Method and system for personalized content recommendation based on multi-dimensional user portrait
By using a multi-dimensional user profiling method to dynamically update user interests, the problem of untimely reflection of interest changes in existing technologies is solved, enabling timely matching and improved credibility of personalized content recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- BEIJING YUYUN TECHNOLOGY CO LTD
- Filing Date
- 2026-03-25
- Publication Date
- 2026-06-05
AI Technical Summary
Existing content recommendation technologies fail to reflect the dynamic changes in user interests in a timely manner, resulting in a lack of personalized recommendation results and an inability to effectively integrate multiple features and information.
By employing a multi-dimensional user profiling approach and pre-setting an interest update cycle, user interest is dynamically updated by combining user browsing information, interaction information, and social network data. This includes enthusiasm, preference value, activity level, preference similarity, and influence, thereby refining user interest to improve personalized recommendations.
It enables timely matching of recommendation results with users' current interests, improves the personalization and credibility of recommendations, dynamically adapts to changes in user interests, and reduces the impact of short-term fluctuations.
Smart Images
Figure CN122153166A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of content recommendation technology, specifically to a personalized content recommendation method and system based on multi-dimensional user profiles. Background Technology
[0002] With the popularization of the internet and the continuous advancement of information technology, massive amounts of information are being produced, disseminated, and accumulated rapidly. Whether it's news, product information, entertainment content, or academic resources, all are experiencing explosive growth. Faced with this deluge of information, users struggle to quickly and accurately find the content they truly need and are interested in, making information overload a widespread problem.
[0003] In existing content recommendation technologies, user profiles do not fully consider the timeliness of user interests and their susceptibility to social influences. This results in delays in updating user profiles, failing to reflect dynamic changes in user interests in a timely manner. Consequently, recommendation algorithms struggle to effectively integrate multiple features and information, leading to a lack of personalization in recommendation results. Summary of the Invention
[0004] In light of the above, it is necessary to provide a personalized content recommendation method and system based on multi-dimensional user profiles. Compared with traditional personalized content recommendation methods based on multi-dimensional user profiles, this method improves the personalization of recommendations and ensures that the recommendation results always match the user's current interests. In a first aspect, embodiments of this application provide a personalized content recommendation method based on multi-dimensional user profiles, the method comprising the following steps: For a single user, the update cycle for the user's interest in various fields is preset, and each cycle is divided into multiple intervals. The interest level is updated at the beginning of the current period based on browsing and interaction information from the previous period. Specifically: By measuring the quantity and duration of content viewed by users in each field within each time period, we can obtain the user's enthusiasm for each field within each time period. Combined with the number of interactions users have with each field within each time period, we can obtain the user's preference value for each field within each time period. Furthermore, we can introduce the frequency of user browsing each field within each time period to obtain the user's activity level in browsing each field within each time period, and thus obtain the user's basic interest in each field. By comparing the overlap in browsing areas and basic interests between users and their friends, the similarity of preferences between users and their friends is obtained. Combined with the interaction between users and their friends, the influence of each friend on the user is obtained, so as to filter target friends from all friends. Based on the influence of each target friend and the basic interest, the user's basic interest in each area is adjusted to obtain the updated user interest in each area, thereby updating the user profile for recommending personalized content.
[0005] In one embodiment, the process of obtaining the enthusiasm level is as follows: The maximum browsing time of a user across all fields within each browsing period in each interval is calculated, and the ratio of the user's browsing time across all fields within each browsing period in each interval to the maximum value is recorded as the browsing time ratio. The level of enthusiasm is positively correlated with the amount of content and the ratio of browsing time.
[0006] In one embodiment, the process of obtaining the preference value is as follows: Calculate the average number of user interactions across all browsing periods within each interval; The ratio of the number of times a user interacts with each field during each browsing period in each interval to the mean value is denoted as the interaction ratio. The preference values are positively correlated with the enthusiasm level and the ratio of interaction times, respectively.
[0007] In one embodiment, the process of obtaining the activity level is as follows: Calculate the arithmetic mean of the user's preference values for each field across all browsing periods within each interval; The activity level is positively correlated with the frequency and the arithmetic mean, respectively.
