Cache optimization method based on file freshness and content recommendation

Abstract

The invention relates to the technical field of wireless communication, in particular to a cache optimization method based on file freshness and content recommendation, and the method comprises the following steps: a BS collects file requests initiated by users before each time slot and randomly responds to requests of R users, and when a user i requests a file j and is responded by the BS, the user i sends the file j to the BS; the method comprises the following steps: firstly, determining whether a file j meeting an information age threshold requirement exists in users in a D2D communication range of a user i, and if so, acquiring the file j from the users in the D2D communication range of the user i for caching; if the file does not exist, a recommendation algorithm is executed, an appropriate file is selected from the files cached in the users within the D2D communication range of the user i to be recommended to the user i, and if the user i refuses to accept the recommended file, the requested file j is obtained from the BS to be cached. The system energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a cache optimization method based on file freshness and content recommendation. Background Art

[0002] The explosive growth of mobile data traffic in wireless networks and users' requirements for quality of service (QoS) have put tremendous pressure on base stations (BS) in traditional cellular networks. Compared with storing files directly in the BS, caching content objects on network edge devices can respond to some user requests locally, thereby reducing transmission delays and lowering system energy consumption.

[0003] Although content caching at the edge of the network can reduce the cost of network traffic transmission and communication energy consumption, there is also the problem of cached files becoming outdated, which affects the user experience. Summary of the invention

[0004] On the basis of maintaining file freshness, the present invention introduces content recommendation mechanism into cache decision, aims to reduce system energy consumption and improve cache hit rate, and proposes a cache optimization method based on file freshness and content recommendation.

[0005] The present invention adopts the following technical solutions:

[0006] A cache optimization method based on file freshness and content recommendation, comprising:

[0007] Construct a cellular D2D communication network model with U users and a BS. The BS is equipped with a file library of size F, and the size of each file is x. j ,The cache capacity of each user device is M, and there are T time slots of equal length in each transmission frame. The user can obtain the requested file within one time slot, and the BS file library always caches the latest file;

[0008] The BS collects file requests initiated by users before each time slot and randomly responds to the requests of R users. When user i requests file j and is responded to by the BS, it first confirms whether there is file j that meets the information age threshold requirement among the users within the D2D communication range of user i. If so, file j is obtained from the users within the D2D communication range of user i for caching. If not, the recommendation algorithm is executed to select a suitable file from the files cached among the users within the D2D communication range of user i and recommend it to user i. If user i refuses to accept the recommended file, the requested file j is obtained from the BS for caching.

[0009] A further improvement is that, if there are multiple files j meeting the information age threshold requirement among users within the D2D communication range of user i, a file j with the smallest information age is selected for caching.

[0010] A further improvement is that the recommendation algorithm includes:

[0011] Calculate the user interest matrix I based on the file topic according to the user's historical request records U×F , and then calculate the cosine similarity between users according to the user interest matrix, find the user k with the highest interest similarity to user i among the users within the D2D communication range of user i, find at most N files that user k is most interested in and meet the information age threshold requirements in the files cached by user k and recommend them to user i. If user k does not have files that meet the requirements, find the user with the second highest interest similarity to user i, find at most N files that the second highest user is most interested in and meet the information age threshold requirements in the files cached by the second highest user and recommend them to user i, and so on.

[0012] A further improvement is to calculate the user interest matrix I based on the file topic according to the user's historical request records U×F The methods include:

[0013] Assume that all the documents in the document library cover a total of Y topics, using the set Ω = {ω1,ω2,...ω k ,...ω Y} indicates that the subject ω k The attribute function between and file f is defined as:

[0014]

[0015] use Represents user u's opinion on the document topic ω k The preference of is formulated using the concept of mutual information and is expressed as:

[0016]

[0017] Among them, Z(ω k ) represents the subject ω k The collection of all files of Φ U represents the historical request records of user u, P(Z(ω k )|Φ U ) represents the theme ω k The probability of appearing in the historical request records of user u, and P(Z(ω k )) represents the theme ω k The probability of appearing in the entire file library, therefore, user u’s interest in file f is defined as:

[0018]

[0019] Among them, user u’s interest in file f is η u,f ∈[-1,1],η u,f The closer the value is to -1, the less interested the user is, and the closer it is to 1, the more interested the user is.

