A method for pushing activities on a big data-based cultural cloud platform

Abstract

This invention discloses an activity recommendation method for a big data-based cultural cloud platform, belonging to the field of activity recommendation technology. The method includes the following steps: constructing a multi-scale interest graph based on user behavior to obtain a user graph model; calculating the similarity of activity nodes based on the user graph model to obtain a candidate activity set; performing behavioral preference inference based on the candidate activity set to obtain a refined activity list; recommending activities based on the refined activity list to obtain recommendation results; and collecting user feedback behavior based on the recommendation results to obtain a feedback interest graph. This invention can more accurately reflect users' multi-dimensional interests, capture changes in user interests over different time periods, handle complex user interest structures and behavioral patterns, improve the accuracy and flexibility of recommendations, enhance the adaptability to dynamic changes in user interests, improve the timeliness and personalized recommendation capabilities of the recommendation system, and increase the accuracy of recommendations.

Description

Technical Field

[0001] This invention relates to the field of event push technology, specifically to an event push method based on a big data-driven cultural cloud platform. Background Technology

[0002] With the deepening development of cultural digitalization, cultural cloud platforms have gradually become an important carrier connecting the public with public cultural service resources. The activity push notifications from cultural cloud platforms provide users with personalized and diversified cultural service offerings.

[0003] Existing technologies have shortcomings in user interest modeling: most existing methods use static or single-time-scale interest graph models, which cannot dynamically adjust the changing trends of interests. Furthermore, the modeling of deep relationships among user interests is relatively simple, which can easily lead to problems such as low recommendation accuracy and poor timeliness.

[0004] Existing technologies for behavioral preference reasoning have shortcomings: existing behavioral preference reasoning methods usually rely on simple historical behavior statistics or recommendations based on a single interest tag, which cannot handle the diversity and changes in user interests well, and are difficult to cope with the dynamic reasoning needs of users with cross-category interest changes or cold-start users. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an activity push method based on a big data-driven cultural cloud platform to solve the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide an activity push method based on a big data-driven cultural cloud platform, comprising the following steps:

[0008] S1. Construct a multi-scale interest graph based on user behavior to obtain a user graph model;

[0009] S2. Calculate the similarity of activity nodes based on the user graph model to obtain a candidate activity set;

[0010] S3. Perform behavioral preference reasoning based on the candidate activity set to obtain a refined activity list;

[0011] S4. Push activities based on the refined activity list and obtain the push results;

[0012] S5. Collect user feedback behavior based on the push results to obtain a feedback interest graph.

[0013] To further optimize this technical solution, the multi-scale interest map construction in step S1 includes:

[0014] Based on the historical operation behavior of each user in the cultural cloud platform, each behavior is assigned a unique identifier and sorted by timestamp to extract the user's behavior patterns in different time periods, thereby obtaining the multi-dimensional time behavior features of each user, and then using the user graph construction model to construct a multi-scale interest graph.

[0015] To further optimize this technical solution, the user map construction model includes:

[0016] ;

[0017] in:

[0018] : No. Activities at a time scale interest level;

[0019] Time scale The weights range from 0 to 1, and the sum of all weights is 1;

[0020] :user For the activity Overall interest level;

[0021] : Number of time scales.

[0022] To further optimize this technical solution, the degree of interest includes:

[0023] ;

[0024] in:

[0025] :user In time scale The set of behaviors on;

[0026] :user In behavior The weights on the scale range from 0 to 1, and the sum of all weights is 1.

[0027] :Behavior In time scale The above activities The influence coefficient ranges from 0 to 1.

[0028] To further optimize this technical solution, the influence coefficient includes:

[0029] ;

[0030] in:

[0031] Time decay parameter;

[0032] Time scale Length;

[0033] :Behavior With activities The strength of the association;

[0034] On a time scale Above, behavior In the activity The number of times it is triggered;

[0035] :Behavior The total number of triggers across all activities is greater than or equal to 1.

[0036] To further optimize this technical solution, the correlation strength includes:

[0037] ;

[0038] in:

[0039] Among all users, behavior Appeared at the event Total number of times;

[0040] :Behavior Total number of times it appears across all activity types.

[0041] To further optimize this technical solution, the behavioral preference reasoning in S3 includes:

[0042] Based on the user's interest graph and the resulting candidate activity set, a preference inference model is used to infer the user's interests and preferences, thereby filtering the candidate activity set and pushing a more accurate list of activities to the user.