[0008] In one embodiment, the process of obtaining the basic interest level is as follows: For each domain, the intervals in which the normalized value of the activity level in the previous period is greater than the preset activity threshold are recorded as the user's activity intervals. The basic interest level is the weighted sum of the user's activity levels across all active periods, where the weight of the user's activity level in each active period is the inverse proportional mapping result of the time difference between each active period and the first period in the current period.
[0009] In one embodiment, the process of obtaining the preference similarity is as follows: Calculate the intersection of browsing areas between users and their friends; record the ratio of the number of elements in the intersection to the total number of all areas as the quantity ratio; Each domain in the intersection is denoted as a common domain. The difference between the normalized values of the basic interest degree of the user and each friend in each common domain is denoted as the interest difference degree of the user and each friend in each common domain. The average value of the interest difference degree of the user and each friend in all common domains is calculated. The preference similarity is positively correlated with the quantity ratio and negatively correlated with the average value.
[0010] In one embodiment, obtaining the influence of each friend on the user to filter target friends from all friends includes: By analyzing the interactions between users and their friends, we can determine the intensity of those interactions. The influence degree is the product of the preference similarity and the interaction strength; Friends whose normalized influence on a user is greater than or equal to a preset influence threshold are considered as the user's target friends.
[0011] In one embodiment, the process of obtaining the interaction intensity is as follows: The interaction intensity is the weighted sum of the number of interactions between the user and each friend across all intervals, where the weight of the number of interactions between the user and each friend in each interval is the inverse proportional mapping result of the time difference between each interval and the first interval in the current period.
[0012] In one embodiment, the step of revising the user's basic interest in each domain to obtain the updated user interest in each domain includes: The product of the influence of each target friend on the user and the basic interest of each target friend in each field is used as the interest adjustment degree of each target friend on the user in each field. Calculate the normalized result of the sum of the interest corrections of all target friends to the user in each field; Based on the user's basic interest in each field, the product of the normalized result and the user's basic interest in each field is superimposed, and the superimposed result is used as the updated user interest in each field.
[0013] Secondly, embodiments of this application also provide a personalized content recommendation system based on multi-dimensional user profiles, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described personalized content recommendation methods based on multi-dimensional user profiles.
[0014] This application has at least the following beneficial effects: This application quantifies the quantity and duration of user browsing content into enthusiasm, thereby capturing changes in user interests over different time periods and improving the dynamic adaptability of recommendations. Furthermore, it considers user interaction behavior, enabling preference values to more accurately reflect users' true interests and thus enhance the personalization of recommendations. In addition, it combines browsing frequency to assess user activity and obtain basic interest level, which can more comprehensively reflect the degree of user attention to different fields. At the same time, calculating basic interest level can provide a relatively stable interest assessment, avoiding the impact of short-term fluctuations on recommendation results. Furthermore, considering the overlap in browsing areas between users and their friends allows us to capture the impact of social relationships on user interests. Combining this with interaction data, we can assess the influence of friends on users, enabling us to more accurately identify friends with significant influence. This helps us leverage social relationships for recommendations, improving their relevance and credibility. By analyzing the influence of target friends on users and their basic interests in various areas, we can adjust users' basic interests in each area, resulting in updated user interests. This updated interest level more promptly and comprehensively reflects changes in user interests, enhancing personalized recommendations and ensuring that recommendations always match users' current interests. Attached Figure Description
[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 A flowchart illustrating the steps of a personalized content recommendation method based on a multi-dimensional user profile, provided in one embodiment of this application. Figure 2 This is a schematic diagram of the interest update process. Detailed Implementation
[0017] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.
[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".
[0019] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.
[0020] The following section, in conjunction with the accompanying drawings, details the specific scheme of the personalized content recommendation method and system based on multi-dimensional user profiles provided in this application.
[0021] Please see Figure 1 The diagram illustrates a flowchart of a personalized content recommendation method based on multi-dimensional user profiles according to an embodiment of this application. The method includes the following steps: Step 1: For a single user, preset the update cycle of the user's interest in each field, and divide each cycle into multiple intervals.