[0020] Finally, according to user u's interest in file f, u,f , get the user interest matrix I based on the file topic U×F .

[0021] A further improvement is that the information age of the file at the BS is set to 0, and the information age is gradually increased as the file spreads in the network, using δ ij (t) represents the information age of file j cached by user i in time slot t, and it is assumed that all requested content is delivered within one time slot. If file j cached at user i is not updated, then when the time slot increases by 1, the age of file j cached at user i will increase by 1. When user i caches a new version of file j, if file j is obtained from the BS, the information age of file j cached at user i will drop to 1. If file j is obtained from user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j cached at user k. If file j is recommended by user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j recommended by user k. The information age function of file j cached by user i is:

[0022]

[0023] Let binary array and represents the source of cached file j at user i in time slot t. If the file j requested by user i is obtained from a user within the D2D communication range, then If the file j requested by user i is obtained from BS, then If the file j requested by user i is recommended by a user within the D2D communication range, then

[0024] A further improvement is to use the binary variable a ij (t) is used to record whether the request for cached file j at user i in time slot t is served. If it is successfully served, then a ij (t)=1, otherwise, a ij (t) = 0; using binary variable b ij (t) to record whether user i has cached file j in time slot t. If it has, then bij (t) = 1, otherwise, b ij (t) = 0, when the cache file j at user i is unloaded at time slot t, b ij (t) = 0, the total size of the user's cached files cannot exceed the user's maximum cache capacity M, then:

[0025]

[0026] in,

[0027]

[0028] in,

[0029] A further improvement is that, assuming that the user set within the D2D communication range of user i is represented by D i To express, the minimum age of file j cached by user i among users within the D2D communication range at time slot t is The minimum age of file j obtained by user i through recommendation within the D2D communication range at time slot t is According to formula (4), we can get:

[0030]

[0031] Assume that the age of the file information obtained by the user cannot exceed an information age threshold δ th ,Right now:

[0032]

[0033] This ensures that the files obtained by users are not outdated and meet the information age threshold requirements.

[0034] A further improvement is to assume that the binary variable c ij (t) indicates whether there is a file j that meets the cache request in the D2D communication range of user i in time slot t. If yes, c ij (t)=1, otherwise c ij (t) = 0, c ij (t) is expressed as:

[0035]

[0036] in,

[0037] When user i’s request for cached file j is served in time slot t, i.e., a ij (t) = 1, then the file can be obtained from users or BSs within the D2D communication range or by accepting recommended files, so:

[0038]

[0039] in,

[0040] A further improvement is that the method also includes: when user i receives file j, if there is an old version of file j in the cache of user i, the old version of file j is directly replaced with the new version of file j; if the cache space of user i is sufficient, when caching the new file j, it is directly cached; if the cache space of user i is full, a content uninstallation operation is performed, following the principle of uninstalling outdated old files first.

[0041] A further improvement is that the method further includes analyzing the system energy consumption: assuming that the BS and the user each have a fixed power P B and P D To transfer files, the user recommends that the transfer power of the file be P R , assuming that large-scale fading is considered in the communication process, SNR represents the signal-to-noise ratio, then the energy consumption of file j transmission is expressed as:

[0042]

[0043] Among them, P x is the transmission power, x j is the size of file j, B is the system bandwidth, SNR = P x ψ -α / σ 2 ,ψ represents the communication distance, α is the path loss index, σ 2 is the power average of Gaussian white noise;

[0044] Assume that the energy consumption from the BS, from the users within the D2D communication range, and the files transmitted through the recommendation are E B ,E D and E R , then the energy consumption of the system in a single time slot is expressed as:

[0045]

[0046] Among them, 1≤t≤T.

