multiple tolerance ranges; wherein each tolerance range corresponds to a group, and the tolerance range is defined by the maximum and minimum values ​​of the tolerance; determine the tolerance range in which each historical recommendation object's tolerance to the historical recommendation information falls, and define the group corresponding to the tolerance range as the target group to which the historical recommendation object belongs; divide the acceptance range of the historical recommendation objects to the historical recommendation information into multiple acceptance ranges; wherein each acceptance range corresponds to a subgroup in the target group, and the acceptance range is defined by the maximum and minimum values ​​of the acceptance; determine the acceptance range in which the historical recommendation object's acceptance to the historical recommendation information falls, and define the subgroup corresponding to the determined acceptance range as the subgroup to which the historical recommendation object belongs in the target group.

[0192] In some embodiments, the processing module 5552 is further configured to perform the following processing for each of the plurality of insertion positions in the information stream: determining the information promotion benefit when inserting the recommendation information at the insertion position based on the insertion position characteristics of the insertion position and the recommendation characteristics of the recommendation information; wherein, the insertion position is any one of the plurality of insertion positions in the information stream for inserting the recommendation information for the target recommendation object; determining the information interference loss when inserting the recommendation information at the insertion position based on the insertion position characteristics of the insertion position; and taking the remaining portion after the information promotion benefit is offset by the information interference loss as the degree of influence when inserting the recommendation information at the insertion position.

[0193] In some embodiments, the processing module 5552 is further configured to take the difference between the information promotion benefit and the information interference loss when inserting the recommendation information at the insertion position as the degree of influence when inserting the recommendation information at the insertion position; or, multiply the loss quantification coefficient by the information interference loss, and take the result of the multiplication as the information interference loss after loss quantification; and take the difference between the information promotion benefit and the information interference loss after loss quantification when inserting the recommendation information at the insertion position as the degree of influence when inserting the recommendation information at the insertion position.

[0194] In some embodiments, the insertion location features include the exposure rate of the insertion location and the click-through rate of historical recommended information inserted at the insertion location; the recommendation features include the exposure cost and the estimated click-through rate; the processing module 5552 is further configured to determine the information promotion revenue weight of the recommended information as the ratio of the click-through rate of the historical recommended information inserted at the insertion location to the estimated click-through rate of the recommended information; and multiply the information promotion revenue weight of the recommended information, the exposure cost of the recommended information, and the exposure rate of the insertion location, and use the result of the multiplication as the information promotion revenue corresponding to inserting the recommended information at the insertion location.

[0195] In some embodiments, the information interference loss includes interference loss and information loss; wherein the interference loss is used to quantify the following information: the number of new recommended information that can be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream continues; wherein the information loss is used to quantify the following information: the number of new recommended information that cannot be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream stops.

[0196] In some embodiments, the insertion position features of the insertion position include the probability of continuing to browse the information stream after inserting the recommendation information at the insertion position, and the expected exposure of the insertion position; the processing module 5552 is further configured to multiply the probability of continuing to browse the information stream after inserting the recommendation information at the insertion position by the expected exposure of the insertion position, and use the product result as the interference loss when inserting the recommendation information at the insertion position.

[0197] In some embodiments, the insertion position feature includes the probability of stopping browsing the information stream after inserting the recommendation information at the insertion position, and the expected exposure of subsequent insertion positions at the insertion position; wherein, the priority of the information corresponding to the insertion position in the information stream is higher than the priority of the information corresponding to the subsequent insertion position in the information stream, and the priority sorting method is consistent with the priority sorting method of the information in the information stream; wherein, the priority sorting method of the information in the information stream includes at least one of the following: ascending or descending order based on publication time; ascending or descending order based on recent comment time; ascending or descending order based on popularity; ascending or descending order based on clicks; ascending or descending order based on reposts; and the order of information arrangement in the information stream; the processing module 5552 is further configured to multiply the probability of stopping browsing the information stream after inserting the recommendation information at the insertion position by the expected exposure of subsequent insertion positions at the insertion position, and use the product result as the information loss when inserting the recommendation information at the insertion position.

[0198] In some embodiments, the determining module 5553 is further configured to sort the degree of influence corresponding to the plurality of insertion positions in descending order, filter out at least one degree of influence that is ranked first, and determine the insertion position corresponding to the filtered degree of influence as the target insertion position.

[0199] This invention provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the information recommendation method described above in this invention.

[0200] This invention provides a computer-readable storage medium storing executable instructions. When these executable instructions are executed by a processor, they cause the processor to execute the information recommendation method provided in this invention, for example... Figure 3-7 The information recommendation method is shown.

[0201] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0202] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0203] As an example, executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0204] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0205] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of the present invention are included within the scope of protection of the present invention.

Claims

1. An information recommendation method, characterized in that, include: Presenting the information flow in the human-computer interaction interface; Based on the historical recommendation information of the target recommendation object, determine the group to which the target recommendation object belongs; Obtain multiple insertion position features corresponding to the group to which the target recommended object belongs; Based on the characteristics of multiple insertion positions corresponding to the group to which the target recommendation object belongs, determine the degree of influence when inserting recommendation information at multiple insertion positions in the information stream; Based on the degree of influence corresponding to the plurality of insertion positions, the target insertion position is determined from the plurality of insertion positions; The recommendation information is displayed at the target insertion position in the information stream.

