Intelligent pushing method and system based on dynamic interest attenuation analysis

Abstract

The invention provides an intelligent pushing method and system based on dynamic interest attenuation analysis, and relates to the technical field of data pushing.The method comprises the steps that when a target user triggers a video recommendation request, a first recommended video list is determined based on the video recommendation request and historical watching records of the target user; obtaining a first list watching feature of the target user for the first recommended video list, performing dynamic interest attenuation analysis based on the first list watching feature, and determining a second recommended video list according to an interest attenuation analysis result; and recommending the second recommended video list to the user side of the target user. The technical problem of deviation between the short video recommendation content and the real-time interest of the user in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data push, and in particular to an intelligent push method and system based on dynamic interest decay analysis. Background Art

[0002] As users continuously watch short videos, their interest in specific content categories tends to naturally wane over time. However, traditional short video recommendation algorithms rely on historical user behavior data to build static recommendation models, making it difficult to capture the dynamic evolution of user interests in real time. Consequently, they lack effective perception and response to this waning interest trend, often leading to discrepancies between recommended content and users' real-time interests. Summary of the Invention

[0003] The present invention addresses the technical problem in the prior art that short video recommendation content deviates from the user's real-time interests, and provides an intelligent push method and system based on dynamic interest decay analysis.

[0004] The technical solution of the present invention to solve the above technical problems is as follows: In a first aspect, the present invention provides an intelligent push method based on dynamic interest decay analysis, comprising: When a target user triggers a video recommendation request, determining a first recommended video list based on the video recommendation request and the target user's historical viewing record; obtaining a first list viewing feature of the target user for the first recommended video list, performing a dynamic interest decay analysis based on the first list viewing feature, and determining a second recommended video list according to the interest decay analysis result; The second recommended video list is recommended to the user terminal of the target user.

[0005] In a second aspect, the present invention provides an intelligent push system based on dynamic interest decay analysis, comprising: A first recommendation module is configured to, when a target user triggers a video recommendation request, determine a first recommended video list based on the video recommendation request and the target user's historical viewing history; an interest decay analysis module, configured to obtain a first list viewing feature of the target user for the first recommended video list, perform a dynamic interest decay analysis based on the first list viewing feature, and determine a second recommended video list according to the interest decay analysis result; The second recommendation module is used to recommend the second recommended video list to the user terminal of the target user.

[0006] The beneficial effects of the present invention are: Compared with the existing technology, this application first determines the first recommended video list based on the video recommendation request and the historical viewing record of the target user when the target user triggers a video recommendation request. The first recommended video list is in line with the historical preferences of the target user, providing a reliable basis for the subsequent dynamic interest decay analysis through real-time feedback. Secondly, the viewing characteristics of the first list of the first recommended video list by the target user are obtained, and a dynamic interest decay analysis is performed based on the viewing characteristics of the first list. The second recommended video list is determined according to the results of the interest decay analysis. The current interest of the target user is quantified by the ratio of the current average viewing time to the historical average viewing time. The interest weight is updated accordingly to generate the second recommended video list, realizing the transformation of the recommendation strategy from static history-driven to dynamic real-time feedback-driven, ensuring that the recommended content is always synchronized with the user's current interests. Finally, the second recommended video list is recommended to the user end of the target user to ensure that the recommended video received by the user is the result of real-time optimization and fits the current interest, thereby improving the user's acceptance of the recommended content and viewing experience.

[0007] Through the above technical solution, this application first generates a first list of recommended videos based on the target user's historical viewing history. It then records the target user's real-time viewing characteristics of the first list of recommended videos, analyzes the difference between current and historical interests, quantifies the trend of interest decay or increase, and then adjusts the interest weights to generate and push a second list of recommended videos that better matches the user's real-time preferences. This achieves a precise match between recommended content and the user's current interests, reduces recommendation bias, and improves the user's viewing experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 A schematic diagram of the process of the intelligent push method based on dynamic interest decay analysis provided by the present invention; Figure 2 This is a structural diagram of the intelligent push system based on dynamic interest decay analysis provided by the present invention.

[0009] In the accompanying drawings, the components represented by the reference numerals are as follows: A first recommendation module 11 , an interest decay analysis module 12 , and a second recommendation module 13 . DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0011] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the specified features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0012] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or illustration". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0013] Example 1, as Figure 1 As shown, an embodiment of the present invention provides an intelligent push method based on dynamic interest decay analysis, including: S10: When a target user triggers a video recommendation request, a first recommended video list is determined based on the video recommendation request and the target user's historical viewing record.