[0047] The beneficial effects of the present invention are:

[0048] The present invention proposes a cache optimization method based on file freshness and content recommendation, which can improve cache hit rate and reduce system energy consumption while maintaining file freshness. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Schematic diagram of the cellular D2D communication network model system in the present invention;

[0050] Figure 2 A schematic diagram of the information age evolution process of user cache files of the present invention;

[0051] Figure 3 A flow chart of the algorithm implementation method for the cache decision process;

[0052] Figure 4 A flowchart of the algorithm implementation method for the file recommendation strategy;

[0053] Figure 5 is the relationship between the average information age and the number of documents;

[0054] Figure 6 is the relationship diagram between cache hit rate and information age threshold;

[0055] Figure 7 This is a graph showing the relationship between cache hit rate and the number of files;

[0056] Figure 8 It is the relationship diagram between system energy consumption and user cache capacity;

[0057] Fig. 9 is the relationship diagram between system energy consumption and information age threshold;

[0058] Fig.10 Figure 2 is a graph showing the relationship between system energy consumption and the number of files. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0060] Please refer to the attached Figure 1 -Attached Fig.10 , an embodiment of the present invention proposes a cache optimization method based on file freshness and content recommendation, including:

[0061] like Figure 1 As shown in the figure, a cellular D2D communication network model with U users and one BS is constructed. The BS is equipped with a file library of size F, and the size of each file is x j, the cache capacity of each user device is M, assuming that the users in the cell are uniformly distributed, there are T time slots of equal length in each transmission frame, and time slot-based scheduling is implemented. The user can obtain the requested file within a time slot, and the BS file library always caches the latest file.

[0062] It can be understood that, assuming that OFDMA (Orthogonal Frequency Division Multiple Access) is used in each time slot in the network, before the start of each time slot, the BS collects file requests from all users and can serve R user requests at the same time. According to the user location and the unique identifier and file freshness of the file cached in each user device, the BS determines the access route of the file requester, allocates orthogonal channel resources to them and returns a request failure signal to the unserved users. The user's requests are divided into new files that are not currently cached and new versions of files that are currently cached.

[0063] In the research of caching strategies in the prior art, most of the studied user request models for files are Zipf distribution models, which cannot accurately reflect the user's actual request content needs. Therefore, the present invention predicts the user's content preference by analyzing the real data set, thereby establishing a user request model to more realistically simulate the cache network scenario.

[0064] To this end, the present invention constructs a file request model based on user preference prediction. Specifically, in order to obtain the user's request probability for a file, the present invention obtains the user's rating of the movie as the user's historical request record for the file by analyzing the MovieLens1M dataset, and forms a rating matrix S with the user's rating of the movie. The value s of the rating matrix S is u,f ∈{0,1,2,3,4,5} represents the degree of user u’s liking for file f. u,f The larger the value, the more user u likes file f, s u,f =0 means user u has not rated file f.

[0065] In real life, the number of files in the database is extremely large, and it is impossible for users to rate all of them, which means that it is relatively sparse and cannot accurately reflect user preferences.

[0066] Therefore, the present invention adopts a collaborative filtering algorithm to predict the user's preference for unrated files: First, the correlation coefficient μ between the scores of user i and user j on file f is calculated based on the Pearson correlation coefficient: i,j :

[0067]

[0068] Among them, s i,frepresents the score of user i on file f, s j,f represents the rating of user j on file f, represents the average rating of all files by user i, represents the average rating of all files by user j.

[0069] Since formula (14) does not take into account the number of rated files, when the number of files rated together is small, the similarity between the two users calculated will be very high, resulting in low reliability of the obtained user similarity. Therefore, it is necessary to perform maximum and minimum normalization correction on the obtained correlation coefficient, which is expressed as follows:

[0070]

[0071] μ i,j =θμ i,j (16)

[0072] Where Q i,j represents the number of files that user i and user j have rated together, and They represent the minimum and maximum number of files rated by user i and other users respectively.

