CN108668287A - A kind of active cache method based on user content popularity and movement rule - Google Patents
A kind of active cache method based on user content popularity and movement rule Download PDFInfo
- Publication number
- CN108668287A CN108668287A CN201810354251.8A CN201810354251A CN108668287A CN 108668287 A CN108668287 A CN 108668287A CN 201810354251 A CN201810354251 A CN 201810354251A CN 108668287 A CN108668287 A CN 108668287A
- Authority
- CN
- China
- Prior art keywords
- user
- base station
- content
- probability
- migration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/18—Network planning tools
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W16/00—Network planning, e.g. coverage or traffic planning tools; Network deployment, e.g. resource partitioning or cells structures
- H04W16/22—Traffic simulation tools or models
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/0289—Congestion control
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W28/00—Network traffic management; Network resource management
- H04W28/02—Traffic management, e.g. flow control or congestion control
- H04W28/10—Flow control between communication endpoints
- H04W28/14—Flow control between communication endpoints using intermediate storage
Abstract
The active cache method based on user content popularity and movement rule that the invention discloses a kind of, it is excavated by introducing social relationships, division to the progress community of the user in small cell network, according to zipf law, analyze the content popularit distribution of each community, and by Mining Algorithms of Frequent Patterns, the movement rule of user is excavated, the probability distribution that user's subsequent time period accesses each base station is obtained according to regular confidence level.The probability distribution of each base station is distributed and accessed according to the content requests of user, and under the limited buffer memory capacity in base station, cache hit rate is maximized by optimizing cache policy.Active cache method using the present invention can reduce backhaul load, reduce user's time delay, while this method has resolution characteristic, algorithm complexity low.
Description
Technical field
The invention belongs to field of communication technology, more particularly to a kind of active based on user content popularity and movement rule
Caching method.
Background technology
The rapid development of mechanics of communication is brought deeply while having pushed the development of society for the life style of people
Influence.Especially smart mobile phone is universal, not only enriches the experience of mobile subscriber, also brings a large amount of novel radio services,
Lead to the explosive growth of mobile flow.To meet unprecedented huge flow demand, a kind of promising approach is exactly chalcid fly
Data traffic on macro wireless network can be unloaded on Microcell by the deployment of nest network, small base station, when in conjunction with such as macrocellular
When with other Radio Access Network technologies such as Wi-Fi unloadings, small cell base station can bring more preferably moving body for terminal user
It tests and wireless coverage, by disposing additional network node within the scope of LAN, or shortens the distance between network and user,
Small small cell network can largely effectively improve the spatial multiplex ratio, network coverage, storage capacity of network
With flow relieving capacity.
However, since the large scale deployment of small base station causes infrastructure construction expense to increase substantially, operator can not
For the backhaul link for the large capacity that small base station deployment is connect with core network.In traditional network, the content of user's request must
It must be obtained from server by backhaul link, but when network busy or when having a large amount of requests to pour in, Netowrk tape will be given
Carry out great flow load, demand of the user to Internet resources such as link bandwidth, memory space and processor processing capacity etc. is super
When having crossed the intrinsic capacity of network, network is caused greatly to bear, congestion will occur for small base station backhaul link, lead to user
The time delay for obtaining content greatly increases, once and congestion occurs for network, just will appear loss of data, delay increases, under handling capacity
The situations such as drop, can cause " congestion collapse " when serious, and network performance drastically declines, and QoS of customer is deteriorated, the experience of user
Quality will also become difficult to endure.Therefore, under this passive service mode, small base station network can not play it and increase network
The advantage of capacity.