[0014] The target user's historical viewing records often contain their past interest preferences, including preferred content categories, topics of interest, and other historical preference characteristics. By extracting and analyzing these historical viewing records, we can identify the target user's interest categories and corresponding interest weights, and then generate a first list of recommended videos that match their past interest characteristics.

[0015] To address the above issues, this application triggers a video recommendation request when the target user clicks on any video in the short video application to enter the video playback interface. This binds the video recommendation request to the user's active interaction, ensuring that the initiation of the recommendation process is synchronized with the user's actual viewing behavior. Furthermore, when the target user triggers a video recommendation request, a first list of recommended videos is determined based on the video recommendation request and the target user's historical viewing history.

[0016] Specifically, step S10 in the method includes: Connecting to a full video pool according to the video recommendation request, wherein the full video pool includes all recommended videos; extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and an interest weight of each of the user interest categories based on the historical viewing features; Videos are selected from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list.

[0017] In an embodiment of the present application, first, the full video pool is connected according to the video recommendation request. The full video pool contains all the recommendable videos of the short video application platform to ensure that the sources of the recommended content are fully covered.

[0018] Secondly, historical viewing features are extracted based on the target user's historical viewing records, and multiple user interest categories and interest weights of each user interest category are determined based on the historical viewing features. Among them, interest categories refer to video categories such as food, travel, and fitness, and the interest weight of each user interest category is calculated. The interest weight can reflect the target user's preference for the interest category. The higher the interest weight, the stronger the target user's preference for the interest category in the past.

[0019] Finally, based on multiple user interest categories and the interest weights of each user interest category, videos are selected from the entire video pool to obtain the first recommended video list. The first recommended video list is the initial recommendation result generated based on the historical preferences of the target user, providing basic data for subsequent dynamic interest decay analysis.

[0020] Specifically, the “extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and an interest weight of each of the user interest categories based on the historical viewing features” includes: Traversing the historical viewing records and extracting a first historical viewing record; When the viewing time of the first historical viewing record exceeds a preset time threshold, obtaining a first video category of the first historical viewing record; Add 1 to the occurrence frequency of the first video category in all video categories according to the first video category; After processing the historical viewing records, the video categories whose appearance frequency is not 0 in the full video categories are used as multiple user interest categories of the target user; The ratio of the occurrence frequency of each user interest category to the total occurrence frequency of all user interest categories is calculated as the interest weight of the corresponding user interest category.

[0021] In the embodiment of the present application, the historical viewing records are first traversed to extract the first historical viewing record. For example, 100 historical viewing records of the target user are traversed, and detailed information of each historical viewing record is read one by one, including video ID, viewing time, etc.

[0022] Secondly, when the viewing time of the first historical viewing record exceeds the preset time threshold, the first video category of the first historical viewing record is obtained, wherein the preset time threshold can be flexibly set according to the actual scenario, such as 10 seconds, 15 seconds, etc., and is mainly used to filter invalid viewing behaviors such as accidental clicks and second slides. The video category can be determined based on the classification system preset by the short video application platform, such as food, travel, fitness, music, cute pets, etc. For example, if the preset time threshold is 10 seconds, traverse the 100 historical viewing records of the target user, and filter out 60 records with a viewing time of more than 10 seconds as valid historical viewing records, and extract the video categories corresponding to the valid historical viewing records, for example, 30 corresponding to food, 20 corresponding to travel, and 10 corresponding to fitness. In this way, food, travel, and fitness are used as the first video category.

[0023] Next, the frequency of occurrence of the first video category in the full video category is increased by 1. For example, if the frequencies of food, travel, and fitness in the full video category are initially 12, 8, and 3, respectively, and based on 60 valid historical viewing records (30 food, 20 travel, and 10 fitness), the frequency of occurrence of the first video category is accumulated, and the frequency of food is 12 + 30 = 42 times, the frequency of travel is 8 + 20 = 28 times, and the frequency of fitness is 3 + 10 = 13 times.

[0024] After processing the historical viewing history, the video categories with a non-zero frequency of occurrence in the total video categories are used as the target user's multiple interest categories. For example, after the above processing, the frequencies of food, travel, and fitness are 42, 28, and 13 times, respectively. These three categories are then determined as the target user's interest categories.