[0073] Set a threshold μ th Used to filter users who meet the threshold requirements. i,j >μ th When , it means that user i and user j have similar rating behaviors and are put into user set U is , that is, j∈U is , according to the obtained user set U is Predict the files that user i has not rated, expressed as:

[0074]

[0075] Then, the file score matrix obtained by predicting the supplementary score is used to construct the request probability p of user u for file f. u,f :

[0076] First, the value of the scoring matrix is ​​normalized to the maximum and minimum to obtain To represent the difference in user ratings for different files, it is defined as:

[0077]

[0078] in, and Represent the lowest and highest scores of the user in the file library respectively, then the probability of user u requesting file f is expressed as:

[0079]

[0080] Finally, the popularity of the file is obtained based on the user's preferences for the file:

[0081]

[0082] The method of the present invention is described in more detail below:

[0083] The BS collects file requests initiated by users before each time slot and randomly responds to the requests of R users. When user i requests file j and is responded to by the BS, it first confirms whether there is file j that meets the information age threshold requirement among the users within the D2D communication range of user i. If so, file j is obtained from the users within the D2D communication range of user i for caching. If not, the recommendation algorithm is executed to select a suitable file from the files cached among the users within the D2D communication range of user i and recommend it to user i. If user i refuses to accept the recommended file, the requested file j is obtained from the BS for caching.

[0084] In this embodiment, further, if there are multiple files j meeting the information age threshold requirement among users within the D2D communication range of user i, a file j with the smallest information age is selected for caching to meet the information age requirement to the greatest extent.

[0085] Furthermore, the recommendation algorithm includes:

[0086] Calculate the user interest matrix I based on the file topic according to the user's historical request records U×F , and then calculate the cosine similarity between users based on the user interest matrix, find the user k with the highest interest similarity to user i in the D2D communication range of user i, find at most N files that user k is most interested in and meet the information age threshold requirements in the files cached by user k and recommend them to user i. If user k does not have files that meet the requirements, find the user with the second highest interest similarity to user i, find at most N files that the second highest user is most interested in and meet the information age threshold requirements in the files cached by the second highest user and recommend them to user i, and so on. User i selects a file based on the degree of interest and recommends it to user i with probability p. r Accept the file.

[0087] Specifically, the user interest matrix I based on the file topic is calculated according to the user's historical request records: U×F The methods include:

[0088] Assume that all the documents in the document library cover a total of Y topics, using the set Ω = {ω1,ω2,...ω k ,...ω Y} indicates that when content is recommended to users, users are more likely to accept content related to topics that interest them.

[0089] For example, if the user is most interested in content related to sports or games, then when the user searches for content, the probability that he clicks on the recommended content related to the topic is very high, while the probability of being clicked on content recommended to the user about travel or opera is very small.

[0090] Themeω k The attribute function between and file f is defined as:

[0091]

[0092] use Represents user u's opinion on the document topic ω k The preference of is formulated using the concept of mutual information and is expressed as:

[0093]

[0094] Among them, Z(ω k ) represents the subject ω k The collection of all files of U represents the historical request records of user u, P(Z(ω k )|Φ U ) represents the theme ω k The probability of appearing in the historical request records of user u, and P(Z(ω k )) represents the theme ω k The probability of appearing in the entire file library, therefore, user u’s interest in file f is defined as:

[0095]

[0096] Among them, user u’s interest in file f is η u,f ∈[-1,1],η u,f The closer the value is to -1, the less interested the user is, and the closer it is to 1, the more interested the user is.

[0097] When a user accepts recommended files, priority is given to files that are of greatest interest and that meet the information age threshold requirements.

[0098] Finally, according to user u's interest in file f, u,f , get the user interest matrix I based on the file topic U×F .

[0099] In the embodiment of the present invention, the age of information (AoI) of a file is used as an indicator for evaluating the freshness of a file, such as Figure 2As shown in , it is defined as the time from the latest file generation to the present, that is, the information age of the file at the BS is set to 0, and the information age number gradually increases as the file spreads in the network.

