Digital media transmission method and system based on user behaviors
By collecting and analyzing user situation-aware data on smart terminal nodes, using long-term memory networks and federated learning technology to dynamically optimize user behavior patterns, and combining decision tree algorithms to predict user future behavior, the problems of insufficient dynamic optimization mechanism and insufficient privacy protection in the existing technology are solved, and personalized recommendations and effective protection of user privacy are achieved.
Patent Information
- Application Number
- CN202510190935.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-20
- Publication Date
- 2025-05-30
AI Technical Summary
In the prior art, there is insufficient dynamic optimization mechanism and insufficient privacy protection, making it difficult to adapt to real-time changing needs and effectively protect user privacy.
By collecting user situation perception data, using long-term and short-term memory network algorithm to analyze user behavior patterns, and forming dynamic optimization of user behavior patterns through federated learning, combining decision tree algorithm to predict user future behavior, generate preloading strategies, and protect user privacy through homomorphic encryption technology.
Dynamic optimization of user behavior patterns is achieved, ensuring that recommendation strategies are always close to users' immediate needs, and user privacy is protected through federated learning and homomorphic encryption technology.
Smart Images
Figure CN120067448A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of digital media transmission and personalized recommendation, and particularly to a method and system for digital media transmission based on user behavior. Background Art
[0002] With the development of information technology, digital media transmission and personalized recommendation have become important components of modern Internet services. In recent years, the progress of Internet of Things (IoT) devices and sensor technologies has made the collection and analysis of context-aware data a research hotspot. By integrating multiple sensors (such as GPS, accelerometers, etc.), rich user behavior data can be obtained, combined with external information such as calendar applications, social media interactions, and weather forecasts, enhancing the level of intelligence.
[0003] Traditional methods rely on static CDNs and rule engines, making it difficult to adapt to real-time changing requirements and performing poorly in dealing with complex user behavior patterns. In recent years, the application of technologies such as long short-term memory networks (LSTMs) and federated learning has significantly improved the accuracy of behavior pattern recognition and the level of privacy protection. Summary of the Invention
[0004] In view of the above existing problems, the present invention is proposed.
[0005] Therefore, the present invention provides a method for digital media transmission based on user behavior, which solves the problems of insufficient dynamic optimization mechanism and insufficient privacy protection in the prior art.
[0006] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for digital media transmission based on user behavior, which includes: Collecting user context-aware data; Analyzing the user context-aware data using the long short-term memory network algorithm to obtain the user's daily behavior pattern; Performing federated learning using the user context-aware data and the user's daily behavior pattern to form a dynamically optimized user behavior pattern, and predicting the user's future behavior through a decision tree algorithm; Generating a preloading strategy based on the user's future behavior; Loading digital media according to the user's current context-aware data and the preloading strategy, and recording the user's feedback; Adjusting the preloading strategy according to the user's feedback.
[0007] As a preferred solution of the method for digital media transmission based on user behavior of the present invention, wherein: the specific steps of collecting the user context-aware data are as follows: Collecting user context-aware data through IoT devices and sensors; The user context-aware data refers to the user's geographical location, timestamp, activity type, temperature, humidity, and light intensity.
[0008] As a preferred solution of the method for digital media transmission based on user behavior according to the present invention, wherein: analyzing the user context-aware data by using the long short-term memory network algorithm to obtain the user's daily behavior pattern, the specific steps are as follows. Perform normalization and encoding processing on the user context-aware data, and input the normalized and encoded user context-aware data into a multi-dimensional space to obtain a numerical vector of the user context-aware data features. Use the long short-term memory network algorithm to analyze the numerical vector of the user context-aware data features to form the user's daily behavior pattern, and the expression is: ; Wherein, is the user's daily behavior pattern, is the geographical location coordinate at time , is the activity type encoding vector at time , is the temperature value at time , is the humidity value at time , is the light intensity value at time , is the long short-term memory network algorithm.