[0022] This application is mainly used for "recommendation of similar content", such as short videos or text and image information.
[0023] User interests and preferences are dynamic and influenced by various factors, such as emerging trends, changes in life stages, and the evolution of social relationships. Regularly updating user profiles can capture these changes in a timely manner, thereby ensuring that recommended content always matches the user's current interests.
[0024] For individual users, a preset update cycle for the user's interest in various fields is established, with each cycle divided into multiple intervals. Through seamless event tracking technology, the user's browsing and interaction information from the previous cycle is captured. Specifically: On mobile content pages, such as article detail pages and video playback pages, an SDK (Software Development Kit) is embedded to collect user ID, user browsing content ID, timestamp of when the user started browsing each piece of content, duration of the user's stay on each piece of content, and user interaction behavior while browsing each piece of content. Specifically, user interaction behavior while browsing each piece of content includes: liking, commenting, sharing, and saving. Based on user authorization, the system uses the social platform's API to access the user's social network data, obtain the user's friend list, and monitor and record every interaction between the user and each friend, including: likes, comments, and shares.
[0025] In this embodiment, the duration of the period and the interval are one week and one day, respectively. The duration of both the period and the interval are preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions.
[0026] Step 2: Update the interest level at the beginning of the current cycle based on the browsing and interaction information from the previous cycle.
[0027] Step 2.1: By measuring the number and duration of content viewed by users in each field during each browsing period within each interval, we obtain the user's enthusiasm for each field during each browsing period within each interval. Combined with the number of interactions between users and each field during each browsing period within each interval, we obtain the user's preference value for each field during each browsing period within each interval. Then, we introduce the frequency of user browsing each field within each interval to obtain the user's activity level in browsing each field within each interval, and thus obtain the user's basic interest in each field.
[0028] User interests are dynamic, especially as user preferences, needs, and behaviors change over time. For example, on a video recommendation platform, a user might enjoy watching game gameplay videos at one point, but over time, they might develop a greater interest in tech news. Failure to update user profiles in a timely manner may result in recommending outdated content, leading to a poor user experience.
[0029] Within the previous period, each consecutive online session of a user within each interval is recorded as a browsing session. Users may browse content from different fields at different times of the day. The viewing time can be used to measure whether the content is of interest to the user. Therefore, the field with the longest viewing time is usually the field that the user is most interested in. By comparing the user's viewing time in each field with the time spent in the field of greatest interest, the user's level of enthusiasm for each field can be reflected.
[0030] Based on the above analysis, the user's enthusiasm for each field is obtained by measuring the quantity and duration of content viewed by users in each field during each browsing period within each interval. The specific process is as follows: The maximum browsing time of a user across all fields within each browsing period in each interval is calculated, and the ratio of the user's browsing time across all fields within each browsing period in each interval to the maximum value is recorded as the browsing time ratio. The level of enthusiasm is positively correlated with the amount of content and the ratio of browsing time.
[0031] It should be noted that positive correlation means that the variables change in the same direction; when one variable increases, the other variable also increases, and when one variable decreases, the other variable also decreases.
[0032] In this embodiment, the expression for the user's enthusiasm for each field during each browsing time period within each interval is as follows: In the formula, This represents the user's enthusiasm for the i-th domain during the c-th browsing period within the d-th interval; ln() represents the logarithmic function with the natural constant as the base, used to analyze... Scaling is performed to avoid Too large an amount can cause a computer crash; This represents the number of times a user viewed content in the i-th domain within the c-th browsing time period of the d-th interval; This represents the duration of a user's browsing of the i-th field during the c-th browsing period within the d-th interval; This represents the maximum browsing time for a user across all categories within the c-th browsing period of the d-th interval. Wherein, This is recorded as the ratio of browsing time.
[0033] It should be added that: if a user does not browse the i-th field during the c-th browsing period, then... The value is assigned to 0.01.
[0034] It should be noted that if a user browses more content in the i-th field during the c-th browsing period, and for a longer duration, it indicates that the user's enthusiasm for the i-th field is higher during the c-th browsing period.