[0100] Use δ ij (t) represents the information age of file j cached by user i in time slot t, and it is assumed that all requested content is delivered within one time slot. If file j cached at user i is not updated, then when the time slot increases by 1, the age of file j cached at user i will increase by 1; when user i caches a new version of file j, the file information age will drop to a smaller value. If file j is obtained from the BS, the information age of file j cached at user i will drop to 1. If file j is obtained from user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j cached at user k. If file j is recommended by user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j recommended by user k. The information age function of file j cached by user i is:

[0101]

[0102] Let binary array and represents the source of cached file j at user i in time slot t. If the file j requested by user i is obtained from a user within the D2D communication range, then If the file j requested by user i is obtained from BS, then If the file j requested by user i is recommended by a user within the D2D communication range, then

[0103] If there is no cached file j at user k in time slot t, then δ kj (t)=0.

[0104] In the present invention, the cache decision process is constrained, including:

[0105] (1) Cache capacity constraints:

[0106] Using binary variable a ij (t) is used to record whether the request for cached file j at user i in time slot t is served. If it is successfully served, then a ij (t)=1, otherwise, a ij (t) = 0; using binary variable b ij (t) to record whether user i has cached file j in time slot t. If it has, then b ij (t) = 1, otherwise, b ij(t) = 0, when the cache file j at user i is unloaded at time slot t, b ij (t) = 0, the total size of the user's cached files cannot exceed the user's maximum cache capacity M, then:

[0107]

[0108] in,

[0109]

[0110] in,

[0111] (2) Age of Information (AoI) Constraints:

[0112] Assume that the user set within the D2D communication range of user i is represented by D i To express, the minimum age of file j cached by user i among users within the D2D communication range at time slot t is The minimum age of file j obtained by user i through recommendation within the D2D communication range at time slot t is According to formula (4), we can get:

[0113]

[0114] Assume that the age of the file information obtained by the user cannot exceed an information age threshold δ th ,Right now:

[0115]

[0116] This ensures that the files obtained by users are not outdated and meet the information age threshold requirements.

[0117] (3) File access constraints:

[0118] Assume that the binary variable c ij (t) indicates whether there is a file j that meets the cache request in the D2D communication range of user i in time slot t. If yes, c ij (t)=1, otherwise c ij (t)=0.

[0119] Whether the users within the D2D communication range can provide the requested file depends on the type of file requested. When a user requests a new file, it depends on whether the users within the D2D communication range contain the requested file. When a user requests a new version, it depends on whether the users within the D2D communication range contain an updated version of the requester's cached file. ij (t) is expressed as:

[0120]

[0121] in,

[0122] When user i’s request for cached file j is served in time slot t, i.e., a ij (t) = 1, then the file can be obtained from users or BSs within the D2D communication range or by accepting recommended files, so:

[0123]

[0124] in,

[0125] In this embodiment, the method also includes: when user i receives file j, if there is an old version of file j in the cache of user i, the old version of file j is directly replaced with the new version of file j; if the cache space of user i is sufficient, when caching the new file j, it is directly cached; if the cache space of user i is full, a content uninstallation operation is performed, following the principle of uninstalling outdated old files first.

[0126] In this embodiment, the method further includes analyzing the system energy consumption: assuming that the BS and the user each use a fixed power P B and P D To transfer files, the user recommends that the transfer power of the file be P R , assuming that large-scale fading is considered in the communication process, SNR represents the signal-to-noise ratio, then the energy consumption of file j transmission is expressed as:

[0127]

[0128] Among them, P x is the transmission power, x j is the size of file j, B is the system bandwidth, SNR = P x ψ -α / σ 2 ,ψ represents the communication distance, α is the path loss index, σ 2 is the average power of Gaussian white noise.

[0129] Assume that the energy consumption from the BS, from the users within the D2D communication range, and the files transmitted through the recommendation are E B ,E D and E R , then the energy consumption of the system in a single time slot is expressed as:

[0130]

[0131] Among them, 1≤t≤T.