[0025] Finally, the ratio of the frequency of each user's interest category to the total frequency of all user interest categories is calculated as the interest weight of the corresponding user interest category. For example, the total frequency of food, travel, and fitness is 42 + 28 + 13 = 83. Then, the interest weight of food = 42 / 83 ≈ 0.506, the interest weight of travel = 28 / 83 ≈ 0.337, and the interest weight of fitness = 13 / 83 ≈ 0.157. In this way, invalid behavior is filtered out by the preset duration threshold, and the user's preference for different categories is quantified by the frequency ratio, ensuring that the interest categories and weights truly reflect the user's historical and stable preferences and avoiding the interference of accidental behavior on the recommendation logic.

[0026] Furthermore, the step of “selecting videos from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list” includes: Get the number of videos in the preset recommendation list; Obtaining a preset number of recommendations for each user interest category based on the number of videos in the preset recommendation list and the interest weight of each user interest category; Based on the preset recommendation quantity of each user interest category, videos are randomly selected from the corresponding video categories in the full video pool to obtain the first recommended video list.

[0027] In this embodiment of the present application, the preset number of videos in the recommendation list is first obtained. The preset number of videos in the recommendation list is the total number of recommended videos per batch, such as 20, that the short video application platform pre-sets based on factors such as the user's historical browsing data. This is to control the length of the recommendation list and avoid overloading the user with choices due to too many videos, or insufficient content supply due to too few videos. For example, if the short video application platform sets a limit of 20 videos per batch, the preset number of videos in the recommendation list is 20.

[0028] Next, based on the number of videos in the preset recommendation list and the interest weight of each user's interest category, the preset number of recommendations for each user's interest category is calculated: Preset number of recommendations = Preset number of videos in the recommendation list × interest weight. The calculated result is fine-tuned by rounding, rounding up, and rounding down to ensure that the sum of the preset number of recommendations for each user's interest category equals the number of videos in the preset recommendation list. For example, if the preset number of videos in the recommendation list is 20, the interest weight for food is 0.506, the interest weight for travel is 0.337, and the interest weight for fitness is 0.157, then the preset number of recommendations for food = 20 × 0.506 ≈ 10 (rounded off), the preset number of recommendations for travel = 20 × 0.337 ≈ 7 (rounded off), and the preset number of recommendations for fitness = 20 × 0.157 ≈ 3 (rounded off). The total number of preset recommendations for food, travel, and fitness is 20, which is consistent with the number of videos in the preset recommendation list. In this way, the interest weight is converted into a specific preset number of recommendations, ensuring that the proportion of recommended content matches the user's historical preferences.

[0029] Finally, based on the preset number of recommendations for each user's interest category, videos are randomly selected from the corresponding video category in the full video pool to obtain the first recommended video list. Specifically, after determining the preset number of recommendations for each interest category, videos are randomly selected from the video subset of the corresponding video category in the full video pool and combined to form the first recommended video list. The purpose of random selection is to avoid homogenization of recommended content and ensure that users see more diverse content. For example, in the full video pool, the food category contains 1,000 videos, the travel category contains 800 videos, and the fitness category contains 500 videos. According to the preset number of recommendations, 10 videos are randomly selected from the food video subset, 7 videos are randomly selected from the travel video subset, and 3 videos are randomly selected from the fitness video subset. These 20 videos are randomly arranged or arranged alternately according to interest categories to form the first recommended video list. The first recommended video list is the initial recommendation result generated based on the historical preferences of the target user.

[0030] In summary, compared to existing technologies, when a target user triggers a video recommendation request, this application determines a first list of recommended videos based on the video recommendation request and the target user's historical viewing history. This extracts interest features based on historically valid behaviors, ultimately generating a first list of recommended videos that matches the target user's historical preferences. This provides a reliable foundation for subsequent dynamic interest decay analysis through real-time feedback.

[0031] S20: Obtain a first list viewing feature of the target user for the first recommended video list, perform a dynamic interest decay analysis based on the first list viewing feature, and determine a second recommended video list according to the interest decay analysis result.

[0032] While watching videos, users' interests may change dynamically based on factors such as exposure to content and viewing time. Traditional recommendation methods, which rely on historical user behavior to build static recommendation models, struggle to capture these dynamic changes in real time. This often leads to continued push of content that has waned in users' interest, resulting in recommendations that are out of sync with their current interests.