[0009] As a preferred solution of the method for digital media transmission based on user behavior according to the present invention, wherein: using the user context-aware data and the user's daily behavior pattern for federated learning to form a dynamically optimized user behavior pattern, and predicting the user's future behavior through a decision tree algorithm, the specific steps are as follows. Initialize the federated learning client on the intelligent terminal node, and the intelligent terminal node uses the user context-aware data and the user's daily behavior pattern to train the parameters of the long short-term memory network algorithm through the federated learning algorithm. Calculate the parameter gradient of the long short-term memory network algorithm through the parameter gradient formula, and the expression is: ; Wherein, is the parameter gradient of the long short-term memory network algorithm, is the user's daily behavior pattern of the th intelligent terminal node, is the geographical location coordinate at time of the th intelligent terminal node, is the The moment of an intelligent terminal node The activity type coding vector of is the moment of the th intelligent terminal node, is the moment of the th intelligent terminal node, is the moment of the th intelligent terminal node, is the partial derivative of the loss function of the th intelligent terminal node with respect to the parameters of the long short-term memory network algorithm; Use homomorphic encryption technology to encrypt the parameter gradients of the long short-term memory network algorithm, and upload the encrypted parameter gradients of the long short-term memory network algorithm to the central server through a secure channel; After the central server receives the encrypted parameter gradients of the long short-term memory network algorithm from all participating nodes, it decrypts and aggregates the parameter gradients of the long short-term memory network algorithm, and at the same time distributes them to the intelligent terminal nodes to update the parameters of the long short-term memory network algorithm; Through the updated long short-term memory network algorithm, a dynamic optimized user behavior pattern is formed, and the expression is: ; Among them, is the dynamic optimized user behavior pattern, is the geographical location coordinate at the moment , is the activity type coding vector at the moment , is the temperature value at the moment , is the humidity value at the moment , is the light intensity value at the moment , is the updated long short-term memory network algorithm; Normalize the dynamic optimized user behavior pattern, select the C4.5 algorithm as the decision tree algorithm, and use the normalized dynamic optimized user behavior pattern to calculate the probability of the user's future behavior through the decision tree algorithm. The expression is: ; Among them, is the probability of the user's future behavior, is the weight of the th decision tree, is the prediction output of the th decision tree for the dynamic optimized user behavior pattern; Select the behavior with the highest probability based on the probability of the user's future behavior as the user's future behavior.
[0010] As a preferred solution of the method for digital media transmission based on user behavior according to the present invention, wherein: generating a preloading strategy based on the user's future behavior, the specific steps are as follows. Use the intelligent terminal node to collect the user's historical interaction data, and use the collaborative filtering algorithm to screen out candidate digital media based on the user's historical interaction data; Use the matrix factorization algorithm to convert the candidate digital media and the user's future behavior into the encoded vector of the candidate digital media and the encoded vector of the user's future behavior; Calculate the correlation score between the candidate digital media and the user's future behavior through the correlation score formula, and the expression is: ; Wherein, is the correlation score between the candidate digital media and the user's future behavior, is the weight of the th activity type, is the th candidate digital media's encoded vector on the th activity type, is the encoded vector of the user's future behavior on the th activity type, is the number of activity types, is the smoothing factor, is the user's future behavior; According to the correlation score between the candidate digital media and the user's future behavior, use the sorting algorithm to sort all candidate digital media to obtain the preloading strategy.
[0011] As a preferred solution of the method for digital media transmission based on user behavior according to the present invention, wherein: loading digital media according to the user's current context awareness data and the preloading strategy, and recording the user's feedback, the specific steps are as follows. Form a preloading digital media list according to the preloading strategy, and use the intelligent terminal node to collect the current user context awareness data; Calculate the current user behavior's probability distribution through the logistic regression algorithm for the current user context awareness data; Use the matrix factorization algorithm to convert the preloading digital media and the current user behavior's probability distribution into the encoded vector of the preloading digital media and the encoded vector of the current user behavior's probability distribution; Calculate the matching score between the preloading digital media and the current user behavior's probability distribution through the matching score formula, and the expression is: ; Among them, is the matching degree score of the pre-loaded digital media and the probability distribution of the current user behavior, is the weight of the th activity type, is the th encoding vector of the th activity type of the th pre-loaded digital media, is the encoding vector of the probability distribution of the current user behavior on the th activity type, is the probability of the current context awareness data on the th activity type, is the smoothing factor, is the probability distribution of the current user behavior, is the index of the pre-loaded digital media;
[0012] As a preferred solution of the method for transmitting digital media based on user behavior according to the present invention, wherein: the preloading strategy is adjusted according to user feedback, and the specific steps are as follows, Normalize the user feedback, and obtain the user feedback score by weighted summation; Based on the correlation score between the candidate digital media and the user's future behavior and the user feedback score, update the correlation score formula to form a user feedback enhanced correlation score formula, and the expression is: ; Among them, is the correlation score between the user feedback enhanced candidate digital media and the user's future behavior, is the adjustment parameter of the user feedback score, is the user feedback score; Adjust the preloading strategy according to the correlation score between the user feedback enhanced candidate digital media and the user's future behavior.