[0132] The following is a simulation and analysis of a cache optimization method based on file freshness and content recommendation proposed in an embodiment of the present invention:

[0133] The present invention simulates the relationship between average information age, system energy consumption, cache hit rate and the number of files, cache capacity and age threshold to evaluate the cache optimization method based on file freshness and content recommendation proposed in the present invention.

[0134] This paper compares the proposed cache optimization method (Our Proposed Strategy) with the following two file access strategies:

[0135] Content-oriented strategy: When a user sends a request to the BS and the BS responds, the requested content, whether it is a new file or a new version of the file, can be obtained from neighboring users or BS.

[0136] Version-oriented strategy: After a user sends a request and receives a response, if the request is for a new version of a file, it is only obtained from the BS. If the request is for a new file, it can be obtained from neighboring users or BSs.

[0137] The data used to calculate user preferences in the present invention comes from the MovieLens1M dataset, and the specific parameter settings of the simulation are shown in Table 1.

[0138] Parameters Value BS radius: RB 500m D2D coverage radius:RD 150m Total number of time slots:T 500 The maximum number of orthogonal subcarriers:R 80 <![CDATA[The transmission power of BS:P b ]]> 10w <![CDATA[The transmission power of D2D User:P d ]]> 0.5w <![CDATA[The transmission power of Recommendation:P r ]]> 1w The number of files:F 200 The number of users: 100 Size of cache on a user:M 20 <![CDATA[File size:x j ]]> 1Mbits <![CDATA[AoI threshold:δ th ]]> 15 System bandwidth:B 10MHz Path loss exponent: α 2 <![CDATA[Average Gaussian white noise power:σ 2 ]]> 0.0001

[0139] Table 1

[0140] like Figure 5 As shown in the figure, as the number of files increases, the average information age of files of the three strategies shows a downward trend. This is because the repetition rate of files requested by users in each time slot will decrease as the number of files increases, and the proportion of new file requests will increase. Therefore, users will obtain new files with high freshness more frequently and continuously discard unpopular and outdated files in the cache device.

[0141] like Figure 6 As shown, as the information age threshold increases, the cache hit rates of the three strategies all show an upward trend. This is because when the information age threshold increases, the number of files in the system that can meet the information age requirements of the user's requested files increases, and the proportion of the system obtaining the required files within the D2D communication range increases, so the cache hit rate will continue to increase. Among them, the strategy of the present invention performs the best, because when the user cannot find the requested file within the D2D communication range, the recommendation algorithm will recommend content that is similar to the requested file and of interest to the user, thereby further improving the cache hit rate.

[0142] like Figure 7 As shown, as the number of files increases, the cache hit rates of the three strategies all show a downward trend. This is because when the number of files in the network increases, the repetition rate of files requested by users decreases, and the probability that the same files are cached in the users within the D2D communication range decreases, thereby reducing the cache hit rate. When the number of files increases, the downward trend of the strategy of the present invention tends to be stable. This is because when the file requested by the user is not within the D2D communication range, the content will be recommended to the user first. When the number of files increases, the recommendation algorithm works more, and the cache hit rate will not continue to decrease but tend to be stable.

[0143] like Figure 8 As shown in the figure, as the cache capacity increases, the system energy consumption of the content-based strategy and the version-based strategy increases. This is because higher cache capacity will leave more outdated files in the user, and in order to meet the age constraint requirements, more files will be sourced from the BS. The system energy consumption of the strategy of the present invention tends to be stable and much lower than the other two strategies. This is because the recommendation algorithm will recommend similar files to users whose information age meets the requirements, and there is no need to frequently update files from the BS, thereby greatly reducing system energy consumption.

[0144] like Fig. 9 As shown, with the increase of the information age threshold, the system energy consumption of the three strategies all show a downward trend. This is because the increase of the information age threshold means that the user's requirements for file freshness are reduced, and the probability that the file requested by the user can be obtained nearby becomes greater, thereby reducing the system energy consumption. Among them, the strategy of the present invention performs best. The application of the recommendation algorithm makes it unnecessary for users to frequently obtain files from the BS even if they cannot obtain files nearby.