[0043] To further optimize this technical solution, the preference reasoning model includes:

[0044] ;

[0045] in:

[0046] : The merged users For the activity The total preference value;

[0047] :user Short-term behavior on activities Preferences;

[0048] :user Long-term behavior on activities Preferences;

[0049] : Behavioral fusion coefficient, ranging from 0 to 1.

[0050] To further optimize this technical solution, the user behavior preferences for activities include:

[0051] ;

[0052] in:

[0053] Users on a time scale The above activities The preference is determined by the size of the time scale, indicating whether it is a short-term or long-term preference.

[0054] On a time scale Attention weights, ranging from 0 to 1, are applied on this timescale, and the sum of all weights is 1.

[0055] To further optimize this technical solution, the attention weights include:

[0056] ;

[0057] in:

[0058] :user In time scale Above the behavior Preference level;

[0059] Apply an exponential function to the preference for each behavior;

[0060] : Normalization factor.

[0061] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program instructions, when executed by the processor, implement the steps of an activity push method for a big data-based cultural cloud platform as described in the first aspect of the present invention.

[0062] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of an activity push method for a big data-based cultural cloud platform as described in the first aspect of the present invention.

[0063] Compared with existing technologies, this invention provides an activity push method based on a big data-driven cultural cloud platform, which has the following beneficial effects:

[0064] This activity recommendation method based on big data in the cultural cloud platform, through user graph modeling and the introduction of interest modeling at multiple time scales, can capture changes in user interests over different time periods, improving the accuracy and flexibility of recommendations, enhancing the adaptability to dynamic changes in user interests, and improving the timeliness and personalized recommendation capabilities of the recommendation system.

[0065] By introducing an attention mechanism through a preference inference model, the influence of short-term and long-term user behavior is adjusted by dynamically adjusting attention weights, thereby more accurately reflecting the user's multidimensional interests, handling the user's complex interest structure and behavioral patterns, and improving the accuracy and personalization of recommendations. Attached Figure Description

[0066] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 This is a flowchart illustrating an activity push method for a big data-based cultural cloud platform proposed in this invention.

[0068] Figure 2 This is a flowchart illustrating the user graph construction model of the activity push method for a big data-based cultural cloud platform proposed in this invention.

[0069] Figure 3 This is a flowchart illustrating the preference reasoning model of an activity push method based on a big data cultural cloud platform proposed in this invention.

[0070] Figure 4 This is a flowchart illustrating the user feedback model of an activity push method based on a big data cultural cloud platform proposed in this invention. Detailed Implementation

[0071] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0072] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0073] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0074] Example 1:

[0075] Reference Figures 1-4 This is the first embodiment of the present invention, which provides an activity push method based on a big data-driven cultural cloud platform, including the following steps:

[0076] S1. Construct a multi-scale interest graph based on user behavior to obtain a user graph model.

[0077] In this embodiment, the construction of the multi-scale interest map includes:

[0078] Based on the historical operational behavior of each user in the cultural cloud platform, including activity clicks, details browsing, reservations, and actual participation, each behavior is assigned a unique identifier and sorted by timestamp. Long Short-Term Memory Network (existing technology) is used for processing to extract the user's behavioral patterns in different time periods. On this basis, the time dimension is segmented (such as by day, week, or holiday) to obtain the multi-dimensional time behavior characteristics of each user. Then, a user graph construction model is used to construct a multi-scale interest graph.

[0079] Furthermore, the user graph construction model includes:

[0080] User Interest Graph Composed of interest node graphs at multiple time scales , This is a set of nodes in the graph, representing the user's points of interest and activity. The edge set of the graph represents the interactive behaviors between users and activities, including clicks, favorites, participation, etc.

[0081] This model obtains a comprehensive interest score by weighting and fusing users’ interest in activities at different time scales, and then judges users’ activity interest points based on the interest score.

[0082] ;

[0083] in:

[0084] : No. Activities at a time scale interest level;

[0085] Time scale The weight represents the importance of interest information at that time scale, ranging from 0 to 1. The sum of all weights is 1. The larger the weight, the greater the impact of the activity interest at that time scale on the overall interest of the activity.

[0086] :user For the activity Overall interest level;

[0087] : Number of time scales.

[0088] Furthermore, the degree of interest includes:

[0089] ;

[0090] in:

[0091] :user In time scale The set of behaviors represents all of a user's behavioral records, including clicks, browsing, participation, etc.

[0092] :user In behavior The weights on the scale range from 0 to 1, and the sum of all weights is 1, adjusted based on the user's historical behavior.