[0035] Furthermore, enthusiasm typically reflects a user's level of interest in a particular area, primarily determined by browsing time and the amount of content viewed. This is a short-term behavioral indicator, potentially influenced by environment, context, or specific content. To more deeply consider users' long-term attitudes and preferences towards each area, it's necessary to also consider their interactive behavior.
[0036] Based on the above analysis, by measuring the user's enthusiasm for each domain during each browsing period within each interval, and the number of interactions the user has with each domain during each browsing period within each interval, we obtain the user's preference value for each domain during each browsing period within each interval. The specific process is as follows: Calculate the average number of user interactions across all browsing periods within each interval; The ratio of the number of times a user interacts with each field during each browsing period in each interval to the mean value is recorded as the interaction ratio, and is used as the user's participation level in each field during each browsing period in each interval. The preference values are positively correlated with the enthusiasm and participation, respectively.
[0037] In this embodiment, the expression for the user's preference values for each field during each browsing time period within each interval is as follows: In the formula, This represents the user's preference value for the i-th domain during the c-th browsing period within the d-th interval; This represents the user's engagement with the i-th domain during the c-th browsing period within the d-th interval; This represents the user's enthusiasm for the i-th domain during the c-th browsing period within the d-th interval.
[0038] It should be added that: if a user interacts with the i-th domain 0 times during the c-th browsing period within the d-th interval, then... The value is assigned to 0.01.
[0039] Furthermore, based on obtaining users' preference values for each field during each browsing time period within each interval, the frequency of users browsing each field within each interval is introduced to obtain the user's activity level in browsing each field within each interval. The specific process is as follows: Calculate the arithmetic mean of the user's preference values for each field across all browsing periods within each interval; The activity level is positively correlated with the frequency and the arithmetic mean, respectively.
[0040] In this embodiment, the expression for the user's activity level in browsing various fields within each interval is: In the formula, This represents the user's activity level when browsing the i-th field within the d-th interval; This represents the number of times a user browses the i-th field within the d-th interval; This represents the total number of times a user browses all fields within the d-th interval; This represents the arithmetic mean of the user's preference values for the i-th domain across all browsing periods within the d-th interval. Where, This represents the frequency with which a user browses the i-th domain within the d-th interval.
[0041] It should be added that: if a user browses the i-th field 0 times within the d-th interval, then... The value is assigned to 0.01.
[0042] It should be noted that the higher the frequency of a user browsing the i-th domain within the d-th interval, and the larger the preference value, the more frequently the user browses the i-th domain and interacts with its content. This indicates that the user shows greater interest and attention to the i-th domain, and is more active in the i-th domain.
[0043] Furthermore, by comprehensively analyzing users' activity levels across all timeframes in the previous period, we can gain a more complete understanding of users' interest in each area, which helps to provide richer recommendations and reduce the monotony of recommending only one type of content.
[0044] Based on the above analysis, and by comprehensively considering the user's activity level across all intervals in the previous period, we can obtain the user's basic interest in each field. The specific process is as follows: For each domain, the normalized value of the activity level in the previous period that is greater than the preset activity threshold is statistically analyzed and recorded as the user's activity interval. The basic interest level is the weighted sum of the user's activity levels across all active periods, where the weight of the user's activity level in each active period is the inverse proportional mapping result of the time difference between each active period and the first period in the current period.
[0045] In this application, all normalization operations are implemented using the Min-Max normalization method, which is a well-known technique and will not be described further in this application.
[0046] In this embodiment, the preset activity threshold is set to 0.5, and the value of the preset activity threshold is calculated from experimental data.
[0047] In this embodiment, taking the i-th domain as an example, the expression for the user's basic interest in the i-th domain is: In the formula, This represents the user's basic level of interest in the i-th domain; This represents the total number of user activity zones. This represents the number of intervals between a user's z-th active interval and the first interval in the current period; This represents the user's activity level when browsing the i-th domain within the z-th active range.