[0033] In response to the above problem, the present application obtains the target user's first list viewing characteristics of the first recommended video list, performs dynamic interest decay analysis based on the first list viewing characteristics, and determines the second recommended video list according to the interest decay analysis results.

[0034] Specifically, step S20 in the method includes: Continuously recording the viewing time of each video in the first recommended video list by the target user to obtain viewing characteristics of the first list; Performing user interest analysis on the viewing features of the first list to obtain user current interest features; Retrieve the user's historical interest features, perform a dynamic interest decay analysis in combination with the user's current interest features, and determine a second recommended video list based on the interest decay analysis results.

[0035] In this embodiment of the present application, the target user's viewing duration of each video in the first recommended video list is first continuously recorded to obtain a first list viewing feature. For example, as the target user watches videos in the first recommended video list, the viewing duration of each video is continuously recorded in real time. For example, the target user may spend 25 seconds watching the first food video in the first recommended video list, 40 seconds watching the second travel video, and so on. These scattered duration data are then aggregated in sequence, such as {25 seconds, 40 seconds, ...}, to obtain the first list viewing feature, which can reflect the target user's real-time viewing interest during the viewing process.

[0036] Next, we perform a user interest analysis on the viewing characteristics of the first list to obtain the user's current interest characteristics. Specifically, we calculate the average viewing time of all videos in each user interest category in the first recommended video list to obtain the average viewing time of multiple user interest categories as the user's current interest characteristics. The user's current interest characteristics can reflect the user's current preference for each interest category.

[0037] Finally, the user's historical interest characteristics are retrieved and combined with the user's current interest characteristics to perform a dynamic interest decay analysis. The second list of recommended videos is determined based on the results of the interest decay analysis. Specifically, the ratio of the average viewing time of the current user's interest category to the average historical viewing time is calculated to obtain the interest adjustment coefficient for the current user's interest category. The corresponding interest weight is adjusted accordingly to obtain the adjusted interest weight for each user's interest category as the interest decay analysis result. The second list of recommended videos is then determined based on the interest decay analysis results.

[0038] Specifically, the “performing a user interest analysis on the viewing features of the first list to obtain the user's current interest features” includes: obtaining a first user interest category from the plurality of user interest categories; Calculating the average viewing time of all videos in the first recommended video list belonging to the first user interest category to obtain the average viewing time of the first category; The average viewing time of the remaining user interest categories is obtained in the same manner as that of the first category, and the average viewing time of the plurality of user interest categories is obtained as the user's current interest feature.

[0039] In the embodiment of the present application, a first user interest category is first obtained from multiple user interest categories. For example, an interest category, such as food, is randomly selected from multiple user interest categories of the target user as the first user interest category.

[0040] Next, the average viewing time of all videos in the first recommended video list belonging to the first user interest category is calculated to obtain the average viewing time for the first category. For example, the viewing time of all videos in the first recommended video list belonging to the first user interest category is counted. For example, if the target user watches 10 food videos in the first recommended video list for 25 seconds, 30 seconds, 20 seconds, and so on, respectively, and the total viewing time is 250 seconds, then the average viewing time = 250 / 10 = 25 seconds, which is the average viewing time for the first category.

[0041] Finally, using the same method for obtaining the average viewing time for the first category, we obtain the average viewing time for the remaining user interest categories, resulting in the average viewing time for multiple user interest categories as the user's current interest profile. For example, the average viewing time for the remaining user interest categories is calculated using the same method. For example, the average viewing time for seven travel videos is 40 seconds, and the average viewing time for three fitness videos is 20 seconds. These constitute the user's current interest profile: the average viewing time for food is 25 seconds, the average viewing time for travel is 40 seconds, and the average viewing time for fitness is 20 seconds. In this way, the average viewing time quantifies the user's current interest. A longer average viewing time indicates a higher user interest in videos in that category; conversely, a lower average viewing time indicates a decrease in interest.