[0013] In a second aspect, the present invention provides a digital media transmission system based on user behavior, including, a collection module for collecting user context awareness data; a behavior module for analyzing user context awareness data by using the long short-term memory network algorithm to obtain the user's daily behavior pattern; a prediction module for performing federated learning using user context awareness data and the user's daily behavior pattern to form a dynamically optimized user behavior pattern, and predicting the user's future behavior through a decision tree algorithm; A preloading module that generates a preloading strategy based on the user's future behavior; A recording module that loads digital media according to the user's current context awareness data and the preloading strategy, and records the user's feedback; An adjustment module that adjusts the preloading strategy according to the user's feedback.
[0014] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the method for digital media transmission based on user behavior as described in the first aspect of the present invention is implemented.
[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the method for digital media transmission based on user behavior as described in the first aspect of the present invention is implemented.
[0016] The beneficial effects of the present invention are as follows: By initializing the federated learning client on the intelligent terminal node and using the user context awareness data and the user's daily behavior pattern to train the parameters of the long short-term memory network algorithm, the dynamic optimization of the user behavior pattern is realized, ensuring that the recommendation strategy always closely matches the user's immediate needs. By uploading the encrypted parameter gradients to the central server through the homomorphic encryption technology, the central server decrypts and aggregates these gradients, updates the global parameters, and then distributes them to the intelligent terminal node, realizing the continuous optimization of the global parameters while protecting the user's privacy. Description of the Drawings
[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0018] Figure 1 It is a flowchart of the method for digital media transmission based on user behavior in Embodiment 1.
[0019] Figure 2 It is a schematic diagram of the system for digital media transmission based on user behavior in Embodiment 1. Detailed Embodiments
[0020] To make the above objects, features, and advantages of the present invention more obvious and understandable, the detailed embodiments of the present invention will be described in detail below with reference to the drawings in the specification.
[0021] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.
[0022] Secondly, as used herein, an "embodiment" or "embodiments" refer to specific features, structures, or characteristics that may be included in at least one implementation of the present invention. The phrase "in an embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it an individual or alternative embodiment that is mutually exclusive with other embodiments.
[0023] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a method for digital media transmission based on user behavior, including the following steps: S1: Collect user context awareness data.
[0024] The specific steps are as follows. The intelligent terminal node initializes its built-in or external sensors and API interfaces.
[0025] The intelligent terminal node refers to a smart phone, a smart watch, a temperature and humidity sensor, and a light sensor.
[0026] Collect user context awareness data through a smart phone, a smart watch, a temperature and humidity sensor, and a light sensor; Start the GPS module to obtain the accurate geographical location coordinates of the user at a frequency of once per minute.
[0027] Using the accelerometer and gyroscope, the intelligent terminal node records the attitude changes of the device once per second, and combines machine learning algorithms to analyze and identify the user's activity type in real time. The activity type can be walking, running, or driving.
[0028] The temperature and humidity sensor and the light sensor sample the data of the surrounding environment once every 5 minutes, and record the temperature, humidity, and light intensity respectively.
[0029] The user context awareness data refers to the user's geographical location, timestamp, activity type, temperature, humidity, and light intensity.
[0030] It should also be noted that such extensive data collection not only enhances the depth of understanding of user behavior but also provides a detailed basis for subsequent behavior pattern analysis.
