A short video advertisement recommendation method and system

Abstract

The application belongs to the field of internet advertisement service, and provides a short video advertisement recommendation method, which comprises the following steps: capturing real-time interaction data of a user; obtaining a user label according to the real-time interaction data of the user; obtaining an advertisement sequence according to the user label, and calculating a real-time concurrent performance index of the advertisement; generating an optimal advertisement sequence combination according to the real-time concurrent performance index, and performing advertisement recommendation according to the optimal advertisement sequence combination. According to the advertisement recommendation method of the embodiment of the application, the timeliness and concurrent load of the advertisement can be comprehensively evaluated, the effect and efficiency of advertisement delivery are improved, the stability and user experience of digital content multi-network channels are enhanced, and the advertisement delivery process is comprehensively optimized.

Description

Technical Field

[0001] This invention belongs to the field of internet advertising services, and specifically relates to a method and system for recommending short video ads. Background Technology

[0002] In the short video ecosystem, user interests exhibit highly dynamic changes, influenced by factors such as trending events, social media dissemination, and time-sensitive content, resulting in frequent abrupt shifts. With the rapid development of short video platforms, user interests shift rapidly along with news headlines and content dissemination on social media, making it difficult for traditional recommendation systems to accurately capture these sudden changes. Traditional recommendation algorithms largely rely on static historical profiles, such as weekly / monthly behavioral statistics and users' historical browsing records, to build user profiles. While this static data reflects users' long-term interests to some extent, in the short video ecosystem, traditional recommendation systems often fail to respond promptly to sudden trending events or real-time changes in public opinion due to signal lag, resulting in recommendations that do not accurately meet users' current interest needs.

[0003] As proposed in patent number CN114845170A, "An Advertising Placement Method, Smart Terminal and Computer-Readable Storage Medium", this method recommends advertisements based on static historical user profiles. This method relies on users' past behavioral data. However, in the rapidly changing short video ecosystem, users' interests may have changed. Existing internet advertising services are unable to adapt to such highly dynamic changes in interests, resulting in inefficient and inaccurate advertising placement. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in related technologies. Therefore, the first objective of this invention is to propose a short video advertising recommendation method that, by comprehensively evaluating the timeliness and concurrent load of advertisements, not only improves the effectiveness and efficiency of advertising delivery but also enhances the stability of digital content across multiple network channels and user experience, thus comprehensively optimizing the advertising delivery process.

[0005] The second objective of this invention is to propose a short video advertising recommendation system.

[0006] To achieve the above objectives, a first aspect of the present invention provides a short video advertisement recommendation method, the method comprising the following steps:

[0007] S100 captures real-time user interaction data;

[0008] S200, obtains user tags based on real-time user interaction data;

[0009] S300, obtaining an advertisement sequence according to a user label and calculating a real-time concurrent performance index of the advertisement;

[0010] S400, generating an optimal advertisement sequence combination according to the real-time concurrent performance index and performing advertisement recommendation according to the optimal advertisement sequence combination.

[0011] The advertisement recommendation method according to the embodiment of the application can comprehensively evaluate the timeliness and concurrent load of the advertisement, not only improve the effect and efficiency of advertisement delivery, but also enhance the stability and user experience of digital content multi-network channels, and comprehensively optimize the advertisement delivery process.

[0012] Further, in step S100, the user real-time interaction data includes: a play completion rate, a video duration, a like behavior or a sharing behavior.

[0013] The play completion rate, the video duration, the like behavior or the sharing behavior of the i-th video watched by the user in the last 24 hours are recorded; pec(i) is used to represent the play completion rate of the i-th video watched by the user in the last 24 hours, time(i) is used to represent the video duration of the i-th video watched by the user in the last 24 hours, wherein i is the serial number of the video, K(i) is used to represent the number of like behaviors or sharing behaviors of the i-th video watched by the user in the last 24 hours, wherein the serial number of the video i=1, 2, …, K, K represents the number of videos, and the total duration TZ of the video duration of the videos watched by the user in the last 24 hours is recorded, the maximum value Tmax of the video duration of the videos watched by the user in the last 24 hours is recorded, the minimum value Tmin of the video duration of the videos watched by the user in the last 24 hours is recorded, the average value TK of the video duration of the videos watched by the user in the last 24 hours is recorded, and the average value OCK of the number of like behaviors or sharing behaviors of the videos watched by the user in the last 24 hours is recorded.