2. The method according to claim 1, characterized in that, The step of determining the group to which the target recommendation object belongs based on the target recommendation object's historical recommendation information includes: Determine the tolerance and acceptance level of the target recommendation object towards historical recommendation information; The group to which the target recommendation object belongs is determined based on the target recommendation object's tolerance and acceptance of the historical recommendation information.

3. The method according to claim 2, characterized in that, The method includes: The parameters corresponding to the tolerance level of the target recommendation object to historical recommendation information are determined, and the parameters include: The exposure of the target recommendation object's historical recommendation information, wherein the historical recommendation information is presented during the target recommendation object's browsing of the historical information stream; The duration of the target recommended object's browsing history information stream, wherein the historical information stream contains the historical recommendation information; The parameters corresponding to the degree to which the target recommendation object accepts historical recommendation information are determined include: The click-through rate of the historical recommendation information of the target recommendation object; The number of clicks on the historical recommendation information of the target recommendation object; Determining the tolerance and acceptance level of the target recommendation object towards historical recommendation information includes: The exposure volume of the target recommendation object's historical recommendation information and the duration of the target recommendation object's browsing history information stream are weighted and summed, and the weighted sum is used as the target recommendation object's tolerance level for the historical recommendation information; The click-through rate of the target recommendation object's historical recommendation information and the click volume of the target recommendation object's historical recommendation information are weighted and summed, and the weighted sum is used as the degree to which the target recommendation object accepts the historical recommendation information.

4. The method according to claim 2, characterized in that, The step of determining the group to which the target recommendation object belongs based on the target recommendation object's tolerance and acceptance of the historical recommendation information includes: From multiple tolerance ranges, determine the target tolerance range in which the target recommendation object's tolerance for the historical recommendation information falls, and determine the group corresponding to the target tolerance range as the target group to which the target recommendation object belongs, wherein each tolerance range corresponds to one group; The multiple tolerance ranges are obtained by dividing the range of tolerance values; the range of tolerance values ​​is composed of the tolerance values ​​of multiple historical recommended objects to the historical recommended information, and the range of tolerance values ​​ends with the maximum and minimum values ​​of the tolerance. From multiple acceptance degree intervals, determine the target acceptance degree interval in which the target recommended object accepts the historical recommendation information, and determine the subgroup corresponding to the target acceptance degree interval as the subgroup to which the target recommended object belongs in the target group, wherein each acceptance degree interval corresponds to a subgroup; The multiple acceptance degree intervals are obtained by dividing the value range of acceptance degree; the value range of acceptance degree is composed of the acceptance degree values ​​of the multiple historical recommendation objects to the historical recommendation information, and the value range of acceptance degree is terminated by the maximum and minimum values ​​of the acceptance degree.

5. The method according to claim 1, characterized in that, Before presenting the information flow on the human-computer interaction interface, the method includes: Based on the tolerance and acceptance of historical recommendation information by multiple historical recommendation objects, the multiple historical recommendation objects are divided into multiple groups; Determine the characteristics of each insertion position in the historical information stream of each group, and use them as the insertion position characteristics of the group.

6. The method according to claim 5, characterized in that, The characteristics of the insertion position include the exposure rate of the insertion position; The process of determining the features of each insertion position in the historical information stream of each group includes: Iterate through the historical information stream sent to the historical recommendation objects in each of the groups, and perform the following processing for each insertion position in the traversed historical information stream: Determine a first number of historical recommended objects in the group that have been viewed to the insertion position; Determine a second number of historical recommended objects in the group that have been viewed up to the first insertion position; The ratio of the first quantity to the second quantity is determined as the exposure rate of the insertion position.

7. The method according to claim 5, characterized in that, The step of dividing the multiple historical recommendation objects into multiple groups based on their tolerance and acceptance of historical recommendation information includes: The tolerance range of the multiple historical recommendation objects for the historical recommendation information is divided into multiple tolerance ranges; Each tolerance level interval corresponds to a group, and the range of tolerance level values ​​is defined by the maximum and minimum values ​​of the tolerance level. Determine the tolerance range of each historical recommendation object to the historical recommendation information, and determine the group corresponding to the tolerance range as the target group to which the historical recommendation object belongs; The range of values ​​for the degree of acceptance of the historical recommendation information by the historical recommendation object is divided into multiple acceptance degree ranges; Each of the acceptance degree intervals corresponds to a subgroup in the target group, and the value interval of the acceptance degree is set to the maximum and minimum values ​​of the acceptance degree. Determine the acceptance level range of the historical recommendation object to which it accepts the historical recommendation information, and determine the subgroup corresponding to the determined acceptance level range as the subgroup to which the historical recommendation object belongs in the target group.