[0145] like Fig.10 As shown in the figure, as the number of files increases, the system energy consumption of the content-based strategy and the version-based strategy shows an upward trend. This is because when the number of files increases, the repetition rate of user requests decreases, the proportion of user-requested files obtained from the BS increases, and the system energy consumption increases. The system energy consumption of the strategy of the present invention does not change with the number of files and is much lower than the other two strategies. This is because under the work of the recommendation algorithm, users can obtain files similar to the requested files and of interest to themselves from surrounding users, so they do not need to frequently access the BS, which greatly reduces the system energy consumption.

[0146] The cache optimization method based on file freshness and content recommendation proposed in the present invention can improve cache hit rate and reduce system energy consumption while maintaining file freshness. Simulation experiments show that in terms of file freshness, the strategy of the present invention improves the performance by 29.5% over the content-based strategy and 29.8% over the version-based strategy. In terms of cache hit rate, the strategy of the present invention improves the performance by 17.4% over the content-based strategy and 58.1% over the version-based strategy. In terms of system energy consumption, the strategy of the present invention improves the performance by 67.6% over the content-based strategy and 74.3% over the version-based strategy. It can be seen that the cache strategy proposed in the present invention that comprehensively considers cache decisions, content recommendations, user preferences and file freshness significantly improves the performance of the system.

[0147] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A cache optimization method based on file freshness and content recommendation, characterized in that: include: Construct a cellular D2D communication network model with U users and a BS. The BS is equipped with a file library of size F, and the size of each file is x. j ,The cache capacity of each user device is M, and there are T time slots of equal length in each transmission frame. The user can obtain the requested file within one time slot, and the BS file library always caches the latest file; The BS collects file requests initiated by users before each time slot and randomly responds to the requests of R users. When user i requests file j and is responded to by the BS, it first confirms whether there is file j that meets the information age threshold requirement among the users within the D2D communication range of user i. If so, file j is obtained from the users within the D2D communication range of user i for caching. If not, the recommendation algorithm is executed to select a suitable file from the files cached among the users within the D2D communication range of user i and recommend it to user i. If user i refuses to accept the recommended file, the requested file j is obtained from the BS for caching.

2. A cache optimization method based on file freshness and content recommendation according to claim 1, characterized in that: If there are multiple files j that meet the information age threshold requirement among the users within the D2D communication range of user i, then the file j with the smallest information age is selected for caching.

3. The cache optimization method based on file freshness and content recommendation according to claim 1, characterized in that: The recommendation algorithm includes: Calculate the user interest matrix I based on the file topic according to the user's historical request records U×F , and then calculate the cosine similarity between users according to the user interest matrix, find the user k with the highest interest similarity to user i among the users within the D2D communication range of user i, find at most N files that user k is most interested in and meet the information age threshold requirements in the files cached by user k and recommend them to user i. If user k does not have files that meet the requirements, find the user with the second highest interest similarity to user i, find at most N files that the second highest user is most interested in and meet the information age threshold requirements in the files cached by the second highest user and recommend them to user i, and so on.

4. A cache optimization method based on file freshness and content recommendation according to claim 3, characterized in that: Calculate the user interest matrix I based on the file topic according to the user's historical request records U×F The methods include: Assume that all the documents in the document library cover a total of Y topics, using the set Ω = {ω1,ω2,...ω k ,...ω Y } indicates that the subject ω k The attribute function between and file f is defined as: use Represents user u's opinion on the document topic ω k The preference of is formulated using the concept of mutual information and is expressed as: Among them, Z(ω k ) represents the subject ω k The collection of all files of U represents the historical request records of user u, P(Z(ω k )|Φ U ) represents the theme ω k The probability of appearing in the historical request records of user u, and P(Z(ω k )) represents the theme ω k The probability of appearing in the entire file library, therefore, user u’s interest in file f is defined as: Among them, user u’s interest in file f is η u,f ∈[-1,1],η u,f The closer the value is to -1, the less interested the user is, and the closer it is to 1, the more interested the user is. Finally, according to user u's interest in file f, u,f , get the user interest matrix I based on the file topic U×F .