[0014] In the short video ecology, the user interest is affected by hot events, social spread and time-sensitive content, and presents a high-frequency mutation characteristic, wherein the high-frequency mutation characteristic refers to the rapid, drastic and non-continuous change of the user interest in a short time due to external or internal factors; and the traditional recommendation system relies on static historical portraits such as week / month behavior statistics, for example, the advertisement delivery method proposed in the patent number CN114845170A entitled “Advertisement delivery method, intelligent terminal and computer readable storage medium”; this kind of advertisement delivery method often produces the problem that the historical behavior cannot reflect the current interest focus of the user due to signal lag when the attention is diverted due to sudden news, and the problem that it is difficult to establish an effective portrait for new users or low-active users due to sparse data; in order to solve the above problems, the step S200 is proposed in the application.

[0015] Further, in step S200, the step of obtaining the user label according to the user real-time interaction data comprises:

[0016] S201, calculating the shortest effective preference duration TPHa and the longest effective preference duration TPHb according to the user real-time interaction data;

[0017] Let the interaction adjustment factor KUI be K(i) / (1+OCK+K(i)), and let the effective video duration TOI be ; calculate the first offset duration difference TD1, wherein the first offset duration difference TD1 is TmaxxTOI / TZ; wherein the second offset duration difference TD2 is TminxTOI / TZ; the shortest effective preference duration TPHa is TK-TD1, and the longest effective preference duration TPHb is TK+TD2, to obtain the preference window [TPHa, TPHb];

[0018] The effective video duration TOI is used to measure the patience level of the user; the preference window [TPHa, TPHb] is used to shrink to a short duration when the user's patience is low (TOI is small), and is used to expand when the user's patience is high (TOI is large), so as to obtain an accurate user preference video range through curvature adaptive adjustment, and filter videos that obviously do not meet the user's preferences.

[0019] S202, filtering out videos with a video duration between the shortest effective preference duration TPHa and the longest effective preference duration TPHb according to the shortest effective preference duration TPHa and the longest effective preference duration TPHb, and recording the videos as preference videos;

[0020] S203, obtaining the user label according to the video label of the preference video;

[0021] Obtain the video label with the highest frequency of occurrence in the preference video, and match the user label to which the label belongs to obtain the matched user label.

[0022] The beneficial effect of this step is that the effective interest duration TOI is used as the Riemann metric of the user's patience level, and the dynamic preference window [TPHa, TPHb] is constructed in the manifold space through the coupling calculation of the video duration extreme value and the total duration TZ to obtain curvature adaptive adjustment, so that the duration preference interval can automatically stretch and shrink with the content type and the use scenario, and the portrait accuracy can still be maintained when the data is sparse.

[0023] Further, in step S300, obtaining the advertisement sequence according to the user label and calculating the advertisement real-time concurrency performance index comprises:

[0024] S310, after filtering out the matched advertisements from the database through the user label of the user, calculating the timeliness load time by calculating the waiting time of the server for sending the matched advertisements to other users in the last one hour;

[0025] The time-to-live load time method for calculating the waiting time of the matched advertisement sent to other users by the server in the last 1 hour is:

[0026] S311, using ty(j) to represent the waiting time of the jth matched advertisement sent to other users in the last 1 hour, where j represents the serial number of each matched advertisement, j = 1, 2, …, G, G represents the number of matched advertisements; the waiting time of the matched advertisement after receiving the user request in the last 1 hour is obtained; the maximum value of the waiting time of the jth matched advertisement sent to other users in the last 1 hour is recorded as tmax, the average value of the waiting time of the matched advertisement sent to other users in the last 1 hour is tpj, and the median of the waiting time of the matched advertisement sent to other users in the last 1 hour is tav;

[0027] S312, calculating the time-to-live load time fxt(j) of the jth advertisement; wherein fxt(j) is ty(j) plus the difference level ratio between the maximum waiting time and the average waiting time, wherein the difference level ratio between the maximum waiting time and the average waiting time is the difference between tmax and tpj multiplied by ) and .