8. The method according to claim 1, characterized in that, The step of determining the degree of influence when inserting recommendation information at multiple insertion positions in the information stream based on multiple insertion position features corresponding to the group to which the target recommendation object belongs includes: For each of the multiple insertion positions in the information stream, the following processing is performed: Based on the insertion position characteristics and the recommendation characteristics of the recommendation information, determine the information promotion benefit when inserting the recommendation information at the insertion position; Wherein, the insertion position is any one of a plurality of insertion positions for inserting the recommendation information in the information stream for the target recommendation object; Based on the insertion position characteristics, determine the information interference loss when inserting the recommendation information at the insertion position; The remaining portion after offsetting the information promotion benefits by the information interference losses is taken as the degree of influence when inserting the recommendation information at the insertion position.

9. The method according to claim 8, characterized in that, The remaining portion after offsetting the information promotion benefits by the information interference losses, as the degree of influence when inserting the recommendation information at the insertion position, includes: The difference between the information promotion benefit and the information interference loss when inserting the recommendation information at the insertion position is taken as the degree of influence when inserting the recommendation information at the insertion position; or... Multiply the loss quantization coefficient by the information interference loss, and use the result of the multiplication as the information interference loss after loss quantization. The difference between the information promotion benefit when inserting the recommendation information at the insertion position and the information interference loss after loss quantification is taken as the degree of influence when inserting the recommendation information at the insertion position.

10. The method according to claim 8, characterized in that, The insertion location features include the exposure rate of the insertion location and the click-through rate of historical recommendation information inserted at the insertion location; the recommendation features include exposure cost and estimated click-through rate. The step of determining the information promotion benefit when inserting the recommendation information at the insertion position based on the insertion position characteristics and the recommendation information characteristics includes: The ratio of the click-through rate of the historical recommendation information inserted at the insertion position to the estimated click-through rate of the recommendation information is determined as the information promotion revenue weight of the recommendation information. The information promotion benefit weight of the recommended information, the exposure cost of the recommended information, and the exposure rate of the insertion position are multiplied together, and the result of the multiplication is used as the information promotion benefit when the recommended information is inserted at the insertion position.

11. The method according to claim 8, characterized in that, The information interference loss includes interference loss and information loss; wherein, the interference loss is used to quantify the following information: the number of new recommended information that can be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream continues; The information loss is used to quantify the following information: the number of new recommended information that cannot be presented when the recommended information is displayed at the insertion position and the process of browsing the information stream stops.

12. The method according to claim 8, characterized in that, The insertion position features include the probability of continuing to browse the information stream after inserting the recommendation information at the insertion position, and the expected exposure of the insertion position; The determination of information interference loss when inserting the recommendation information at the insertion position includes: The probability of continuing to browse the information stream after inserting the recommendation information at the insertion position is multiplied by the expected exposure at the insertion position, and the product is used as the interference loss when inserting the recommendation information at the insertion position.

13. The method according to claim 11, characterized in that, The insertion position features include the probability of stopping browsing the information stream after inserting the recommendation information at the insertion position, and the expected exposure of subsequent insertion positions at the insertion position; Wherein, the priority of the information corresponding to the insertion position in the information stream is higher than the priority of the information corresponding to the subsequent insertion position in the information stream, and the priority sorting method is consistent with the priority sorting method of the information in the information stream; The prioritization method for information in the information flow includes at least one of the following: Sort by publication date in ascending or descending order; Sort by the most recent comment time in ascending or descending order; Based on ascending or descending order of popularity; Sort by click count in ascending or descending order; Sort by forwarding volume (ascending or descending); Based on the order of information in the information flow; The determination of information interference loss when inserting the recommendation information at the insertion position includes: The probability of stopping browsing the information stream after inserting the recommendation information at the insertion position is multiplied by the expected exposure of the subsequent insertion position at the insertion position, and the product is used as the information loss when inserting the recommendation information at the insertion position.

14. The method according to claim 1, characterized in that, The step of determining the target insertion position from the plurality of insertion positions based on the degree of influence corresponding to each of the plurality of insertion positions includes: The influence levels corresponding to the multiple insertion positions are sorted in descending order, and at least one influence level that is ranked first is selected. The insertion position corresponding to the selected degree of influence is determined as the target insertion position.

15. An information recommendation device, characterized in that, include: The presentation module is used to display the information flow on the human-computer interaction interface; The processing module is used to determine the group to which the target recommendation object belongs based on the historical recommendation information of the target recommendation object; obtain multiple insertion position features corresponding to the group to which the target recommendation object belongs; and determine the degree of influence when inserting recommendation information at multiple insertion positions in the information stream based on the multiple insertion position features corresponding to the group to which the target recommendation object belongs. The determining module is used to determine the target insertion position from the plurality of insertion positions based on the degree of influence corresponding to each of the plurality of insertion positions; A display module is used to display the recommendation information at the target insertion position in the information stream.

16. An electronic device, characterized in that, include: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the information recommendation method according to any one of claims 1 to 14.

17. A computer-readable storage medium, characterized in that, It stores executable instructions for implementing the information recommendation method according to any one of claims 1 to 14 when executed by a processor.

18. A computer program product comprising computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, the information recommendation method according to any one of claims 1 to 14 is implemented.

Images