5. A cache optimization method based on file freshness and content recommendation according to any one of claims 1 to 4, characterized in that: The information age of the file at the BS is set to 0, and the information age is gradually increased as the file spreads in the network, using δ ij (t) represents the information age of file j cached by user i in time slot t, and it is assumed that all requested content is delivered within one time slot. If file j cached at user i is not updated, then when the time slot increases by 1, the age of file j cached at user i will increase by 1. When user i caches a new version of file j, if file j is obtained from the BS, the information age of file j cached at user i will drop to 1. If file j is obtained from user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j cached at user k. If file j is recommended by user k within the D2D communication range, the information age of file j cached at user i will drop to the information age of file j recommended by user k. The information age function of file j cached by user i is: Let binary array and represents the source of cached file j at user i in time slot t. If the file j requested by user i is obtained from a user within the D2D communication range, then If the file j requested by user i is obtained from BS, then If the file j requested by user i is recommended by a user within the D2D communication range, then 6. A cache optimization method based on file freshness and content recommendation according to claim 5, characterized in that: Using binary variable a ij (t) is used to record whether the request for cached file j at user i in time slot t is served. If it is successfully served, then a ij (t)=1, otherwise, a ij (t) = 0; Using binary variable b ij (t) to record whether user i has cached file j in time slot t. If it has, then b ij (t) = 1, otherwise, b ij (t) = 0, when the cache file j at user i is unloaded at time slot t, b ij (t) = 0, the total size of the user's cached files cannot exceed the user's maximum cache capacity M, then: in, in, 7. A cache optimization method based on file freshness and content recommendation according to claim 6, characterized in that: Assume that the user set within the D2D communication range of user i is represented by D i To express, the minimum age of file j cached by user i among users within the D2D communication range at time slot t is The minimum age of file j obtained by user i through recommendation within the D2D communication range at time slot t is According to formula (4), we can get: Assume that the age of the file information obtained by the user cannot exceed an information age threshold δ th ,Right now: This ensures that the files obtained by users are not outdated and meet the information age threshold requirements.

8. The cache optimization method based on file freshness and content recommendation according to claim 7, characterized in that: Assume that the binary variable c ij (t) indicates whether there is a file j that meets the cache request in the D2D communication range of user i in time slot t. If yes, c ij (t)=1, otherwise c ij (t) = 0, c ij (t) is expressed as: in, When user i’s request for cached file j is served in time slot t, i.e., a ij (t) = 1, then the file can be obtained from users or BSs within the D2D communication range or by accepting recommended files, so: in, 9. The cache optimization method based on file freshness and content recommendation according to claim 1, characterized in that: The method also includes: when user i receives file j, if there is an old version of file j in the cache of user i, the old version of file j is directly replaced with the new version of file j; if the cache space of user i is sufficient, when caching the new file j, it is directly cached; if the cache space of user i is full, a content uninstallation operation is performed, following the principle of uninstalling outdated old files first.

10. The cache optimization method based on file freshness and content recommendation according to claim 8, characterized in that: The method further includes analyzing the system energy consumption: assuming that the BS and the user each use a fixed power P B and P D To transfer files, the user recommends that the transfer power of the file be P R , assuming that large-scale fading is considered in the communication process, SNR represents the signal-to-noise ratio, then the energy consumption of file j transmission is expressed as: Among them, P x is the transmission power, x j is the size of file j, B is the system bandwidth, SNR = P x ψ -α / σ 2 ,ψ represents the communication distance, α is the path loss index, σ 2 is the power average of Gaussian white noise; Assume that the energy consumption from the BS, from the users within the D2D communication range, and the files transmitted through the recommendation are E B ,E D and E R , then the energy consumption of the system in a single time slot is expressed as: Among them, 1≤t≤T.

Images