[0028] Wherein, the time-to-live load time fxt(j) can reflect the difference between the maximum waiting time and the average waiting time, and quantify the response performance of the advertisement in time.

[0029] S320, obtaining the concurrent user number data of the matched advertisement, wherein the concurrent user number data includes the maximum concurrent user number, the current concurrent user number, and the total concurrent user number in the last 1 hour, and recording the maximum concurrent user number of the jth matched advertisement as bfsx(j), that is, the maximum number of simultaneous users that the advertisement can support; the current concurrent user number of the jth advertisement is recorded as bfs(j), that is, the number of users currently recommended the advertisement; and the total concurrent user number of the jth advertisement in the last 1 hour is recorded as bfsp(j), that is, the number of users to which the advertisement is sent in the past 1 hour;

[0030] S330, calculating the real-time concurrent performance index tfh(j) through the concurrent user number data and the time-to-live load time: tfh(j) is fxt(j) multiplied by the ratio of the current concurrent total delay OCK1 and the total concurrent delay OCK2 in the last 1 hour, wherein the current concurrent total delay OCK1 is , OCK2 is G × tpj × at; wherein a is the current concurrent rate, a = bfs(j) / bfsx(j), and at is the total concurrent rate in the last 1 hour, at = bfsp(j) / bfs(j);

[0031] The real-time concurrency performance index tfh(j) is used to measure the real-time response performance of the advertisement by combining the ratio of time-sensitive load time fxt(j) and concurrency delay.

[0032] The beneficial effect of this step is that by analyzing the concurrency load of the advertisement, the recommendation of the advertisement is efficiently managed, avoiding system crashes or response delays caused by excessive concurrency load, and ensuring the stability and smoothness of the advertisement delivery. In addition, by reducing the waiting delay and improving the timeliness of the advertisement and visualizing the quantification of the current real-time concurrency performance index, the advertiser can adjust the delivery strategy according to the performance data of the current real-time concurrency performance index, optimize the budget allocation, and further improve the return on investment of the advertisement.

[0033] Further, in step S400, the optimal advertisement sequence combination is generated according to the real-time concurrency performance index, and the advertisement recommendation is performed according to the optimal advertisement sequence combination.

[0034] The average value of all real-time concurrency performance indexes is calculated and denoted as pjfh, the advertisements with real-time concurrency performance indexes less than pjfh are denoted as optimal preference advertisements, and the advertisement recommendation is performed according to the real-time concurrency performance indexes of the optimal preference advertisements from small to large.

[0035] The beneficial effect of the present application is that by comprehensively evaluating the timeliness and concurrency load of the advertisement, not only the effect and efficiency of the advertisement delivery are improved, but also the stability and user experience of the digital content multi-network channel are enhanced, and the advertisement delivery process is comprehensively optimized.

[0036] To achieve the above-mentioned purpose, the second aspect embodiment of the present application further proposes a short video advertisement recommendation system, which comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a short video advertisement recommendation method when executing the computer program, and the short video advertisement recommendation system runs in a computing device of a satellite, a desktop computer, a notebook computer, a palm computer, and a cloud data center.

[0037] By executing the short video advertisement recommendation method through the short video advertisement recommendation system, the timeliness and concurrency load of the advertisement can be comprehensively evaluated, not only the effect and efficiency of the advertisement delivery are improved, but also the stability and user experience of the digital content multi-network channel are enhanced, and the advertisement delivery process is comprehensively optimized. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The flowchart of a short video advertisement recommendation method is shown.