[0031] S2: Analyze the user context awareness data using the long short-term memory network algorithm to obtain the user's daily behavior pattern.
[0032] The specific steps are as follows: S2.1. Normalize the user context awareness data using Z - score normalization and encode it using embedding encoding to ensure that data in different dimensions has the same scale, eliminate the influence of dimensions, and input the normalized and encoded user context awareness data into a multi - dimensional space to obtain a numerical vector of user context awareness data features. It should also be noted that: The normalization process enables subsequent machine learning algorithms to work more effectively.
[0033] S2.2. Use the long short - term memory network algorithm to analyze the numerical vector of user context awareness data features to form the user's daily behavior pattern. The expression is: ; Where, is the user's daily behavior pattern, is the geographical location coordinate at time , is the activity type encoding vector at time , is the temperature value at time , is the humidity value at time , is the light intensity value at time , is the long short - term memory network algorithm.
[0034] It should also be noted that: By using the LSTM algorithm, the long - term dependence relationships in user behavior can be effectively captured, solving the problem that traditional methods are difficult to handle non - linear and long - time - interval data.
[0035] S3: Use the user context awareness data and the user's daily behavior pattern for federated learning to form a dynamically optimized user behavior pattern, and predict the user's future behavior through a decision tree algorithm.
[0036] The specific steps are as follows: S3.1. Initialize the federated learning client on the intelligent terminal node, including setting configuration parameters and establishing a secure communication channel to ensure that each intelligent terminal node can independently execute local training tasks. The intelligent terminal node uses the user context awareness data and the user's daily behavior pattern to train the parameters of the long short - term memory network algorithm through the federated learning algorithm. The training process includes initializing the federated learning client, calculating the parameter gradients of the long short - term memory network algorithm, encrypting the parameter gradients of the long short - term memory network algorithm, aggregating, decrypting, and updating the parameters of the long short - term memory network algorithm.
[0037] It should also be noted that: This distributed training method of federated learning not only protects the privacy of users, but also reduces the bandwidth requirements for data transmission.
[0038] S3.2. Calculate the parameter gradients of the long short-term memory network algorithm through the parameter gradient formula. The calculation of the parameter gradients is completed by comparing the daily behavior patterns of users on each intelligent terminal node with the context awareness data at that moment and combining the partial derivatives of the loss function with respect to the parameters of the long short-term memory network algorithm. The expression is: ; where is the parameter gradient of the long short-term memory network algorithm, is the daily behavior pattern of users on the th intelligent terminal node, is the geographical location coordinate of the th intelligent terminal node at time , is the activity type coding vector of the th intelligent terminal node at time , is the temperature value of the th intelligent terminal node at time , is the humidity value of the th intelligent terminal node at time , is the light intensity value of the th intelligent terminal node at time , is the partial derivative of the loss function of the th intelligent terminal node with respect to the parameters of the long short-term memory network algorithm; First, calculate the difference between the current user behavior pattern and the actual context awareness data, and then use the chain rule to obtain the partial derivative of the loss function with respect to the parameters of the long short-term memory network algorithm.
[0039] S3.3. Use homomorphic encryption technology to encrypt the parameter gradients of the long short-term memory network algorithm. Homomorphic encryption technology allows computational operations to be performed on encrypted data without decrypting it, and upload the encrypted parameter gradients of the long short-term memory network algorithm to the central server through a secure channel; It should also be noted that: Homomorphic encryption technology can ensure the security of parameter gradients during transmission and processing.
[0040] S3.4. After the central server receives the encrypted long short-term memory network algorithm parameter gradients from all participating nodes, it uses the characteristics of homomorphic encryption technology to perform an aggregation calculation on the long short-term memory network algorithm parameter gradients to obtain the total gradient. The central server decrypts the aggregated total gradient using the private key it holds to obtain the global gradient in plaintext form, and at the same time distributes it to the intelligent terminal nodes to update the parameters of the long short-term memory network algorithm; It should also be noted that: after receiving the encrypted long short-term memory network algorithm parameter gradients, they can be directly aggregated and calculated without decrypting each individual gradient, thus avoiding the exposure of sensitive data.