Invention content
It is an object of the invention to be directed to cache invalidation problem, propose a kind of based on user content popularity and movement rule
Active cache method, this method clusters the user in network, according to the content stream of the neat each community of husband's distributional analysis
Row degree is distributed, the base station that will be accessed by Frequent Pattern Mining, prediction user, before user arrives at, by access base station
Integrally most popular content caching gets off user, to reduce backhaul load, reduces user's time delay；Overcome limited backhaul chain
Limitation of the appearance of a street amount to mobile communication system capacity, meets abundant type of service and user demand, effectively reduces network load,
Using active cache technology the demand of user is predicted using the contextual information of user, and obtain and deposit using predictability
Important resource is stored up, to ensure quality of service requirement, reduces the flow of backhaul link, and then the flow for reducing cellular system is negative
It carries.
In order to achieve the above object, present invention employs following technical schemes：One kind being based on user content popularity and shifting
The active cache method for moving rule, includes the following steps：
Step 1) carries out community's division to the user in small cell network, according to zipf law, each society after analysis divides
The content popularit in area is distributed；
Step 2), by space-time Mining Algorithms of Frequent Patterns, user's movement rule in small cell network is excavated,
The migration probability matrix that user's subsequent time period accesses each base station is established according to regular confidence level；
Step 3), the probability distribution that each base station is distributed and accessed according to the content requests of user, in the limited caching in base station
Under capacity, cache hit rate is maximized by optimizing cache policy.
The step 1) includes the following steps：
1.1) user characteristics vector is established according to the personal information of user, using cosine similarity computational methods, according to spy
Sign vector carrys out the similarity between calculate node, obtains the customer relationship figure of undirected weighting；
1.2) the customer relationship figure by the undirected weighting that step 1.1) obtains is converted into edge graph, is based on node similarity calculation
While while between similarity, to weighting side use bottom-up Hierarchical clustering methods, obtain the non-overlapping community about side
It divides, and then converts the non-overlapping community on side to and divided about an overlapping community；
1.3) community after the division obtained according to step 1.2), obtains the number of user's request content in each community,
And then determine the distribution of each community content popularity.
The step 1.3) includes the following steps：
(1) assume that a contents directory contains F content, these contents are ranked up by popularity descending, first
To be most popular, the last item is most unfashionable, and content popularit distribution follows zipf law；User asks f-th of content
Probability formulaIt indicating, for the wherein value range of δ between 0 to 1.0, which determines the peak of distribution,
Influence the inclined degree of distribution, referred to as tilt parameters；
(2) step 1.2) as a result, obtaining the user of cc ∈ { 1,2,3 } class to content library in being divided according to communities of users
{f1, f2..., fFIn content sequence vector sc=[sc1, sc2..., scF] and each content of request probability value；
(3) probability value that overlapping community asks each content is obtained according to step (2), community is calculated using following formula
Content popularit：
The step 2) includes the following steps：
2.1) processing is carried out to the motion track of user and generates higher-dimension blurring trajectorie sequence training set；
2.2) it to the higher-dimension blurring trajectorie sequence training set generated in step 2.1), is excavated using Mining Algorithms of Frequent Patterns
Go out the frequent migration series of user, migration rules are generated according to migration series；
2.3) migration rules that step 2.2) generates are based on, confidence level is set, establish the migration probability matrix of user.
The step 2) is as follows：
(1) candidate pattern of migration series is generated according to Mining Algorithms of Frequent Patterns；
(2) user's all mobile sequence respective base stations in the residence time of each base station and library in candidate pattern are calculated
The sum of the intersection of residence time；
(3) calculate candidate pattern in user rest on each base station time and；
(4) ratio of step (2) and step (3) result, as time support support_Temporal, space are calculated
Support support_Spacial is defined as sequenceSpace support counting in sequence library BS and sequence sum in library
The support of ratio, sequence entirety is expressed as formula：
Support (α)=k × support_Spacial+ (1-k) × support_Temporal；
(5) all frequent migration sequence patterns for meeting minimum support are excavated；
(6) it is directed to the frequent migration pattern excavated and obtained, the mobility rule of user is generated, constitutes movement rule collection, it is false
Surely there is user's frequent migration sequence patternK >=2, then can be according to this shifting
The mobility rule of the user of dynamic schema creation is as follows：
(＜ b1,t1,t1' ＞) → (＜ b2,t2,t2' ＞ ..., ＜ bk,tk,tk' ＞)
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞) → (＜ b3,t3,t3' ＞ ..., ＜ bk,tk,tk' ＞) ...
