Fair perception edge data caching method based on recommendation system

Abstract

The invention discloses a fair perception edge data caching method based on a recommendation system, and the method comprises the following steps: S1, obtaining historical request data of a user based on a recommendation system user, extracting data features from the historical request data, and calculating the similarity between the data features; s2, performing modeling on the user request by adopting a hox process; predicting a future user request by adopting a Horkes process model; s3, performing quantitative evaluation on the overall experience quality of the user based on the obtained data feature information, and optimizing a caching strategy by adopting a heuristic algorithm based on the obtained prediction result of the user request; and S4, giving a quantitative index of a fair coefficient based on the obtained data feature information and the user experience quality, and optimizing and adjusting the caching strategy in the S3 based on the quantitative index. According to the invention, while the overall system performance is improved, the service balance among different users is guaranteed, and more comprehensive guarantee is provided for the user experience in the edge computing environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of data caching, and in particular relates to a fairness-aware edge data caching method based on a recommendation system. Background Art

[0002] In recent years, with the popularity of mobile devices and the enhancement of their computing and network capabilities, various resource-intensive applications have emerged one after another, such as high-quality video streaming, virtual reality (VR), augmented reality (AR), interactive games, etc. The surge in these applications has not only significantly increased mobile data traffic, but also put forward higher requirements for low latency and high stability of services. According to statistics, global mobile data traffic continues to grow at an average annual rate of about 40%, which poses a huge challenge to network bandwidth, computing resources and data transmission. Due to geographical limitations, the traditional centralized processing model of cloud computing cannot well meet users' demand for low-latency services.

[0003] In order to solve the high latency problem in the cloud computing model, MEC (Mobile Edge Computing) has been proposed as a new distributed computing architecture. In MEC, edge servers are deployed at the edge of the network close to users. By caching data or performing computing tasks close to users, data transmission latency can be significantly reduced, thereby improving user experience. Specifically, edge servers in the MEC environment have a certain caching capability and can actively cache data that users may request, so that they can respond quickly when users make requests, reduce data acquisition time, and improve service real-time performance and user satisfaction.

[0004] For example, a Chinese patent document with publication number CN112105035A discloses a method and device for deploying a mobile edge computing server, which can ensure that each base station location can be covered by the mobile edge computing servers at other base station locations, thereby ensuring the QOE performance of the edge network.

[0005] However, since the cache space of edge servers is limited and their geographical coverage is relatively limited, how to optimize edge data caching strategies has become a major research challenge. Current edge caching strategies mainly focus on improving cache hit rates, reducing bandwidth requirements, and optimizing resource utilization, such as cache allocation based on content popularity prediction and priority cache allocation based on user demand frequency.

[0006] Nevertheless, most existing studies focus on reducing service latency, improving cache hit rate, and reducing bandwidth consumption, but often ignore the differences in quality of experience (QoE) between users and the fairness of caching strategies. When resources are limited, some users' requests may be prioritized, while other users may suffer from poor service experience due to insufficient resources, resulting in unbalanced QoE. In service-intensive application scenarios, such as video streaming and AR / VR applications, unbalanced QoE will directly affect the overall user experience and satisfaction.

[0007] Therefore, how to improve the cache hit rate while taking into account the service experience of different users has become a key issue in edge caching technology. Summary of the invention

[0008] The present invention provides a fairness-aware edge data caching method based on a recommendation system to solve the problems of insufficient utilization of user features and lack of fairness of caching strategies in the edge caching process.