[0093] :Behavior In time scale The above activities The influence coefficient, ranging from 0 to 1, reflects the user's impact on a time scale. Whether the response to such activities is focused on specific behaviors.

[0094] Furthermore, the influence coefficient includes:

[0095] ;

[0096] in:

[0097] Time decay parameter, adjusted during model training;

[0098] Time scale Length;

[0099] :Behavior With activities The strength of the association;

[0100] On a time scale Above, behavior In the activity The number of times it is triggered;

[0101] :Behavior The total number of triggers across all activities is greater than or equal to 1.

[0102] Furthermore, the correlation strength includes:

[0103] Association strength measures the general relationship between behavior types and activity types. It is a long-term statistical value that provides the basic weights for other parameters and can be adjusted according to regional or platform characteristics. For a fixed behavior i, the association strength between it and all activities is... The sum of is 1;

[0104] ;

[0105] in:

[0106] The total number of times behavior i appears in activity a among all users;

[0107] : The total number of times behavior i appears across all activity types.

[0108] Most traditional methods employ static or single-time-scale interest graph models, largely neglecting the multidimensionality of time, including differences between daily and holiday behaviors, and the influence of seasonal factors and specific events. They cannot dynamically adjust the changing trends of interests, and their modeling of deep relationships among user interests is relatively simple, easily leading to problems such as low recommendation accuracy and poor timeliness. This model, by introducing interest modeling at multiple time scales, can capture changes in user interests over different time periods, improving the accuracy and flexibility of recommendations. It can better adapt to the dynamic changes in user interests, enhancing the timeliness and personalized recommendation capabilities of the recommendation system.

[0109] The use of the above model includes:

[0110] Parameter calculation: Based on user behavior records, calculate the correlation strength between behavior and activity. Influence coefficient of behavior on activity ;

[0111] Interest Calculation: Calculates user interest in the activity at different time scales based on parameters. By weighting and aggregating user interest in the activity across different time scales, a comprehensive user interest score for the activity is obtained. ;

[0112] Interest Graph Construction: Based on users' overall interest in the activities and their interactive behaviors with the activities, a user interest graph is constructed.

[0113] S2. Calculate the similarity of activity nodes based on the user graph model to obtain a candidate activity set.

[0114] In this embodiment, similarity calculation includes:

[0115] Based on the user graph model obtained in step S1, the similarity between user activity interest points and nodes of activities to be recommended is calculated. Each activity to be recommended is encoded as a fixed-dimensional feature vector, including attributes such as the activity's time period, spatial coordinates, topic tags, and affiliated organization. Using cosine similarity calculation (existing technology) as the basic metric, the matching degree between the user activity interest point vector and the activity node vector is calculated. Activities with a similarity score higher than a set threshold are included in the candidate activity set. The candidate activity set is a personalized set of activities generated for each user, reflecting the space of potentially interesting activities for the user within the current time period.

[0116] S3. Perform behavioral preference reasoning based on the candidate activity set to obtain a refined activity list.

[0117] In this embodiment, behavioral preference reasoning includes:

[0118] Based on the user's interest graph and the resulting candidate activity set, a preference inference model is used to infer the user's interests and preferences. This allows for the filtering of the candidate activity set and the delivery of a more accurate list of activities to the user. This list not only reflects the user's structured preferences but also incorporates short-term behavioral changes, resulting in more targeted recommendations.

[0119] After obtaining a user's current interests and preferences, it's also possible to determine whether the user exhibits a tendency for cross-tag linkage behavior towards a certain type of activity. For example, if a user recently participated in both "Guqin Experience" and "Chinese Traditional Clothing Showcase" activities, it can be identified that the user has a stable preference for the higher-level theme cluster of "Traditional Culture Experience." This allows for the prioritization of content belonging to this cluster, such as "Calligraphy Interactive Experience" or "Classical Etiquette Open Class," even if these specific activities do not have the highest match with historical behavior tags. This enhances the user's ability to express cross-theme interests in cultural activities.

[0120] Furthermore, the preference inference model includes:

[0121] This model introduces a fusion mechanism of short-term and long-term behavior, and uses the complementarity of users' immediate behavior and historical long-term behavior to perform refined reasoning, making recommendations more personalized and real-time.

[0122] ;

[0123] in:

[0124] : The merged users For the activity The total preference value;

[0125] :user Short-term behavior on activities Preferences;

[0126] :user Long-term behavior on activities Preferences;

[0127] The behavior fusion coefficient, ranging from 0 to 1, represents the relative influence of short-term and long-term behaviors. It dynamically adjusts the weight of short-term behaviors based on the frequency of user activities. The more frequently a user participates in short-term activities, the higher the weight of short-term behaviors, making the recommended activities more in line with the user's current preferences.