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞ ..., ＜ bk-1,tk-1,tk′-1＞) → (＜ bk,tk,tk' ＞)
The confidence calculations formula of movement rule is as follows：
(7) by migration rules come the migration probability matrix of structuring user's, migration probability by mobility rule confidence level
It determines, migration probability matrix is expressed as formula：
Wherein, M, N indicate number of users and base station number, P respectivelyijIndicate that user i moves to the probability of base station j, this is general
The calculation formula of rate value is as follows：
According to formulaMigration probability in matrix is normalized；
8) in forecast period, a current mobile sequence of user is given, is found out in movement rule concentration the most matched
Mobility rule, to predict the base station that the next timestamp of user will access.
The step 3) includes the following steps：
3.1) user obtained according to step 1) is subordinate to community and the migration probability matrix of user that step 2) obtains,
Establish the mathematical model of active cache；
3.2) according to mobility prediction as a result, popular according to the content of community corresponding to the user for being under the jurisdiction of each base station
Degree determines cache contents；
3.3) cache hit rate is defined, cache hit rate is then maximized according to mathematical model, obtains the master of total optimization
Dynamic cache policy.
The step 3.1) includes the following steps：
(1) there is N in small cell network, the memory capacity of a base station (i=1,2 ..., N), each small base station is S, it is assumed that is had
F file, file directory are C={ f1,f2,…,fF, the length of each file is L, for corresponding one of each base station to
Amount is expressed asbij∈ { 0,1 }, (i=1,2 ..., N), wherein bij∈ { 0,1 }, (j=1,2 ...,
F), bijIndicate that base station i has cached file f equal to 1i, equal to 0 expression does not cache, and the file that meet storage is total
Size no more than base station memory capacity S, i.e.,
(2) assume that current network has accessed M, (k=1,2 ..., M) a mobile subscriber, some user's k access base station i's is general
Rate distribution is expressed asWhereinIndicate to number the user for being k in the general of subsequent period access base station i
Rate, we can be obtained the value by the method used in abovementioned steps (two), which meets
(3) user k asks the probability distribution formula of F file in subsequent periodIt indicates,
WhereinIt indicates to number the probability that the user for being k accesses content j in subsequent period, which can be by abovementioned steps (two)
Content popularit setting method obtains, which meets
(4) probability that user k can be obtained to request content in base station i is expressed asAll users can obtain in base station i
The probability of request content is taken to be exactlySo, in entire small cell network, all users can respectively be subordinate to
The probability that respective request content is obtained in the base station of category is represented by
The step 3.2) includes the following steps：
(1) for whole network,Value reaches maximum, means that all users all
The sum of the probability of respective request content maximum is obtained in corresponding base station, i.e. user asks the maximum probability of hit, therefore is based on
Above formula defines cache hit rate, and hit rate is expressed asThe value value range
It is 0 to 1；
(2) optimization aim is set, since the storage capacity of small base station is limited, the content that cannot ask all users is all slow
It leaves and, therefore the final goal of cache policy is exactly to be reached most using the probability that limited spatial cache makes user ask hit
Greatly, the optimization aim indicated such as following formula is designed thus：
N, M and F therein indicate base station number, number of users and quantity of documents in network respectively,Indicate that user k connects
Enter the probability distribution of base station i,Indicate the probability distribution of user's k request contents j, bijIndicate that base station i has cached the general of content j
Rate is distributed；
(3) calculate will all users of access base station i understand the probability of request content jDistribution, use vijIt indicatesJust convert optimization problem to
(4) it is directed to each base station i,Under conditions of, it maximizesTo vij, (j=1,2 ..., F)
It is ranked up to obtain vi'={ vi(1),vi(2),…,vi(F), it willIn with v 'iThe corresponding element of preceding S element is set as 1,
His element is set as 0, allows in this wayObtain maximum value；
(5) method for using first two steps to each base station makes the cache hit probability of all base stations reach maximum, that
The cache hit rate of small cell network entirety has just reached maximum value.