[0009] A fairness-aware edge data caching method based on a recommendation system comprises the following steps:

[0010] S1. Based on the recommendation system user, obtain the user's historical request data, extract data features from the historical request data, and calculate the similarity between the data features;

[0011] S2. Use Hawkes process to model user requests; in the modeling process, an exponential decay function is selected as the kernel function, which exhibits a monotonically decreasing characteristic, and the maximum likelihood estimation method is used to estimate the parameters in the Hawkes process model; use the Hawkes process model to predict future user requests;

[0012] S3: Based on the data feature information obtained in S1, the overall user experience quality is quantitatively evaluated; then, based on the prediction results of user requests obtained in S2, a heuristic algorithm is used to optimize the cache strategy: the cache pair that can maximize the overall experience quality is preferentially selected until the cache space is filled, thereby finding an approximate optimal solution that maximizes the overall user experience quality;

[0013] S4: Based on the data feature information obtained in S1 and the user experience quality calculated in S3, a quantitative indicator of the fairness coefficient is given, and based on this quantitative indicator, the cache strategy in S3 is optimized and adjusted, so as to ensure the fairness of the experience between users while ensuring the high quality of the overall user experience.

[0014] The specific process of step S1 is:

[0015] The user request frequency is used as the score of the data, and the user-data score matrix M is constructed based on historical requests;

[0016] The user-data rating matrix is ​​decomposed by the non-negative matrix factorization algorithm to obtain the user feature matrix P and the data feature matrix Q;

[0017] An iterative update algorithm is used to optimize the user feature matrix P and the data feature matrix Q; the similarity between the data is calculated.

[0018] The goal of the non-negative matrix factorization algorithm is to find matrices P and Q so that the difference between the product of PQ and M is minimal. The formula is:

[0019]

[0020] In the formula, represents the difference between matrix M and the product of matrix P and Q, M ij Represents the number of times user i requests data j.

[0021] The iterative update algorithm is used to optimize the user feature matrix P and the data feature matrix Q. Specifically, the loss calculation is adopted, and the gradient descent method is continuously iterated. The Republic is as follows:

[0022]

[0023] in, Represents the value of the i-th row and j-th column in the N-th cycle of matrix P. Among them, Represents the value of the i-th row and j-th column in the N-th cycle of matrix Q.

[0024] The similarity between data is calculated by calculating the cosine similarity between data feature matrices. The formula is as follows:

[0025]

[0026] In the formula, Sim(d m ,d n ) represents d m ,d n The similarity between them, cos(Q m ,Q n ) is the cosine similarity of the vectorized representation of the two data, |U m ∩U n | represents the intersection of the sets of users who requested the two data, and Θ is the confidence threshold to avoid high similarity due to too few records.

[0027] In step S2, in the Hawkes process, the probability of an event occurring at a given time is affected by the occurrence of similar events in the past, and the intensity function of time slot t is expressed as:

[0028]

[0029] Among them, H tn represents the timestamp set of historical events, β is the background intensity, and φ(t-τ) is the kernel function that captures the impact of past events on the arrival rate of similar events.

[0030] The maximum likelihood estimation method is used to estimate the parameters in the Hawkes process model, and the formula is:

[0031]

[0032] Where λ(x) is the intensity function of the Hawkes distribution, and L(T) is the maximum natural function of the distribution given the distribution occurring at a given time.

[0033] In step S3, the overall user experience quality is quantitatively evaluated using the following formula:

[0034]

[0035] Among them, φ(i,j) is the propagation delay caused by user i requesting data j, Sim(d m ,d n ) is the similarity between the requested data and the returned data. QoE(i,j,m,n) is the quality of experience of the user's request.

[0036] In step S4, the fairness coefficient is defined as follows:

[0037]

[0038] Wherein, σ represents the standard deviation of the overall user experience quality calculated in S3, and H and L represent the upper bound and lower bound of the overall user experience quality, respectively.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention utilizes the recommendation system to effectively mine user preferences and data features. By combining the user's historical request data and utilizing technical means such as user-data rating matrix construction and non-negative matrix decomposition, the present invention can more accurately predict the user's potential needs. At the same time, the present invention introduces fairness indicators to quantitatively analyze the QoE differences of different users, ensuring that the caching strategy not only improves the overall service quality, but also balances the experience gap between users in multi-user scenarios, thereby achieving higher user satisfaction.