[0048] It should be noted that: As a weighting factor, the closer the z-th active interval is to the first interval in the current period, the more vivid the user's memory and feelings will be, and the greater the weight will be given to the activity performance of the z-th active interval.
[0049] Step 2.2: By comparing the overlap of browsing areas and basic interest levels between the user and each friend, the similarity of preferences between the user and each friend is obtained. Combined with the interaction between the user and friends, the influence of each friend on the user is obtained, so as to filter each target friend from all friends. By comparing the influence of each target friend with the basic interest level, the user's basic interest level in each area is corrected to obtain the updated user's interest level in each area.
[0050] In modern society, a person's decisions and interests are often influenced by those around them, especially the opinions and behaviors of close friends. Interactions between friends can create a powerful social influence; when a user's friends show strong interest in a particular area, the user may also be influenced and more inclined to focus on that area. Therefore, it is necessary to further analyze the areas of interest within a user's social circle during the previous period, and then update the user profile in real time based on the latest social interactions to better reflect the user's current interest status.
[0051] Based on the calculation method for users' basic interest in each field, calculate the basic interest of each user's friends in each field.
[0052] Furthermore, by comparing the overlap in browsing areas and basic interests between the user and their friends, the similarity of preferences between the user and their friends is obtained. The specific process is as follows: The intersection of browsing areas between the user and each of their friends in the previous period is statistically analyzed; the ratio of the number of elements in the intersection to the total number of all areas is denoted as the quantity ratio. Each domain in the intersection is denoted as a common domain. The difference between the normalized values of the basic interest degree of the user and each friend in each common domain is denoted as the interest difference degree of the user and each friend in each common domain. The average value of the interest difference degree of the user and each friend in all common domains is calculated. The preference similarity is positively correlated with the quantity ratio and negatively correlated with the average value.
[0053] It should be noted that negative correlation means that the variables change in opposite directions; when one variable increases, the other decreases, and vice versa.
[0054] It should be noted that: difference refers to the degree of distinction between data, which can be achieved by calculating the absolute value of the difference, the square of the difference, the ratio, etc. This application does not impose any special restrictions on this.
[0055] In this embodiment, the difference between the normalized values of the basic interest degree is specifically the absolute value of the difference.
[0056] In this embodiment, the expression for the preference similarity between the user and each of their friends is: In the formula; This represents the similarity of preferences between the user and the f-th friend; This represents the number of elements in the intersection of the browsing domains of the user and the f-th friend; This represents the total number across all fields; This represents the average difference in interest between the user and their f-th friend across all shared areas. Wherein, It is denoted as the quantity ratio.
[0057] It should be noted that the more the current user's interests match those of the f-th friend, and the smaller the difference in their basic interest in all areas of the intersection, the closer their interest in the areas of the intersection is. In other words, the more common their interests are, and the higher the similarity of their preferences.
[0058] Furthermore, users are often more easily influenced by recommendations or content shared by close friends. Therefore, when making personalized content recommendations to users, priority should be given to friends who interact with users frequently, as their opinions and behaviors are more likely to directly affect users' choices.
[0059] Based on the above analysis, the interaction strength between the user and each friend is obtained by analyzing the user's interactions with each friend, specifically as follows: The interaction intensity is the weighted sum of the number of interactions between the user and each friend across all intervals, where the weight of the number of interactions between the user and each friend in each interval is the inverse proportional mapping result of the time difference between each interval and the first interval in the current period.
[0060] In this embodiment, the expression for the interaction strength between the user and each friend is: In the formula, This represents the interaction strength between the user and the f-th friend; D represents the total number of intervals within a single period. This represents the number of intervals between the d-th interval in the previous period and the first interval in the current period. This represents the number of interactions between the user and the f-th friend within the d-th interval.
[0061] Furthermore, since interaction strength measures the closeness of the relationship between a user and their friend, while preference similarity reflects the consistency of their interests, considering only one metric may lead to a misjudgment of a friend's influence. For example, a friend may share similar interests with a user, but if there is a lack of sufficient interaction between them, that friend's recommendations may still have a relatively small impact on the user's behavior. Therefore, a comprehensive assessment of each friend's influence on the user is necessary, combining interaction strength and preference similarity.