[0041] S3.5. By inputting the numerical vector of the user context awareness data features into the updated long short-term memory network algorithm, a dynamic optimization of the user behavior pattern is formed, and the expression is: ; where, is the dynamic optimization of the user behavior pattern, is the geographical location coordinate at time , is the activity type coding vector at time , is the temperature value at time , is the humidity value at time , is the light intensity value at time , is the updated long short-term memory network algorithm; It should also be noted that: the formation of the dynamic optimization of the user behavior pattern ensures that the preloading strategy can adapt to the changes in the user behavior in real time, improving the response speed and the level of personalized service.
[0042] S3.6. Use Z-score standardization to normalize the dynamic optimization of the user behavior pattern to ensure that the respective eigenvalue of the dynamic optimization of the user behavior pattern is within the same scale range. Select the C4.5 algorithm as the decision tree algorithm, and use the normalized dynamic optimization of the user behavior pattern to calculate the probability of the user's future behavior through the decision tree algorithm. The expression is: ; where, is the probability of the user's future behavior, is the weight of the th decision tree, is the th decision tree's prediction output for the dynamic optimization of the user behavior pattern; It should also be noted that a smoothing factor is added to the denominator part to prevent the denominator from being zero, which increases the robustness of the model.
[0043] S3.7. After calculating the probabilities of the user's future behaviors, generate a list containing various possible behaviors and their corresponding probabilities. To determine the most likely future behavior of the user, compare these probability values and select the behavior with the highest probability as the final prediction result.
[0044] It should also be noted that selecting the behavior with the highest probability as the user's future behavior not only simplifies the subsequent processing logic but also provides clear guidance for preloading strategies and other personalized services.
[0045] S4: Generate a preloading strategy based on the user's future behavior.
[0046] The specific steps are as follows. S4.1. Use the intelligent terminal node to collect the user's historical interaction data, including viewing records, likes, comments, and favorites. Based on the user's historical interaction data, use the collaborative filtering algorithm to screen out candidate digital media. Use the matrix factorization algorithm to decompose the candidate digital media and the user's future behavior into the product of two low-dimensional matrices respectively, so as to extract the encoding vectors of the candidate digital media and the encoding vectors of the user's future behavior. It should also be noted that the matrix factorization technology can effectively extract potential features from high-dimensional sparse matrices, can significantly reduce the data dimension while maintaining the integrity of information, thereby accelerating the calculation speed.
[0047] S4.2. Calculate the similarity between the encoding vectors of the candidate digital media and the user's future behavior in different activity types through the correlation degree scoring formula to quantify the correlation degree between the two, and obtain the correlation degree score of the candidate digital media and the user's future behavior. The expression is: ; Where is the correlation degree score of the candidate digital media and the user's future behavior, is the weight of the th activity type, is the th candidate digital media's encoding vector in the th activity type, is the user's future behavior's encoding vector in the th activity type, is the number of activity types, is the smoothing factor, is the user's future behavior; It should also be noted that: This method for calculating the correlation degree based on the multi-dimensional coding vector not only improves the relevance and personalization level of the recommended content, but also provides a scientific basis for subsequent sorting and preloading strategies, ensuring that the content most likely to arouse the user's interest can be preferentially selected for preloading.
[0048] S4.3. Sort all candidate digital media using a sorting algorithm with a lower time complexity and higher efficiency according to the correlation degree score between the candidate digital media and the user's future behavior, and finally form a preloading strategy.
[0049] It should also be noted that: The preloading strategy not only improves the resource utilization efficiency, reduces the preloading of invalid content, but also significantly improves the user's satisfaction and service quality.
[0050] S5: Load digital media according to the user's current context awareness data and the preloading strategy, and record the user feedback.