The present invention proposes a kind of active cache method based on user content popularity and movement rule, by introducing society
Relation excavation analyzes the content of each community to the division of the progress community of the user in small cell network according to zipf law
Popularity is distributed, and by Mining Algorithms of Frequent Patterns, is excavated to the movement rule of user, is obtained according to regular confidence level
User's subsequent time period accesses the probability distribution of each base station.The probability point of each base station is distributed and accessed according to the content requests of user
Cloth maximizes cache hit rate under the limited buffer memory capacity in base station by optimizing cache policy.Active using the present invention is slow
Method is deposited, backhaul load can be reduced, reduces user's time delay, while this method has resolution characteristic, algorithm complexity low.
Compared with prior art, the present invention at least has the advantages that：The present invention is slow using active from bottom to top
Strategy is deposited, is excavated using social groups and mobility is predicted, consider the demand of each user in network in detail, analysis is each
The distribution of the location information of a user and the content of request, by specific category mobile subscriber and mobile application caused by flow
Amount is unloaded to small base station, and active cache mechanism is then used in small base station.The present invention is excavated by introducing social relationships, by net
User in network is divided into different social groups, obtains the content requests distribution situation of each group, pre- by introducing mobility
It surveys, obtains the movement law of all users in network, cache user most it is expected to obtain in the base station that user will reach in advance
Content, can directly acquire content from local after user reaches corresponding base station, to reduce the congestion of backhaul link, contracting
Short user's time delay.There is no simultaneously in view of the mobility of the social relationships and user of user, this hair for existing buffering scheme
The active cache method of bright use can more effectively excavate the true demand of user, and people's answers in the life that preferably reflects reality
Polygamy and Biodiversity Characteristics can improve the diversity of cache contents, can obtain higher cache hit rate；It is most to have algorithm
All it is integer programming problem, N-P Hard problems, and the present invention has resolution characteristic, is easy to solve.
Description of the drawings
Fig. 1 is the active cache method flow diagram based on user content popularity and movement rule；
Fig. 2 is communities of users splitting scheme frame diagram；
Fig. 3 is the flow chart that overlapping community mining is carried out to user；
Fig. 4 is the Frequent Pattern Mining flow chart using user's migration series.
Specific implementation mode
In order to make present disclosure, effect and advantage be more clearly understood, with reference to the accompanying drawings and examples to this
Invention is described in detail.
All users in network are all classified as the same community, the content popularit of network entirety by existing caching method
Neat husband's distribution, the demand without considering individual consumer are obeyed, therefore buffer efficiency is very low.And the present invention is to be based on user clustering
With ambulant caching method, personalized buffer service can be provided to the user.