[0039] Figure 2 The structure diagram of the short video advertisement recommendation system is shown. DETAILED DESCRIPTION

[0040] Embodiments of the present application are described below in detail with reference to the accompanying drawings, wherein the same or similar components are denoted by the same or similar reference numerals throughout the drawings. The embodiments described below are exemplary and are intended to explain the present application, and are not to be understood as limiting the present application.

[0041] Figure 1 A flowchart of a short video advertisement recommendation method is shown.

[0042] With reference to Figure 1 The present application proposes a short video advertisement recommendation method, which comprises the following steps:

[0043] S100, capturing user real-time interaction data;

[0044] S200, obtaining user tags according to the user real-time interaction data;

[0045] S300, obtaining an advertisement sequence according to the user tags and calculating a real-time concurrency performance index of the advertisement;

[0046] S400, generating an optimal advertisement sequence combination according to the real-time concurrency performance index and performing advertisement recommendation according to the optimal advertisement sequence combination.

[0047] According to the advertisement recommendation method of the embodiments of the present application, the user real-time interaction data can be used to obtain the user tags, and the optimal advertisement sequence combination can be generated according to the real-time concurrency performance index.

[0048] Further, in step S100, the user real-time interaction data includes a play completion rate, a video duration, a like behavior or a sharing behavior.

[0049] The play completion rate, the video duration, the like behavior or the sharing behavior of the i-th video watched by the user in the last 24 hours are recorded; pec(i) is used to represent the play completion rate of the i-th video watched by the user in the last 24 hours, time(i) is used to represent the video duration of the i-th video watched by the user in the last 24 hours, wherein i is the serial number of the video, K(i) is used to represent the like behavior or the sharing behavior of the i-th video watched by the user in the last 24 hours, wherein the serial number of the video i=1, 2, …, K, K represents the number of videos, and the total duration TZ of the video duration of the videos watched by the user in the last 24 hours is recorded, Tmax is recorded as the maximum value of the video duration of the videos watched by the user in the last 24 hours, Tmin is recorded as the minimum value of the video duration of the videos watched by the user in the last 24 hours, TK is recorded as the average value of the video duration of the videos watched by the user in the last 24 hours, and OCK is recorded as the average value of the like behavior or the sharing behavior of the videos watched by the user in the last 24 hours.

[0050] Due to the short video ecology, the user interest is affected by the hot events, social communication and time-sensitive content, and the traditional recommendation system relies on static historical portraits such as weekly / monthly behavior statistics, such as the advertisement delivery method proposed in the patent number CN114845170A entitled "Advertisement delivery method, intelligent terminal and computer readable storage medium", which often causes the problem that the historical behavior cannot reflect the current interest focus of the user due to signal lag when the attention is diverted by sudden news, and the problem that it is difficult to establish an effective portrait for new users or low-active users due to data sparseness; in order to solve the above problems, the present application proposes step S200;

[0051] Further, in step S200, the step of obtaining the user label according to the user real-time interaction data comprises:

[0052] S201, calculating the shortest effective preference duration TPHa and the longest effective preference duration TPHb according to the user real-time interaction data;

[0053] Let the interaction adjustment factor KUI be K(i) / (1+OCK+K(i)), and the effective video duration TOI be ; Calculate the first offset duration difference, wherein the first offset duration difference TD1 is TmaxxTOI / TZ; wherein the second offset duration difference TD2 is TminxTOI / TZ; the shortest effective preference duration TPHa is TK-TD1, and the longest effective preference duration TPHb is TK+TD2, to obtain the preference window [TPHa, TPHb];

[0054] Wherein, the effective video duration TOI is used to measure the patience level of the user; [TPHa, TPHb] is the preference window, when the user's patience is low (TOI is small), the preference window shrinks to short duration, otherwise it expands, which is used to obtain the accurate user's preferred video range through curvature adaptive adjustment, and filter the videos that obviously do not meet the user's preference.