[0051] The specific steps are as follows. S5.1. The preloading strategy is an ordered list of digital media. The candidate digital media are arranged in descending order of the correlation degree score between the candidate digital media and the user's future behavior to obtain a preloading digital media list. The intelligent terminal node will preferentially load the content most likely to arouse the user's interest according to this list, ensuring that the user can immediately access the relevant content when needed, and use the intelligent terminal node to collect the current user context awareness data; Calculate the probability distribution of the current user's behavior through a logistic regression algorithm for the current user context awareness data; It should also be noted that: In order to extract valuable behavior patterns from the current context awareness data, a logistic regression algorithm is used to calculate the current context awareness data, and a probability vector is output, indicating the possibility of the user in different behavior patterns.
[0052] S5.2. Use the matrix factorization algorithm to decompose the preloading digital media and the probability distribution of the current user's behavior into the product of two low-dimensional matrices respectively, so as to extract the coding vector of the preloading digital media and the coding vector of the probability distribution of the current user's behavior; It should also be noted that: The converted coding vector can be used for advanced data analysis tasks such as clustering analysis and anomaly detection to further explore the deep-seated association between user behavior patterns and content features.
[0053] S5.3. Quantify the similarity between the preloading digital media and the user's current behavior pattern through a matching degree scoring formula, and calculate the matching degree score between the preloading digital media and the probability distribution of the current user's behavior. The expression is: ; Where For the matching degree score of preloading digital media and the probability distribution of current user behavior, is the weight of the th activity type, is the th preloading digital media's encoding vector on the th activity type, is the encoding vector of the probability distribution of current user behavior on the th activity type, is the probability of current context awareness data on the th activity type, is the smoothing factor, is the probability distribution of current user behavior, is the index of the preloading digital media;
[0054] It should also be noted that: The matching degree scoring formula not only improves the relevance and personalization level of the recommended content, but also provides a scientific basis for subsequent selection and loading, ensuring that the content most likely to arouse the user's interest can be preferentially selected for loading.
[0055] It should also be noted that: The digital media with the highest matching degree score between the preloading digital media and the probability distribution of current user behavior not only reflects the user's current behavior pattern, but also takes into account the predicted results of the user's future behavior, making the recommendation more accurate and personalized.
[0056] S6: Adjust the preloading strategy according to user feedback.
[0057] The specific steps are as follows, S6.1. Use Z - score normalization to normalize user feedback to ensure that different types of user feedback can be compared on the same scale, and obtain the user feedback score through weighted summation; It should also be noted that: Through normalization, all feedback data can be converted to a unified interval, thus avoiding certain high - value feedback types from occupying too much weight in the calculation.
[0058] S6.2. Based on the correlation degree score between candidate digital media and the user's future behavior and the user feedback score, update the correlation degree scoring formula to form a user - feedback - enhanced correlation degree scoring formula, and the expression is: ; where, is the correlation degree score between the user - feedback - enhanced candidate digital media and the user's future behavior, is a regulation parameter for user feedback scoring is the user feedback score Reorder the candidate digital media list according to the relevance score between the enhanced candidate digital media and the user's future behavior based on user feedback
[0059] It should also be noted that: by continuously integrating the latest user feedback information, the recommendation strategy can be adjusted in a timely manner, and those contents that have been verified to be truly popular with users can be preferentially selected
[0060] This embodiment also provides a digital media transmission system based on user behavior, including: A collection module that collects user context awareness data A behavior module that uses the long short-term memory network algorithm to analyze the user context awareness data to obtain the user's daily behavior pattern A prediction module that uses the user context awareness data and the user's daily behavior pattern for federated learning to form a dynamically optimized user behavior pattern, and predicts the user's future behavior through the decision tree algorithm A preloading module that generates a preloading strategy based on the user's future behavior A recording module that loads digital media according to the user's current context awareness data and the preloading strategy, and records the user feedback An adjustment module that adjusts the preloading strategy according to the user feedback
[0061] This embodiment also provides a computer device applicable to the situation of the method for digital media transmission based on user behavior, 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 method for digital media transmission based on user behavior proposed in the above embodiment
[0062] This computer device can be a terminal. This computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the computer device housing, or an external keyboard, touchpad, or mouse, etc
[0063] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for realizing digital media transmission based on user behavior proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (abbreviated as SRAM), electrically erasable programmable read-only memory (abbreviated as EEPROM), erasable programmable read-only memory (abbreviated as EPROM), programmable read-only memory (abbreviated as PROM), read-only memory (abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disc.