[0042] Furthermore, the “retrieving the user's historical interest characteristics, performing a dynamic interest decay analysis in combination with the user's current interest characteristics, and determining a second recommended video list based on the interest decay analysis results” includes: Retrieving a user's historical interest characteristics, wherein the user's historical interest characteristics include an average historical viewing time of each of the user's interest categories; Traversing the user's current interest features, obtaining the average viewing time of the current user's interest category, and obtaining the historical average viewing time of the current user's interest category from the user's historical interest features; Calculating the ratio of the average viewing time of the current user interest category to the average historical viewing time to obtain an interest adjustment coefficient for the current user interest category; Obtaining the interest adjustment coefficient of each user interest category according to a method of obtaining the interest adjustment coefficient of the current user interest category; adjusting the corresponding interest weight based on the interest adjustment coefficient of each user interest category to obtain the adjusted interest weight of each user interest category as the interest decay analysis result; A second recommended video list is determined according to the adjusted interest weights of the user interest categories.

[0043] In the embodiment of the present application, the user's historical interest features are first retrieved. The user's historical interest features include the average historical viewing time of each user's interest category. For example, the target user's historical average viewing time for each user's interest category over a period of time, such as the last three days, a week, or a month, is extracted. For example, the target user's historical viewing time for food, travel, and fitness over the last week is averaged to 30 seconds, 25 seconds, and 25 seconds, respectively. This historical interest feature can reflect the target user's stable preference for different interest categories over a period of time.

[0044] Secondly, the user's current interest features are traversed to obtain the average viewing time of the current user's interest category, and the historical average viewing time of the current user's interest category is obtained from the user's historical interest features. For example, the user's current interest features are traversed to obtain the average viewing time of the current user's interest category. For example, the average viewing time of food is 25 seconds, the average viewing time of travel is 40 seconds, and the average viewing time of fitness is 20 seconds. The corresponding historical average viewing time of food, travel, and fitness is obtained from the user's historical interest features as 30 seconds, 25 seconds, and 25 seconds.

[0045] Next, calculate the ratio of the average viewing time of the current user's interest category to the average viewing time of all previous users to obtain the interest adjustment coefficient for the current user's interest category. For example, the current interest adjustment coefficient for food is 25 / 30 = 0.83, the current interest adjustment coefficient for travel is 40 / 25 = 1.6, and the current interest adjustment coefficient for fitness is 20 / 25 = 0.8. If the interest adjustment coefficient is greater than 1, it indicates that the current interest in the interest category is increasing. If the interest adjustment coefficient is approximately 1, it indicates that the current interest in the interest category is basically stable. If the interest adjustment coefficient is less than 1, it indicates that the current interest in the interest category is decreasing.

[0046] Furthermore, the corresponding interest weights are adjusted based on the interest adjustment coefficients for each user's interest category, resulting in the adjusted interest weights for each user's interest category, which serve as the interest decay analysis results. For example, if the interest weights for each user's interest category are 0.506 for food, 0.337 for travel, and 0.157 for fitness, then the adjusted interest weight for food is 0.506 × 0.83 = 0.42, the adjusted interest weight for travel is 0.337 × 1.6 = 0.54, and the adjusted interest weight for fitness is 0.157 × 0.8 = 0.13. These interest decay analysis results can reflect the real-time status and evolutionary trends of the target user's interests as they dynamically change.

[0047] Finally, a second list of recommended videos is determined based on the adjusted interest weights of each user interest category. For example, the same method as in S10 is used to calculate the product of the number of videos in the preset recommendation list and the adjusted interest weights of each user interest category. The result is fine-tuned by rounding, rounding up, or rounding down to obtain the number of recommendations for each user interest category. Based on the number of recommendations for each user interest category, videos are randomly selected from the corresponding video category in the full video pool to obtain the second list of recommended videos. For example, if the preset number of recommended videos is 20, then the number of food recommendations = 20 × 0.42 ≈ 8 (rounded down), the number of travel recommendations = 20 × 0.54 ≈ 10 (rounded down), and the number of fitness recommendations = 20 × 0.13 ≈ 2 (rounded down). Then, videos are randomly selected from the corresponding video categories in the full video pool. For example, 8 videos are randomly selected from the food video subset, 10 videos are randomly selected from the travel video subset, and 2 videos are randomly selected from the fitness video subset. These 20 videos are randomly arranged or arranged alternately by interest category to form a second recommended video list. This second recommended video list is generated based on the target user's current preferences. In this way, the video recommendation strategy can be dynamically adjusted according to the real-time status of the target user's interests, increasing the number of recommendations for categories with increasing interest and reducing the number of recommendations for categories with decreasing interest.