[0062] Based on the above analysis, and considering the similarity of preferences and the intensity of interaction between the user and each of their friends, the influence of each friend on the user is obtained, specifically: The product of the similarity of preferences between the user and each of their friends and the intensity of their interactions is used as the degree of influence of each friend on the user.
[0063] Furthermore, by comparing the influence of each friend on the user with a preset influence threshold, the user's target friends are selected, specifically as follows: Friends whose normalized influence on a user is greater than or equal to a preset influence threshold are considered as the user's target friends.
[0064] In this embodiment, the preset influence threshold is set to 0.5, and the value of the preset influence threshold is calculated from experimental data.
[0065] Furthermore, by analyzing the influence of each target friend on the user and the basic interest of each target friend in each field, the user's basic interest in each field is adjusted to obtain the updated user interest in each field. The specific process is as follows: The product of the influence of each target friend on the user and the basic interest of each target friend in each field is used as the interest adjustment degree of each target friend on the user in each field. Calculate the normalized result of the sum of the interest corrections of all target friends to the user in each field; Based on the user's basic interest in each field, the product of the normalized result and the user's basic interest in each field is superimposed, and the superimposed result is used as the user's interest in each field after the update at the initial moment of the current period.
[0066] In this embodiment, the expression for the user's interest in each domain after the update at the initial time of the current period is: In the formula, This indicates the user's interest in each area after the update at the initial moment of the current period; This represents the user's basic interest in the i-th domain; norm() represents the normalization operation; M represents the total number of the user's target friends; This represents the degree of interest adjustment of the m-th target friend towards the user in the i-th domain. A diagram illustrating the interest update process is shown below. Figure 2 As shown.
[0067] Step 3: Update the user profile for recommending personalized content.
[0068] Read the user's original interest in each field from the user profile database, and use the updated user interest in each field at the beginning of the current period to overwrite the original interest, thereby updating the user profile.
[0069] Based on the updated user profile, a domain preference distribution is constructed. If the score of a certain domain in the updated user profile increases significantly, the content of that domain will be given priority during the recall phase.
[0070] The system reads the user's interest in various fields after the update at the beginning of the current period, maps the numerical interest values to recall tags, such as "technology" and "entertainment," and retrieves a candidate set from the content library. During the recall phase, the user's interest in each field after the update at the beginning of the current period is used as a weighting coefficient. For example, if a user's interest in the "technology" field at the beginning of the current period is 0.85, then content in the "technology" field will be prioritized during the recall phase. The candidate set refers to a group of content that the user might be interested in, selected from the vast content library in the early stages of the recommendation process.
[0071] At the beginning of each cycle, update the user's interest in each area to ensure that recommendation decisions are based on the latest user interest status.
[0072] Based on the same inventive concept as the above methods, this application also provides a personalized content recommendation system based on multi-dimensional user profiles, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described personalized content recommendation methods based on multi-dimensional user profiles.
[0073] In summary, this application quantifies the quantity and duration of user browsing content as enthusiasm, thereby capturing changes in user interests over different time periods and improving the dynamic adaptability of recommendations. Furthermore, it considers user interaction behavior, enabling preference values to more accurately reflect users' true interests and thus enhancing the personalization of recommendations. Additionally, by combining browsing frequency with an assessment of user activity, a basic interest score is obtained, which more comprehensively reflects the degree of user attention to different areas. Moreover, calculating the basic interest score provides a relatively stable interest assessment, avoiding the impact of short-term fluctuations on recommendation results. Furthermore, considering the overlap in browsing areas between users and their friends allows us to capture the impact of social relationships on user interests. Combining this with interaction data, we can assess the influence of friends on users, enabling us to more accurately identify friends with significant influence. This helps us leverage social relationships for recommendations, improving their relevance and credibility. By analyzing the influence of target friends on users and their basic interests in various areas, we can adjust users' basic interests in each area, resulting in updated user interests. This updated interest level more promptly and comprehensively reflects changes in user interests, enhancing personalized recommendations and ensuring that recommendations always match users' current interests.