[0128] Furthermore, the user behavior preferences for activities include:

[0129] The purpose of this step is to calculate the impact of short-term and long-term user behavior on the activity. Short-term behavior typically refers to a user's behavior over a recent period (such as activities in the past month), while long-term behavior refers to a user's behavioral patterns over a longer time span (such as activities in the past year).

[0130] ;

[0131] in:

[0132] Users on a time scale The above activities Preferences;

[0133] On a time scale Attention weights reflect behavior The degree of influence on activity preferences.

[0134] Furthermore, the attention weights include:

[0135] The impact of user behavior on preference inference is determined by calculating the importance of user behavior, with more frequent or more important behaviors assigned higher weights.

[0136] ;

[0137] in:

[0138] :user In time scale Above the behavior The user's preference level is calculated by information such as the frequency of user behavior and participation, or estimated by historical behavior data, such as the frequency of user clicking on activities and the frequency of participating in activities.

[0139] An exponential function is applied to the preference for each behavior to enhance the weighting of high-frequency behaviors;

[0140] : Normalization factor, ensuring that the sum of all weights is 1.

[0141] Traditional behavioral preference inference methods often rely on simple historical behavior statistics or recommendations based on a single interest tag, which may not be able to handle the diversity and changes in user interests well. For example, simple behavior matching models cannot handle the diversity and changes in user interests well. This model introduces an attention mechanism, which dynamically adjusts the attention weights to adjust the influence of short-term and long-term user behavior, thereby more accurately reflecting the multidimensional interests of users, handling the complex interest structure and behavioral patterns of users, and improving the accuracy and personalization of recommendations.

[0142] The steps for using this model include:

[0143] Weight Calculation: Based on the attention mechanism, the importance of user behavior is calculated to determine its impact on preference inference, thus obtaining attention weights. More frequent or more important behaviors are given higher weight;

[0144] Preference inference: based on the calculated attention weights Perform preference inference to obtain users' short-term behavioral preferences. and long-term behavioral preferences The fused values ​​are then used to obtain the overall preference value. ;

[0145] Activity filtering: Based on the obtained fused overall preference value Based on the candidate activity set obtained in step S2, user interests in the candidate activity set are explored, and a refined list of recommended activities is selected.

[0146] S4. Push activities based on the refined activity list and obtain the push results.

[0147] In this embodiment, the activity push includes:

[0148] Based on the refined activity list, a task scheduling system (such as Apache Airflow) is used to model the user-activity pair scheduling. Each task instruction includes the push user ID, the corresponding activity ID, the push timestamp, the push channel (including APP, official account, mini program) and the push content identifier. After system configuration, a push execution plan is generated and the instruction is triggered periodically to ensure that each user only receives activity notifications that match their preferences.

[0149] S5. Collect user feedback behavior based on the push results to obtain a feedback interest graph.

[0150] In this embodiment, feedback behavior collection includes:

[0151] After pushing activities to users, user feedback behavior is collected, and a user feedback model is used to construct a feedback interest graph. This enhances the feedback correction function of the user interest graph to optimize future recommendation results.

[0152] Furthermore, the user feedback model includes:

[0153] By weighting timeliness and calculating feedback intensity, we can dynamically construct and update users' feedback interest graphs, capture the true intent of users' feedback, and achieve personalized and real-time responsive activity recommendations.

[0154] Feedback intensity calculation:

[0155] ;

[0156] in:

[0157] :user In time scale The intensity of feedback to the activity is quantified based on user behavior types, including clicks, favorites, participation, etc.

[0158] : No. Weights of class feedback;

[0159] : No. The intensity function of feedback is set according to the type of feedback. For example, the intensity of feedback for clicking is low, and the intensity of feedback for participating in an activity is high.

[0160] :user In time scale Feedback on the activity, including clicks, favorites, and shares;

[0161] Feedback timeliness coefficient, used to control the greater impact of more recent feedback on the graph;

[0162] : Number of feedback types;

[0163] Feedback graph edge weight update:

[0164] ;

[0165] in:

[0166] :user Weighting of feedback relationships between activities;

[0167] Activity type weighting reflects the different impacts of different types of activities (such as art exhibitions, cultural lectures, etc.) on user appeal and feedback intensity;

[0168] Feedback graph overall update rules:

[0169] Update the weights of the corresponding nodes and edges in the graph based on the calculated weights;

[0170] ;

[0171] in:

[0172] , : The updated feedback interest graph and the interest graph before the update.