[0048] In summary, compared to existing technologies, this application obtains the target user's viewing characteristics for the first list of recommended videos, performs a dynamic interest decay analysis based on the viewing characteristics of the first list, and determines the second recommended video list based on the results of the interest decay analysis. In this way, the target user's current interest is quantified by the ratio of the current average viewing time to the historical average viewing time. The interest weight is updated accordingly to generate the second recommended video list, achieving a transition from a static history-driven recommendation strategy to a dynamic real-time feedback-driven recommendation strategy, ensuring that the recommended content always stays in sync with the user's current interests.

[0049] S30: Recommending the second recommended video list to the user terminal of the target user.

[0050] The above steps generate a second recommended video list through dynamic interest decay analysis. The second recommended video list is an optimization result obtained after dynamic adjustment based on the user's real-time viewing behavior of the first recommended video list and historical interest characteristics. Compared with the initial recommendation, it can more accurately reflect the current interest preferences of the target user.

[0051] Therefore, in an embodiment of the present application, the second recommended video list is recommended to the user end of the target user, where the user end specifically refers to the terminal carrier actually used by the target user, such as the short video APP interface on a mobile phone, tablet and other devices, or the video playback page on the web page.

[0052] In this way, it can ensure that the recommended content received by users is the result of real-time interest optimization and highly matches current preferences, thereby improving users' acceptance of recommended content and overall viewing experience.

[0053] In summary, the embodiments of the present application have at least the following technical effects: Compared to existing technologies, this application first determines a first list of recommended videos based on the target user's video recommendation request and their historical viewing history when the target user triggers a video recommendation request. This extracts interest features based on historically valid behaviors, ultimately generating a first list of recommended videos that matches the target user's historical preferences. This provides a reliable foundation for subsequent dynamic interest decay analysis using real-time feedback.

[0054] Secondly, this application obtains the target user's viewing characteristics for the first list of recommended videos, performs a dynamic interest decay analysis based on the viewing characteristics of the first list, and determines a second list of recommended videos based on the results of the interest decay analysis. In this way, the target user's current interest is quantified by the ratio of the current average viewing time to the historical average viewing time. The interest weight is updated accordingly to generate the second list of recommended videos. This achieves a transition from a static history-driven recommendation strategy to a dynamic real-time feedback-driven recommendation strategy, ensuring that the recommended content always stays in sync with the user's current interests.

[0055] Finally, the present application recommends the second recommended video list to the target user's client, thereby ensuring that the recommended videos received by the user are the result of real-time optimization and fit the current interests, thereby improving the user's acceptance of the recommended content and viewing experience.

[0056] Through the above technical solution, this application first generates a first list of recommended videos based on the target user's historical viewing history. It then records the target user's real-time viewing characteristics of the first list of recommended videos, analyzes the difference between current and historical interests, quantifies the trend of interest decay or increase, and then adjusts the interest weights to generate and push a second list of recommended videos that better matches the user's real-time preferences. This achieves a precise match between recommended content and the user's current interests, reduces recommendation bias, and improves the user's viewing experience.

[0057] Example 2, as Figure 2 As shown, based on the same inventive concept as the intelligent push method based on dynamic interest decay analysis provided in Example 1, an embodiment of the present invention further provides an intelligent push system based on dynamic interest decay analysis, including: A first recommendation module 11 is configured to, when a target user triggers a video recommendation request, determine a first recommended video list based on the video recommendation request and the target user's historical viewing history; an interest decay analysis module 12, configured to obtain a first list viewing feature of the target user for the first recommended video list, perform a dynamic interest decay analysis based on the first list viewing feature, and determine a second recommended video list according to the interest decay analysis result; The second recommendation module 13 is configured to recommend the second recommended video list to the user terminal of the target user.

[0058] The first recommendation module 11 is specifically configured to: The video recommendation request is triggered when the target user clicks on any video in the short video application to enter the video playback interface.

[0059] Specifically, the “determining a first recommended video list based on the video recommendation request and the target user's historical viewing history” includes: Connecting to a full video pool according to the video recommendation request, wherein the full video pool includes all recommended videos; extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and an interest weight of each of the user interest categories based on the historical viewing features; Videos are selected from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list.