[0074] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0075] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.
Claims
1. A personalized content recommendation method based on multi-dimensional user profiles, characterized in that, The method includes the following steps: For a single user, the update cycle for the user's interest in various fields is preset, and each cycle is divided into multiple intervals. The interest level is updated at the beginning of the current period based on browsing and interaction information from the previous period. Specifically: By measuring the quantity and duration of content viewed by users in each field within each time period, we can obtain the user's enthusiasm for each field within each time period. Combined with the number of interactions users have with each field within each time period, we can obtain the user's preference value for each field within each time period. Furthermore, we can introduce the frequency of user browsing each field within each time period to obtain the user's activity level in browsing each field within each time period, and thus obtain the user's basic interest in each field. By comparing the overlap in browsing areas and basic interests between users and their friends, the similarity of preferences between users and their friends is obtained. Combined with the interaction between users and their friends, the influence of each friend on the user is obtained, so as to filter target friends from all friends. Based on the influence of each target friend and the basic interest, the user's basic interest in each area is adjusted to obtain the updated user interest in each area, thereby updating the user profile for recommending personalized content.
2. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the enthusiasm level is as follows: The maximum browsing time of a user across all fields within each browsing period in each interval is calculated, and the ratio of the user's browsing time across all fields within each browsing period in each interval to the maximum value is recorded as the browsing time ratio. The level of enthusiasm is positively correlated with the amount of content and the ratio of browsing time.
3. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process for obtaining the preference value is as follows: Calculate the average number of user interactions across all browsing periods within each interval; The ratio of the number of times a user interacts with each field during each browsing period in each interval to the mean value is denoted as the interaction ratio. The preference values are positively correlated with the enthusiasm level and the ratio of interaction times, respectively.
4. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the activity level is as follows: Calculate the arithmetic mean of the user's preference values for each field across all browsing periods within each interval; The activity level is positively correlated with the frequency and the arithmetic mean, respectively.
5. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the basic interest level is as follows: For each domain, the intervals in which the normalized value of the activity level in the previous period is greater than the preset activity threshold are recorded as the user's activity intervals. The basic interest level is the weighted sum of the user's activity levels across all active periods, where the weight of the user's activity level in each active period is the inverse proportional mapping result of the time difference between each active period and the first period in the current period.
6. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the preference similarity is as follows: Calculate the intersection of browsing areas between users and their friends; record the ratio of the number of elements in the intersection to the total number of all areas as the quantity ratio; Each domain in the intersection is denoted as a common domain. The difference between the normalized values of the basic interest degree of the user and each friend in each common domain is denoted as the interest difference degree of the user and each friend in each common domain. The average value of the interest difference degree of the user and each friend in all common domains is calculated. The preference similarity is positively correlated with the quantity ratio and negatively correlated with the average value.
7. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the influence of each friend on the user, in order to filter target friends from all friends, includes: By analyzing the interactions between users and their friends, we can determine the intensity of those interactions. The influence degree is the product of the preference similarity and the interaction strength; Friends whose normalized influence on a user is greater than or equal to a preset influence threshold are considered as the user's target friends.
8. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of obtaining the interaction intensity is as follows: The interaction intensity is the weighted sum of the number of interactions between the user and each friend across all intervals, where the weight of the number of interactions between the user and each friend in each interval is the inverse proportional mapping result of the time difference between each interval and the first interval in the current period.
9. The personalized content recommendation method based on multi-dimensional user profiles as described in claim 1, characterized in that, The process of revising the user's basic interest in each field to obtain the updated user interest in each field includes: The product of the influence of each target friend on the user and the basic interest of each target friend in each field is used as the interest adjustment degree of each target friend on the user in each field. Calculate the normalized result of the sum of the interest corrections of all target friends to the user in each field; Based on the user's basic interest in each field, the product of the normalized result and the user's basic interest in each field is superimposed, and the superimposed result is used as the updated user interest in each field.
10. A personalized content recommendation system based on multi-dimensional user profiles, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the personalized content recommendation method based on multi-dimensional user profiles as described in any one of claims 1-9.