[0055] S202, filtering out the videos with a video duration between the shortest effective preference duration TPHa and the longest effective preference duration TPHb according to the shortest effective preference duration TPHa and the longest effective preference duration TPHb, and recording them as preferred videos;

[0056] S203, obtaining the user label according to the video label of the preferred video;

[0057] Obtain the video label with the highest frequency in the preferred video, and match the user label to which the label belongs to obtain the matched user label.

[0058] S300, obtaining the advertisement sequence according to the user label, and calculating the real-time concurrent performance index of the advertisement;

[0059] S310, after filtering the matched advertisements from the database through the user tag of the user, calculating the time load of the waiting time of the matched advertisements sent to other users by the server in the last one hour;

[0060] The method for calculating the time load of the waiting time of the matched advertisements sent to other users by the server in the last one hour is as follows:

[0061] S311, using ty(j) to represent the waiting time of the jth matched advertisement sent to other users in the last one hour, wherein j represents the serial number of each matched advertisement, j = 1, 2, …, G, and G represents the number of matched advertisements; obtaining the waiting time of the matched advertisements in the last one hour after receiving the user request; taking the maximum value of the waiting time of the jth matched advertisement sent to other users in the last one hour as tmax, the average value of the waiting time of the matched advertisements sent to other users in the last one hour as tpj, and the median of the waiting time of the matched advertisements sent to other users in the last one hour as tav;

[0062] S312, calculating the time load fxt(j) of the jth advertisement; wherein fxt(j) is ty(j) plus the difference level ratio between the maximum waiting time and the average waiting time, wherein the difference level ratio between the maximum waiting time and the average waiting time is the difference between tmax and tpj multiplied by the ratio of ) and .

[0063] Wherein, the time load fxt(j) can reflect the difference between the maximum waiting time and the average waiting time, and quantify the response performance of the advertisement in time.

[0064] S320, obtaining the concurrent user number data of the matched advertisements, wherein the concurrent user number data includes the maximum concurrent user number, the current concurrent user number, and the total concurrent user number in the last one hour, and taking the maximum concurrent user number of the jth matched advertisement as bfsx(j), i.e. the maximum number of users that the advertisement can support at the same time; the current concurrent user number of the jth advertisement as bfs(j), i.e. the number of users currently recommended the advertisement; and the total concurrent user number of the jth advertisement in the last one hour as bfsp(j), i.e. the number of users to which the advertisement is sent in the past one hour;

[0065] S330, calculating the real-time concurrent performance index tfh(j) through the concurrent user number data and the time load: tfh(j) is fxt(j) multiplied by the ratio of the current total delay OCK1 and the total concurrent delay OCK2 in the last one hour, wherein the current total delay OCK1 is , OCK2 is Gxtpjxat; wherein a is a current concurrency rate, a = bfs(j) / bfsx(j), at is a total concurrency rate in the last hour, at = bfsP(j) / bfs(j);

[0066] Wherein, the real-time concurrency performance index tfh(j) is used to measure the real-time response performance of the advertisement by combining the time load time fxt(j) and the ratio of concurrency delay.

[0067] S400, generating an optimal advertisement sequence combination according to the real-time concurrency performance index, and performing advertisement recommendation according to the optimal advertisement sequence combination;

[0068] Calculate the average value of all real-time concurrency performance indexes and mark it as pjfh, mark the advertisements with a real-time concurrency performance index less than pjfh as optimal preference advertisements, and perform advertisement recommendation according to the real-time concurrency performance indexes of the optimal preference advertisements from small to large. Figure 2 As shown in the short video advertisement recommendation system structure diagram.

[0069] Referring to Figure 2 The application further provides a short video advertisement recommendation system 20, which comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the steps in the short video advertisement recommendation method are implemented. The short video advertisement recommendation system 20 runs in a computing device of a satellite, a desktop computer, a notebook computer, a palm computer, and a cloud data center.

[0070] The advertisement recommendation system comprises a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the computer program, the computer program runs in the following units of the advertisement recommendation system:

[0071] The acquisition unit 21 is used for capturing user real-time interaction data.