[0064] In summary, the present invention realizes the dynamic optimization of the user behavior pattern by initializing the federated learning client on the intelligent terminal node and training the parameters of the long short-term memory network algorithm using user context awareness data and user daily behavior patterns, ensuring that the recommendation strategy always adheres to the immediate needs of the user. By uploading the encrypted parameter gradients to the central server through the homomorphic encryption technology, the central server decrypts and aggregates these gradients, updates the global parameters and then distributes them to the intelligent terminal node, realizing the continuous optimization of the global parameters while protecting user privacy.
[0065] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.
Claims
1. A method for digital media transmission based on user behavior, characterized in that: include, Collect user context-aware data; Use the long short-term memory network algorithm to analyze user context perception data and obtain the user's daily behavior pattern; Use user context-aware data and user daily behavior patterns for federated learning to form a dynamically optimized user behavior model and predict user future behavior through a decision tree algorithm; Generate preloading strategies based on user future behavior; Load digital media based on user current context awareness data and preloading strategies, and record user feedback; Adjust preloading strategy based on user feedback.
2. The method for digital media transmission based on user behavior as claimed in claim 1, characterized in that: The specific steps of collecting user context awareness data are as follows: Collect user context-aware data through IoT devices and sensors; The user context awareness data refers to the user's geographic location, timestamp, activity type, temperature, humidity, and light intensity.
3. The method for digital media transmission based on user behavior as claimed in claim 2, characterized in that: The method uses the long short-term memory network algorithm to analyze the user context perception data and obtain the user's daily behavior pattern. The specific steps are as follows: Normalizing and encoding the user context perception data, and inputting the normalized and encoded user context perception data into a multidimensional space to obtain a numerical vector of user context perception data features; The long short-term memory network algorithm is used to analyze the numerical vector of the user's contextual perception data characteristics to form the user's daily behavior pattern, which is expressed as: ； in, For users' daily behavior patterns, For the moment The geographic location coordinates of For the moment The activity type encoding vector, For the moment The temperature value, For the moment The humidity value, For the moment The light intensity value, It is a long short-term memory network algorithm.
4. The method for digital media transmission based on user behavior as claimed in claim 3, characterized in that: The method uses user context perception data and user daily behavior patterns for federated learning to form a dynamically optimized user behavior pattern and predicts user future behavior through a decision tree algorithm. The specific steps are as follows: Initialize the federated learning client on the smart terminal node. The smart terminal node uses the user context perception data and the user's daily behavior pattern to train the parameters of the long short-term memory network algorithm through the federated learning algorithm. The parameter gradient formula is used to calculate the parameter gradient of the long short-term memory network algorithm. The expression is: ； in, is the parameter gradient of the LSTM network algorithm, For the The daily behavior patterns of users of smart terminal nodes, For the The time of intelligent terminal nodes The geographic location coordinates of For the The time of intelligent terminal nodes The activity type encoding vector, For the The time of intelligent terminal nodes The temperature value, For the The time of intelligent terminal nodes The humidity value, For the The time of intelligent terminal nodes The light intensity value, For the The partial derivatives of the loss function of the intelligent terminal nodes with respect to the parameters of the LSTM network algorithm; Use homomorphic encryption technology to encrypt the parameter gradient of the long short-term memory network algorithm, and upload the encrypted parameter gradient of the long short-term memory network algorithm to the central server through a secure channel; After receiving the encrypted LSTM algorithm parameter gradients from all participating nodes, the central server decrypts and summarizes the LSTM algorithm parameter gradients and sends them to the intelligent terminal nodes to update the parameters of the LSTM algorithm. Through the updated long short-term memory network algorithm, a dynamic optimization user behavior pattern is formed, which is expressed as: ； in, To dynamically optimize user behavior patterns, For the moment The geographic location coordinates of For the moment The activity type encoding vector, For the moment The temperature value, For the moment The humidity value, For the moment The light intensity value, It is the updated long short-term memory network algorithm; The dynamic optimization user behavior pattern is normalized, and the C4.5 algorithm is selected as the decision tree algorithm. The probability of the user's future behavior is calculated by the decision tree algorithm using the normalized dynamic optimization user behavior pattern. The expression is: ； in, is the probability of the user's future behavior, For the The weight of a decision tree, For the The prediction output of a decision tree for dynamically optimizing user behavior patterns; According to the probability of the user's future behavior, the behavior with the highest probability is selected as the user's future behavior.