The strategic process that the present invention is illustrated in Fig. 1 considers from what is cached, which is buffered in and how to cache three angles
Buffering scheme carries out community's division to user by weighting side cluster, analyzes the content popularit distribution of each community, pass through frequency
Numerous pattern mining algorithm excavates user in the migration rules of minizone, obtains the probability point that user's subsequent time period accesses each base station
Cloth establishes the mathematical model of caching in conjunction with preceding two-part result, each base station is distributed and accessed according to the content requests of user
Probability distribution obtains the probability i.e. cache hit rate that each user obtains request content from the spatial cache of base station, has in base station
Under the buffer memory capacity of limit, cache hit rate is improved by optimizing cache policy, reduces backhaul load, reduced user and obtain content
Time delay；Specifically, the present invention includes the following steps：
Step (1) communities of users splitting scheme and the community content popularity plan of establishment
1. referring to Fig. 2, communities of users, which divides, is generally divided into two big modules：User's similarity calculation module is gathered with based on side
The user community of class excavates module, specifically includes following steps：
1) referring to Fig. 3, the feature vector for first obtaining binary crelation figure and user between user is concentrated from user related data,
The customer relationship figure of the relationship strength between two users i.e. undirected weighting is weighed using the computational methods of cosine similarity
Weights；
2) referring to Fig. 3, in the clustering algorithm based on side structural similarity, user's binary crelation figure is regarded as one
Undirected weighted graph is indicated, wherein V is a non-empty set of node with a triple G=(V, E, L)；E be a triple (x,
Y, l), x, y ∈ V, l ∈ L, x ≠ y, for any two tuple (x, y, l) ∈ E, (x ', y ', l ') ∈ E, if x=x ', y=
Y ', then l ≠ l ', L are the set on side.Edge graph adjacency matrix is obtained from the adjacency matrix of customer relationship figure, based on a similarity
Calculate while while between similarity, side similarity is defined as formula：
Wherein liAnd ljIt is two sides, vi1、vi2And vj1、vj2It is side l respectivelyi、ljTwo endpoints, the similarity calculation letter
Several value ranges is [0,1], meets surjection principle；
3) referring to Fig. 3, each edge in weighting edge graph is numbered, and preserve two node numbers of each edge, using such as
Lower formula calculates the distance between any two cluster：
The dimension of distance matrix and the number of cluster are all reduced with the merging of cluster, when final number of clusters is equal to setting value,
Output cluster converts non-overlapping side cluster to overlapping a little and clusters according to the two of the side preserved endpoints.
2. the community content popularity plan of establishment specifically includes following steps：
1) assume that a contents directory contains F content, these contents are ranked up by popularity descending, and first is
Most popular, the last item is most unfashionable, and content popularit distribution follows zipf law；User's f-th of content of request
Probability formulaIt indicates, for the wherein value range of δ between 0 to 1.0, which determines the peak of distribution, shadow
Ring the inclined degree of distribution, referred to as tilt parameters；
2) step 3) as a result, obtaining the user of cc ∈ { 1,2,3 } class to content library in being divided according to communities of users
{f1, f2..., fFIn content sequence vector sc=[sc1, sc2..., scF] and each content of request probability value；
3) probability value that overlapping community asks each content is obtained according to step 2), is calculated in community using following formula
Hold popularity：
Movement rule method for digging of the step (2) based on Frequent Sequential Patterns
For reduce user between request content at a distance from, reduce the load of backhaul link, the position of caching determined, by drawing
Enter the prediction to user mobility, by content caching in the base station that user will access, the periodic feature of user's movement shows
The movement of user has followed certain specific Move Mode, and user's motion track is mobile sequence of the user about base station, can be with
With set BSs={ BT1,BT2,…,BTsIndicate, BTi={ bsi,tsi,tei, wherein bsi、tsiAnd teiBase station is indicated respectively
Number, into the base station time and leave time of the base station, user's movement rule mining process includes the following steps：
1) referring to Fig. 4, the candidate pattern of migration series is generated according to Mining Algorithms of Frequent Patterns；
2) referring to Fig. 4, user's all mobile sequences pair in the residence time of each base station and library in candidate pattern are calculated
Answer the sum of the intersection of the residence time of base station；
3) referring to Fig. 4, calculate the user in candidate pattern rest on each base station time and；
4) ratio of step 2) and step 3) result, as time support support_Temporal, space branch are calculated
Degree of holding support_Spacial is defined as sequenceThe ratio of sequence sum in space support counting and library in sequence library BS
Value, the support of sequence entirety are expressed as formula：
Support (α)=k × support_Spacial+ (1-k) × support_Temporal；
5) referring to Fig. 4, all frequent migration sequence patterns for meeting minimum support are excavated；
6) it is directed to the frequent migration pattern excavated and obtained, the mobility rule of user is generated, constitutes movement rule collection, it is assumed that
There is user's frequent migration sequence patternK >=2, then can be according to this movement
The mobility rule of the user of schema creation is as follows：
(＜ b1,t1,t1' ＞) → (＜ b2,t2,t2' ＞ ..., ＜ bk,tk,tk' ＞)
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞) → (＜ b3,t3,t3' ＞ ..., ＜ bk,tk,tk' ＞) ...