[0060] Furthermore, the “extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and an interest weight of each of the user interest categories based on the historical viewing features” includes: Traversing the historical viewing records and extracting a first historical viewing record; When the viewing time of the first historical viewing record exceeds a preset time threshold, obtaining a first video category of the first historical viewing record; Add 1 to the occurrence frequency of the first video category in all video categories according to the first video category; After processing the historical viewing records, the video categories whose appearance frequency is not 0 in the full video categories are used as multiple user interest categories of the target user; The ratio of the occurrence frequency of each user interest category to the total occurrence frequency of all user interest categories is calculated as the interest weight of the corresponding user interest category.

[0061] Furthermore, the step of “selecting videos from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list” includes: Get the number of videos in the preset recommendation list; Obtaining a preset number of recommendations for each user interest category based on the number of videos in the preset recommendation list and the interest weight of each user interest category; Based on the preset recommendation quantity of each user interest category, videos are randomly selected from the corresponding video categories in the full video pool to obtain the first recommended video list.

[0062] The interest decay analysis module 12 is specifically configured to: Continuously recording the viewing time of each video in the first recommended video list by the target user to obtain viewing characteristics of the first list; Performing user interest analysis on the viewing features of the first list to obtain user current interest features; Retrieve the user's historical interest features, perform a dynamic interest decay analysis in combination with the user's current interest features, and determine a second recommended video list based on the interest decay analysis results.

[0063] Specifically, the “performing a user interest analysis on the viewing features of the first list to obtain the user's current interest features” includes: obtaining a first user interest category from the plurality of user interest categories; Calculating the average viewing time of all videos in the first recommended video list belonging to the first user interest category to obtain the average viewing time of the first category; The average viewing time of the remaining user interest categories is obtained in the same manner as that of the first category, and the average viewing time of the plurality of user interest categories is obtained as the user's current interest feature.

[0064] Furthermore, the “retrieving the user's historical interest characteristics, performing a dynamic interest decay analysis in combination with the user's current interest characteristics, and determining a second recommended video list based on the interest decay analysis results” includes: Retrieving a user's historical interest characteristics, wherein the user's historical interest characteristics include an average historical viewing time of each of the user's interest categories; Traversing the user's current interest features, obtaining the average viewing time of the current user's interest category, and obtaining the historical average viewing time of the current user's interest category from the user's historical interest features; Calculating the ratio of the average viewing time of the current user interest category to the average historical viewing time to obtain an interest adjustment coefficient for the current user interest category; Obtaining the interest adjustment coefficient of each user interest category according to a method of obtaining the interest adjustment coefficient of the current user interest category; adjusting the corresponding interest weight based on the interest adjustment coefficient of each user interest category to obtain the adjusted interest weight of each user interest category as the interest decay analysis result; A second recommended video list is determined according to the adjusted interest weights of the user interest categories.

[0065] The second recommendation module 13 is specifically configured to: The second recommended video list is recommended to the user terminal of the target user.

[0066] In summary, the embodiments of the present application have at least the following technical effects: Compared to the existing technology, this application first uses a first recommendation module. When a target user triggers a video recommendation request, it determines a first recommended video list based on the video recommendation request and the target user's historical viewing history. The first recommended video list conforms to the target user's historical preferences, providing a reliable foundation for subsequent dynamic interest decay analysis through real-time feedback. Secondly, through the interest decay analysis module, the target user's viewing characteristics of the first list of recommended videos are obtained. Dynamic interest decay analysis is performed based on the first list viewing characteristics. The second recommended video list is determined based on the interest decay analysis results. The target user's current interest is quantified by the ratio of the current average viewing time to the historical average viewing time. The interest weight is updated accordingly to generate the second recommended video list. This achieves a transition from a static history-driven recommendation strategy to a dynamic real-time feedback-driven recommendation strategy, ensuring that the recommended content always keeps pace with the user's current interests. Finally, through the second recommendation module, the second recommended video list is recommended to the target user's user terminal, ensuring that the recommended videos received by the user are the result of real-time optimization and fit the current interests, thereby improving the user's acceptance of the recommended content and viewing experience. In this way, the recommended content is accurately matched with the user's current interests, reducing recommendation bias and improving the user's viewing experience.

[0067] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0068] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0069] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0070] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0071] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0072] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0073] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An intelligent push method based on dynamic interest decay analysis is characterized by: The method comprises: When a target user triggers a video recommendation request, determining a first recommended video list based on the video recommendation request and the target user's historical viewing record; obtaining a first list viewing feature of the target user for the first recommended video list, performing a dynamic interest decay analysis based on the first list viewing feature, and determining a second recommended video list according to the interest decay analysis result; The second recommended video list is recommended to the user terminal of the target user.