[0041] 2. The present invention combines the recommendation system with fairness analysis to optimize cache allocation by intelligently predicting user needs, while ensuring that all users obtain relatively fair service quality when sharing resources. This method is particularly suitable for service-intensive application scenarios, and can ensure service balance among different users while improving overall system performance, thereby providing more comprehensive protection for user experience in edge computing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of a fairness-aware edge data caching method based on a recommendation system according to an embodiment of the present invention.

[0043] Figure 2 The working status of the edge cache architecture of the recommendation system enabled for the embodiment of the present invention at a certain moment. DETAILED DESCRIPTION

[0044] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0045] like Figure 1 As shown, a fairness-aware edge data caching method based on a recommendation system includes the following steps:

[0046] S1. Data feature extraction: Extract data features from historical records for data value assessment and cache soft hit determination. The specific steps are as follows:

[0047] S11. User-data scoring matrix construction: The user request frequency is used as the score. The higher the request frequency, the higher the score. The scoring matrix M is constructed based on historical requests, where M ij Indicates the number of times user ui requests data dj.

[0048] S12, non-negative matrix decomposition: The non-negative matrix decomposition method is used to obtain the user feature matrix and the data feature matrix. Specifically, the algorithm tries to find matrices P and Q that can satisfy the minimum difference between the product of PQ and M, that is, minimize:

[0049]

[0050] The specific update is to use loss calculation and gradient descent to iterate continuously, that is,

[0051]

[0052] Finally, the user feature matrix P and data feature matrix Q can be obtained for subsequent calculations.

[0053] S13, matrix iterative update: use iterative update algorithm to optimize matrices P and Q.

[0054] S14, data similarity calculation: The similarity between data is determined by calculating the cosine similarity between data feature vectors, that is:

[0055]

[0056] S2. Predict user requests: Use the Hawkes process to predict future user requests and capture time correlation to improve cache hit rate. The specific steps are as follows:

[0057] S21. Behavior correlation analysis: Analyze the strong correlation between the user's historical behavior and future behavior. For example, if a user has frequently requested certain data recently, the user is more likely to continue requesting the data in the short term.

[0058] S22. Hawkes process modeling: Hawkes process is used to model user requests. Hawkes process is a self-exciting process in which the probability of an event occurring at a certain point in time is affected by the occurrence of similar events in the past. In Hawkes process, the probability of an event occurring at a given time is affected by the occurrence of similar events in the past. The intensity function of time slot t can be expressed as:

[0059]

[0060] S23, kernel function selection: The present invention selects an exponential decay function as the kernel function, and exhibits a monotonically decreasing characteristic to reflect the more significant impact of recent events on future request rates, namely:

[0061] φ(x)=αe -σ

[0062] Here, α and σ are the parameters of the function.

[0063] S24, parameter estimation and probability density function and likelihood function calculation: The maximum likelihood estimation method is used to estimate the parameters in the Hawkes process model. After constructing the maximum likelihood function, the expectation maximization (EM) algorithm is used to iteratively optimize the likelihood function, and further derive the probability density function f(t) and the corresponding Hawkes process parameters.

[0064] S3. Maximizing the allocation of user experience quality: The present invention considers that QoE depends on the latency of data retrieval and the similarity between the retrieved data and the requested data. However, the relationship between these factors and QoE is often nonlinear. For example, when the similarity is below a certain threshold, the returned data will no longer be consistent with the data expected by the user, resulting in a significant decrease in QoE. Therefore, an s-type function is used to describe this relationship:

[0065]

[0066] The data similarity can be obtained by extracting data features and decomposing feature matrices in S1. Then, the present invention designs a heuristic algorithm to optimize the cache strategy based on the prediction results of user requests obtained in S2: the cache pairs that can maximize the overall experience quality are preferentially selected until the cache space is filled, aiming to find an approximate optimal solution that can maximize the overall user experience quality. The specific steps are as follows:

[0067] S31. Initialize data, input: request set, cache block set, user set, edge server network G(S,E).