5. The method for digital media transmission based on user behavior as claimed in claim 4, characterized in that: The specific steps of generating a preloading strategy based on future user behavior are as follows: Use intelligent terminal nodes to collect user historical interaction data, and use collaborative filtering algorithms to filter out candidate digital media based on user historical interaction data; Using a matrix decomposition algorithm, converting the candidate digital media and the user's future behavior into a coding vector of the candidate digital media and a coding vector of the user's future behavior; The correlation score between the candidate digital media and the user's future behavior is calculated using the correlation score formula, which is expressed as follows: ； in, Score the relevance of candidate digital media to the user's future behavior, For the The weight of the activity type, For the The candidate digital media is The encoding vectors of the activity types, For the user's future behavior in The encoding vectors of the activity types, is the number of activity types, is the smoothing factor, For the user's future behavior; According to the correlation scores between the candidate digital media and the user's future behavior, a sorting algorithm is used to sort all the candidate digital media to obtain a preloading strategy.
6. The method for digital media transmission based on user behavior as claimed in claim 5, characterized in that: The steps of loading digital media according to the user's current context perception data and preloading strategy and recording user feedback are as follows: Form a preloaded digital media list according to the preload strategy, and use smart terminal nodes to collect current user context awareness data; The current user context perception data is calculated through a logistic regression algorithm to obtain the probability distribution of the current user behavior; Using a matrix decomposition algorithm, converting the probability distribution of the preloaded digital media and the current user behavior into a coding vector of the preloaded digital media and a coding vector of the probability distribution of the current user behavior; The matching score formula is used to calculate the matching score of the probability distribution of the preloaded digital media and the current user behavior. The expression is: ； in, Score the match between the probability distribution of preloaded digital media and current user behavior, For the The weight of the activity type, For the Preloaded digital media The encoding vectors of the activity types, is the probability distribution of the current user behavior in The encoding vectors of the activity types, is the probability distribution of the current user behavior in The probability of a type of activity, is the smoothing factor, is the probability distribution of the current user behavior, for preloading an index of digital media; The preloaded digital media having the highest matching score between the probability distribution of the preloaded digital media and the current user behavior is selected for loading, and the user feedback is recorded using the intelligent terminal node.
7. The method for digital media transmission based on user behavior as claimed in claim 6, characterized in that: The specific steps of adjusting the preloading strategy according to user feedback are as follows: Normalize user feedback and obtain user feedback scores through weighted summation; Based on the correlation scores between the candidate digital media and the user's future behavior and the user feedback scores, the correlation score formula is updated to form a user feedback enhanced correlation score formula, which is expressed as follows: ； in, Scoring the relevance of user feedback-enhanced candidate digital media to the user's future behavior, Tuning parameters for user feedback ratings, Rate user feedback; The preloading strategy is adjusted based on the relevance score of the candidate digital media enhanced by user feedback and the user's future behavior.
8. A digital media transmission system based on user behavior, based on the digital media transmission method based on user behavior according to any one of claims 1 to 7, characterized in that: include, A collection module collects user context awareness data; The behavior module uses the long short-term memory network algorithm to analyze the user's contextual perception data and obtain the user's daily behavior pattern; The prediction module uses user context perception data and user daily behavior patterns for federated learning to form a dynamically optimized user behavior model and predict the user's future behavior through a decision tree algorithm; Preloading module, which generates preloading strategies based on user future behavior; A recording module that loads digital media based on the user's current context awareness data and preloading strategy, and records user feedback; Adjust the module and adjust the preloading strategy based on user feedback.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for digital media transmission based on user behavior described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for digital media transmission based on user behavior described in any one of claims 1 to 7 are implemented.