2. The method according to claim 1, characterized in that The video recommendation request is triggered when the target user clicks on any video in the short video application to enter the video playback interface.

3. The method according to claim 1, characterized in that Determining a first recommended video list based on the video recommendation request and the target user's historical viewing record includes: Connecting to a full video pool according to the video recommendation request, wherein the full video pool includes all recommended videos; extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and an interest weight of each of the user interest categories based on the historical viewing features; Videos are selected from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list.

4. The method according to claim 3, characterized in that Extracting historical viewing features according to the historical viewing records of the target user, and determining a plurality of user interest categories and interest weights of the user interest categories based on the historical viewing features, including: Traversing the historical viewing records and extracting a first historical viewing record; When the viewing time of the first historical viewing record exceeds a preset time threshold, obtaining a first video category of the first historical viewing record; Add 1 to the occurrence frequency of the first video category in all video categories according to the first video category; After processing the historical viewing records, the video categories whose appearance frequency is not 0 in the full video categories are used as multiple user interest categories of the target user; The ratio of the occurrence frequency of each user interest category to the total occurrence frequency of all user interest categories is calculated as the interest weight of the corresponding user interest category.

5. The method according to claim 4, characterized in that Selecting videos from the full video pool based on multiple user interest categories and interest weights of the user interest categories to obtain the first recommended video list includes: Get the number of videos in the preset recommendation list; Obtaining a preset number of recommendations for each user interest category based on the number of videos in the preset recommendation list and the interest weight of each user interest category; Based on the preset recommendation quantity of each user interest category, videos are randomly selected from the corresponding video categories in the full video pool to obtain the first recommended video list.

6. The method according to claim 4, characterized in that Obtaining a first list viewing feature of the target user for the first recommended video list, performing a dynamic interest decay analysis based on the first list viewing feature, and determining a second recommended video list according to the interest decay analysis result, including: Continuously recording the viewing time of each video in the first recommended video list by the target user to obtain viewing characteristics of the first list; Performing user interest analysis on the viewing features of the first list to obtain user current interest features; Retrieve the user's historical interest features, perform a dynamic interest decay analysis in combination with the user's current interest features, and determine a second recommended video list based on the interest decay analysis results.

7. The method according to claim 6, characterized in that Performing user interest analysis on the viewing features of the first list to obtain the user's current interest features includes: obtaining a first user interest category from the plurality of user interest categories; Calculating the average viewing time of all videos in the first recommended video list belonging to the first user interest category to obtain the average viewing time of the first category; The average viewing time of the remaining user interest categories is obtained in the same manner as that of the first category, and the average viewing time of the plurality of user interest categories is obtained as the user's current interest feature.

8. The method according to claim 7, characterized in that Retrieving the user's historical interest characteristics, performing a dynamic interest decay analysis in combination with the user's current interest characteristics, and determining a second recommended video list based on the interest decay analysis results, including: Retrieving a user's historical interest characteristics, wherein the user's historical interest characteristics include an average historical viewing time of each of the user's interest categories; Traversing the user's current interest features, obtaining the average viewing time of the current user's interest category, and obtaining the historical average viewing time of the current user's interest category from the user's historical interest features; Calculating the ratio of the average viewing time of the current user interest category to the average historical viewing time to obtain an interest adjustment coefficient for the current user interest category; Obtaining the interest adjustment coefficient of each user interest category according to a method of obtaining the interest adjustment coefficient of the current user interest category; adjusting the corresponding interest weight based on the interest adjustment coefficient of each user interest category to obtain the adjusted interest weight of each user interest category as the interest decay analysis result; A second recommended video list is determined according to the adjusted interest weights of the user interest categories.

9. Intelligent push system based on dynamic interest decay analysis, characterized by: Used to perform the method according to any one of claims 1 to 8, comprising: A first recommendation module is configured to, when a target user triggers a video recommendation request, determine a first recommended video list based on the video recommendation request and the target user's historical viewing history; an interest decay analysis module, configured to obtain a first list viewing feature of the target user for the first recommended video list, perform a dynamic interest decay analysis based on the first list viewing feature, and determine a second recommended video list according to the interest decay analysis result; The second recommendation module is used to recommend the second recommended video list to the user terminal of the target user.

Images