[0068] S32. First, initialize the gain wj of the cache block to the overall user experience quality of the system, and traverse each request in the request set to find the cache pair related thereto and update the gain of the cache block to the overall experience quality.

[0069] S33, iteratively selecting cache pairs: when the cache space is not full, continuously selecting cache blocks with the maximum gain.

[0070] S34, check cache space: if the cache space on the edge server is larger than the selected data size, cache the data on the server and update the server cache space.

[0071] S35. Update the system experience quality gain of the cache block: traverse the request set associated with the cache block, and update the system experience quality gain value of each request in the current cache state.

[0072] S36, repeat until the cache space is full: repeat the above steps, and continuously select the cache pair with the maximum gain until the cache space is fully utilized.

[0073] S4, cache allocation re-optimization: When maximizing the user experience quality allocation in step S4, the fairness of the cache strategy is first quantitatively evaluated based on the data feature information obtained in S1 and the user experience quality calculated in S3. The present invention introduces a fairness index F to evaluate the fairness of the cache strategy. The fairness index F is defined as follows:

[0074]

[0075] Among them, σ represents the standard deviation of the overall QoE of users calculated in S3, and H and L represent the upper and lower bounds of QoE, respectively. By minimizing the fluctuation of QoE (i.e., the standard deviation σ), the fairness index ensures that the cache strategy can improve the overall user experience quality while minimizing the experience gap between different users. Then, based on the fairness index, the cache strategy in S3 is optimized and adjusted, so as to ensure the fairness of the experience between users while ensuring the high quality of the overall user experience. The specific steps are as follows:

[0076] S41. Select the user with the lowest QoE: First find the user with the lowest QoE, and try to gradually optimize the requests of those users whose QoE is lower than the median to improve the fairness of the caching strategy.

[0077] S42, finding a cache pair with the minimum gain: finding a cache pair associated with the request to be optimized, and selecting a cache pair with the minimum gain.

[0078] S43, cache replacement: replace the data currently stored in the cache with the requested data to improve QoE.

[0079] S44, target weighting and checking: Check whether the overall value of the weighted target has been improved after replacement. If not, restore the data to the original.

[0080] S45, repeat iteration: repeat the above process until all requests to be optimized are completed.

[0081] Recommendation systems can not only mine the characteristics of users and data, but also significantly influence the pattern of user requests. By integrating recommendation systems into mobile edge data caches, the overall benefits of caching strategies are significantly improved. It is worth noting that the data requested by users is usually not irreplaceable. When the requested data is not cached in nearby edge servers, recommending other highly relevant content can also meet the user's needs. This process is called "soft hit".

[0082] Figure 2 The figure shows the working of the edge cache architecture that enables the recommendation system at a certain moment. When user u1 requests data d1, edge server s1 returns the cached data d1 to the user. In the case of user u2 requesting data d2, the system can choose to obtain data d1 from a nearby edge server (s1 or s3), or obtain it from the cloud depending on the latency. If user u3 requests data d3, he has two options: 1) Direct hit cache: Get data d3 from the cloud, although it provides accurate content, but the latency is higher; 2) Soft hit cache: Accept data d1 cached in edge server s3 that is similar to data d3, thereby reducing latency.

[0083] Through the cache strategy supported by this recommendation system, the present invention not only achieves significant results in improving the quality of user experience, but also further reduces the overall response delay of the system. The concept of soft hit combines the advantages of the recommendation system and the cache strategy, so that the system can replace the missed data to a certain extent in the resource-constrained edge environment, thereby achieving optimal utilization of resources.