[0173] Traditional feedback behavior analysis methods are usually based on a single feedback, which cannot comprehensively describe the intensity of users' interest in activities and cannot adjust recommendations based on users' latest behavior, resulting in lag and inflexibility. In contrast, this model comprehensively describes the intensity of users' interest in activities by integrating multiple feedback intensities (clicks, favorites, participation, etc.), thereby gaining a more detailed and accurate understanding of user behavior. It can adjust recommendations in real time based on users' latest behavior, improving the flexibility and real-time nature of feedback.

[0174] The steps for using the model include:

[0175] Feedback intensity calculation: When a user interacts with an activity (such as clicking, adding to favorites, or participating), the intensity function of each feedback is used. and feedback timeliness Calculate the strength of each feedback. ;

[0176] Map Update: Based on Feedback Strength and activity type weight Update the feedback weights between users and activities. The feedback interest graph is obtained by updating the graph based on the feedback weights. ;

[0177] Recommendation Adjustment: Based on the updated interest graph, we will provide users with more personalized and real-time activity recommendations.

[0178] Example 2:

[0179] This embodiment also provides a computer device applicable to an activity push method for a big data-based cultural cloud platform, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the activity push method for a big data-based cultural cloud platform as proposed in the above embodiment.

[0180] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements an activity push method for a big data-based cultural cloud platform as proposed in the above embodiments.

[0181] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0182] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0183] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0184] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0185] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0186] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An activity pushing method of a big data-based cultural cloud platform, characterized in that, The method comprises the following steps: S1, constructing a multi-scale interest graph based on user behavior to obtain a user graph model; The user graph construction model comprises: ； Wherein: : first : interest of activity : in a first time scale : time scale weights, ranging from 0 to 1, the sum of all weights being 1; : user to activity : overall interest : number of time scales; The interest degree comprises: ； Wherein: : user a collection of behaviors on a time scale ; : user in the behavior weight, ranging from 0 to 1, with the sum of all weights being 1; :Behavior In time scale The above activities The influence coefficient ranges from 0 to 1; The influence coefficient comprises: ； Wherein: : time decay parameter; : time scale : length of time : behavior with activity strength of association : on a time scale : the number of times a behavior : on an activity : is triggered; : behavior Total number of triggers in all activities, greater than or equal to 1; The correlation strength comprises: ； Wherein: : total number of times the behavior appeared on the activity ; : behavior Total number of occurrences across all activity types; S2, calculating the similarity of activity nodes according to the user graph model to obtain a candidate activity set; S3, reasoning the behavior preference according to the candidate activity set to obtain a refined activity list; S4, pushing activities according to the refined activity list to obtain a pushing result; S5, collecting feedback behavior according to the pushing result to obtain a feedback interest graph. 2.The activity pushing method of the cultural cloud platform based on big data according to claim 1, wherein, The multi-scale interest graph construction in step S1 comprises: Based on the historical operation behavior of each user in the cultural cloud platform, a unique identifier is assigned to each behavior and time-stamped to extract the behavior pattern of the user at different time periods, obtaining the multi-dimensional time behavior characteristics of each user, and then using the user graph construction model to construct a multi-scale interest graph. 3.The activity pushing method of the cultural cloud platform based on big data according to claim 1, wherein, The behavior preference reasoning in S3 comprises: According to the user's interest graph, combined with the obtained candidate activity set, the user's interest preference is reasoned using a preference reasoning model, so as to screen the candidate activity set and push a more accurate activity list to the user. 4.The activity pushing method of the cultural cloud platform based on big data according to claim 3, characterized in that, The preference reasoning model comprises: ； Wherein: : the user after fusion total preference value for the activity ​ : user short-term behavior to activity preferences; : user long-term behavior to the activity preferences; : behavioral fusion coefficient, ranging from 0 to 1.

5. The activity pushing method of the big data-based cultural cloud platform according to claim 4, characterized in that, The user's preference for activities comprises: ； Wherein: : preferences of the user on a time scale for activities ; : attention weights on a time scale ranging from 0 to 1, the sum of all weights being 1 on this time scale.

6. The activity pushing method of the cultural cloud platform based on big data according to claim 5, characterized in that, The attention weight comprises: ； Wherein: : user on a time scale of behavior preference; : applying an exponential function to the preference degree of each behavior; : Normalization factor.

Images