[0072] The conversion unit 22 is used for acquiring user tags according to the user real-time interaction data.

[0073] The loading unit 23 is used for acquiring advertisement sequences according to the user tags and calculating advertisement real-time concurrency performance indexes.

[0074] The management unit 24 is used for generating an optimal advertisement sequence combination according to the real-time concurrency performance indexes and performing advertisement recommendation according to the optimal advertisement sequence combination.

[0075] The short video advertisement recommendation system can run in desktop computers, notebooks, palmtop computers and cloud servers and other computing devices. The short video advertisement recommendation system can run in an advertisement recommendation system which can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the example is only an example of the short video advertisement recommendation system 20, and does not constitute a limitation on the short video advertisement recommendation system 20, and can include more or fewer components, or combine certain components, or different components, for example, the short video advertisement recommendation system can also include input and output devices, network access devices, buses, etc.

[0076] By executing the short video advertisement recommendation method through the short video advertisement recommendation system 20, the timeliness and concurrent load of the advertisement can be comprehensively evaluated, not only to improve the effect and efficiency of the advertisement, but also to enhance the stability and user experience of the digital content multi-network channel, and to optimize the advertisement process.

[0077] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a list of executable instructions for implementing logic functions, which can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute the instructions, or in conjunction with these instructions execution system, apparatus or device. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by or in conjunction with an instruction execution system, apparatus or device. More specific examples (non-exhaustive list) of computer-readable medium include the following: electrical connections having one or more wires (electronic devices), portable computer disk boxes (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), fiber optic devices, and portable compact disc read-only memories (CDROMs). In addition, the computer-readable medium can even be paper or other suitable medium on which the program can be printed, because the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by editing, interpretation or necessary processing, if necessary, in other suitable ways, and then stored in a computer memory.

[0078] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, through software or firmware in storage media which are executed by suitable instruction executing systems. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or their combinations, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0079] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that a specific feature, structure, material or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. Descriptive expressions of the above terms in the present specification do not necessarily refer to the same embodiment or example. Also, the described specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0080] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0081] In addition, the terms "first", "second", and the like used in the embodiments of the present application are only for the purpose of description, and can not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features indicated in the embodiments. Therefore, the features defined with "first", "second" and the like in the embodiments of the present application can be explicitly or implicitly indicated to include at least one of the features in the embodiments. In the description of the present application, the meaning of the word "plurality" is at least two or two or more, such as two, three, four, and the like, unless otherwise specifically limited in the embodiments.

[0082] In the present application, unless otherwise explicitly specified or limited in the embodiments, the terms "mounting", "connecting", "connecting" and "fixing" and the like appearing in the embodiments should be understood broadly, for example, the connection can be fixed connection, or detachable connection, or integral, which can be understood, or mechanical connection, electrical connection, etc. Of course, it can also be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements, or the interaction relationship of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific implementation situation.

[0083] In the present application, unless otherwise explicitly specified and limited, the first feature is "on" or "under" the second feature. The first and second features can be in direct contact, or the first and second features can be indirectly contacted through an intermediate medium. Moreover, the first feature "above", "above" and "above" the second feature can be directly above or obliquely above the first feature, or only indicate that the horizontal height of the first feature is higher than that of the second feature. The first feature "below", "below" and "below" the second feature can be directly below or obliquely below the first feature, or only indicate that the horizontal height of the first feature is less than that of the second feature.