[0084] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A fairness-aware edge data caching method based on a recommendation system, characterized in that: The following steps are involved: S1. Based on the recommendation system user, obtain the user's historical request data, extract data features from the historical request data, and calculate the similarity between the data features; S2. Use Hawkes process to model user requests; in the modeling process, an exponential decay function is selected as the kernel function, which exhibits a monotonically decreasing characteristic, and the maximum likelihood estimation method is used to estimate the parameters in the Hawkes process model; use the Hawkes process model to predict future user requests; S3: Based on the data feature information obtained in S1, the overall user experience quality is quantitatively evaluated; then, based on the prediction results of user requests obtained in S2, a heuristic algorithm is used to optimize the cache strategy: the cache pair that can maximize the overall experience quality is preferentially selected until the cache space is filled, thereby finding an approximate optimal solution that maximizes the overall user experience quality; S4: Based on the data feature information obtained in S1 and the user experience quality calculated in S3, a quantitative indicator of the fairness coefficient is given, and based on this quantitative indicator, the cache strategy in S3 is optimized and adjusted, so as to ensure the fairness of the experience between users while ensuring the high quality of the overall user experience.

2. The fairness-aware edge data caching method based on the recommendation system according to claim 1, characterized in that: The specific process of step S1 is: The user request frequency is used as the score of the data, and the user-data score matrix M is constructed based on historical requests; The user-data rating matrix is ​​decomposed by the non-negative matrix factorization algorithm to obtain the user feature matrix P and the data feature matrix Q; An iterative update algorithm is used to optimize the user feature matrix P and the data feature matrix Q; the similarity between the data is calculated.

3. The fairness-aware edge data caching method based on the recommendation system according to claim 2 is characterized in that: The goal of the non-negative matrix factorization algorithm is to find matrices P and Q so that the difference between the product of PQ and M is minimal. The formula is: In the formula, represents the difference between matrix M and the product of matrix P and Q, M ij Represents the number of times user i requests data j.

4. The fairness-aware edge data caching method based on the recommendation system according to claim 2, characterized in that: The iterative update algorithm is used to optimize the user feature matrix P and the data feature matrix Q. Specifically, the loss calculation is adopted, and the gradient descent method is continuously iterated. The Republic is as follows: in, Represents the value of the i-th row and j-th column in the N-th cycle of matrix P; Represents the value of the i-th row and j-th column in the N-th cycle of matrix Q.

5. The fairness-aware edge data caching method based on the recommendation system according to claim 2, characterized in that: The similarity between data is calculated by calculating the cosine similarity between data feature matrices. The formula is as follows: In the formula, Sim(d m ,d n ) represents d m ,d n The similarity between them, cos(Q m ,Q n ) is the cosine similarity of the vectorized representation of the two data, |U m ∩U n | represents the intersection of the sets of users who requested the two data, and Θ is the confidence threshold to avoid high similarity due to too few records.

6. The fairness-aware edge data caching method based on the recommendation system according to claim 1, characterized in that: In step S2, in the Hawkes process, the probability of an event occurring at a given time is affected by the occurrence of similar events in the past, and the intensity function of time slot t is expressed as: Among them, H tn represents the timestamp set of historical events, β is the background intensity, and φ(t-τ) is the kernel function that captures the impact of past events on the arrival rate of similar events.

7. The fairness-aware edge data caching method based on the recommendation system according to claim 1, characterized in that: The maximum likelihood estimation method is used to estimate the parameters in the Hawkes process model, and the formula is: Where λ(x) is the intensity function of the Hawkes distribution, The maximum natural function of a distribution given the distribution of occurrences at a given time.

8. The fairness-aware edge data caching method based on recommendation system according to claim 1, characterized in that: In step S3, the overall user experience quality is quantitatively evaluated using the following formula: Among them, φ(i,j) is the propagation delay caused by user i requesting data j, Sim(d m ,d n ) is the similarity between the requested data and the returned data; QoE(i,j,m,n) is the quality of experience of the user's request.

9. The fairness-aware edge data caching method based on the recommendation system according to claim 1, characterized in that: In step S4, the fairness coefficient is defined as follows: Wherein, σ represents the standard deviation of the overall user experience quality calculated in S3, and H and L represent the upper bound and lower bound of the overall user experience quality, respectively.

Images