[0084] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1.A short video advertisement recommendation method, characterized in that, The method comprises the following steps: S100, capturing user real-time interaction data; recording the playback completion rate of the i-th video watched by the user in the last 24 hours, the video duration, the number of like behaviors or sharing behaviors; using pec(i) to represent the playback completion rate of the i-th video watched by the user in the last 24 hours, using time(i) to represent the video duration of the i-th video watched by the user in the last 24 hours, wherein i is the serial number of the video, using K(i) to represent the number of like behaviors or sharing behaviors of the i-th video watched by the user in the last 24 hours, wherein the serial number of the video is i=1, 2, …, K, K represents the number of videos, and the total duration TZ of the video duration of the videos watched by the user in the last 24 hours is recorded, the maximum value Tmax of the video duration of the videos watched by the user in the last 24 hours is recorded, the minimum value Tmin of the video duration of the videos watched by the user in the last 24 hours is recorded, the average value TK of the video duration of the videos watched by the user in the last 24 hours is recorded, and the average value OCK of the number of like behaviors or sharing behaviors of the videos watched by the user in the last 24 hours is recorded; S200, obtaining user tags according to the user real-time interaction data; S201, calculating the shortest effective preference duration and the longest effective preference duration according to the user real-time interaction data; Let the interaction adjustment factor KUI be K(i) / (1+OCK+K(i)), and let the effective video duration TOI be ; calculate a first offset duration difference, wherein the first offset duration difference TD1 is TmaxxTOI / TZ; wherein the second offset duration difference TD2 is TminxTOI / TZ; the shortest effective preference duration TPHa is TK-TD1, and the longest effective preference duration TPHb is TK+TD2, to obtain the preference window [TPHa, TPHb]; S202, filtering out videos with a video duration between the shortest effective preference duration and the longest effective preference duration according to the shortest effective preference duration and the longest effective preference duration, and recording them as preference videos; S203, obtaining user tags according to the video tags of the preference videos; S300, obtaining an advertisement sequence according to the user tags, and calculating an advertisement real-time concurrency performance index; S310, after filtering out matched advertisements from the database through the user tags of the user, calculating the time efficiency load time through the waiting time of the matched advertisements sent to other users by the server in the last one hour; S311, using ty(j) to represent the waiting time of the j-th matched advertisement sent to other users in the last one hour, wherein j represents the serial number of each matched advertisement, j=1, 2, …, G, and G represents the number of matched advertisements; obtaining the waiting time of the matched advertisements after receiving the user request in the last one hour; recording the maximum value tmax of the waiting time of the j-th matched advertisement sent to other users in the last one hour, the average value tpj of the waiting time of the matched advertisements sent to other users in the last one hour, and the median value tav of the waiting time of the matched advertisements sent to other users in the last one hour; S312, calculating the time effectiveness load time fxt(j) of the jth advertisement; wherein fxt(j) is ty(j) plus the difference level ratio between the maximum waiting time and the average waiting time, wherein the difference level ratio between the maximum waiting time and the average waiting time is the difference between tmax and tpj multiplied by the ratio of ) and . S320, obtaining the concurrent user number data of the matched advertisements, wherein the concurrent user number data comprises the maximum concurrent user number, the current concurrent user number and the total concurrent user number in the last one hour, and taking the maximum concurrent user number of the jth matched advertisement as bfsx(j), that is, the maximum simultaneous sending user number that the advertisement can support; the current concurrent user number of the jth advertisement as bfs(j), that is, the current recommended user number of the advertisement; and the total concurrent user number of the jth advertisement in the last one hour as bfsp(j), that is, the number of users to which the advertisement is sent in the past one hour; S330, calculating real-time concurrent performance index tfh(j) through concurrent user number data and time-sensitive load time: tfh(j) is the ratio of fxt(j) multiplied by current concurrent total delay OCK1 and total concurrent delay OCK2 in the last one hour, wherein the current concurrent total delay OCK1 is , OCK2 is G*tpj*at; wherein a is the current concurrent rate, a=bfs(j) / bfsx(j), at is the total concurrent rate in the last one hour, at=bfsp(j) / bfs(j); S400, generating an optimal advertisement sequence combination according to the real-time concurrent performance index, and performing advertisement recommendation according to the optimal advertisement sequence combination. 2.A short video advertisement recommendation system, characterized in that, The short video advertisement recommendation system comprises a processor, a memory, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the short video advertisement recommendation method in claim 1 are implemented. The short video advertisement recommendation system runs in a desktop computer, a notebook computer, a palm computer or a cloud data center.

Images