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞ ..., ＜ bk-1,tk-1,tk′-1＞) → (＜ bk,tk,tk' ＞)
The confidence calculations formula of movement rule is as follows：
7) by migration rules come the migration probability matrix of structuring user's, migration probability is determined by the confidence level of mobility rule
Fixed, migration probability matrix is expressed as formula：
Wherein, M, N indicate number of users and base station number, P respectivelyijIndicate that user i moves to the probability of base station j, this is general
The calculation formula of rate value is as follows：
According to formulaMigration probability in matrix is normalized；
8) in forecast period, a current mobile sequence of user is given, is found out in movement rule concentration the most matched
Mobility rule, to predict the base station that the next timestamp of user will access.
Active cache method of the step (3) based on user content popularity and movement rule
1. the structure of mathematical model includes the following steps：
1) there is N in small cell network, the memory capacity of a base station (i=1,2 ..., N), each small base station is S, it is assumed that has F
A file, file directory are C={ f1,f2,…,fF, the length of each file is L, and a vector is corresponded to for each base station
It is expressed asbij∈ { 0,1 }, (i=1,2 ..., N), wherein bij∈ { 0,1 }, (j=1,2 ..., F),
bijIndicate that base station i has cached file f equal to 1i, equal to 0 expression does not cache, and to meet the file total size of storage
No more than the memory capacity S of base station, i.e.,
2) assume that current network has accessed M, (k=1,2 ..., M) a mobile subscriber, some user's k access base station i's is general
Rate distribution is expressed asWhereinIndicate to number the user for being k in the general of subsequent period access base station i
Rate, we can be obtained the value by the method used in abovementioned steps (two), which meets
3) user k asks the probability distribution formula of F file in subsequent periodIt indicates,
WhereinIt indicates to number the probability that the user for being k accesses content j in subsequent period, which can be by abovementioned steps (two)
Content popularit setting method obtains, which meets
4) probability that user k can be obtained to request content in base station i is expressed asAll users can obtain in base station i
The probability of request content is taken to be exactlySo, in entire small cell network, all users can respectively be subordinate to
The probability that respective request content is obtained in the base station of category is represented by
2. the optimization process of cache hit rate includes the following steps：
For reduction network load as much as possible, most of flow is unloaded in small base station, end-user access is worked as
When base-station node, when node B cache data to be accessed if be called hit, conversely, if do not hit, if do not had
There is hit just to need to obtain into core net, since the process of access evidence and user's access are synchronous progress, if do not ordered
In, delay will be generated because being asked into core net.Therefore, cache hit rate be influence one of time delay it is very important because
Element, hit rate is higher, and for user's time delay with regard to smaller, user experience is better, and the load of backhaul link is also just smaller.
1) for whole network,Value reaches maximum, means that all users all
The sum of the probability of respective request content maximum is obtained in corresponding base station, i.e. user asks the maximum probability of hit, therefore is based on
Above formula defines cache hit rate, and hit rate is expressed asThe value value range
It is 0 to 1.
2) optimization aim is set, since the storage capacity of small base station is limited, the content that cannot ask all users is all slow
It leaves and, therefore the final goal of cache policy is exactly to be reached most using the probability that limited spatial cache makes user ask hit
Greatly, the optimization aim indicated such as following formula is designed thus：
N, M and F therein indicate base station number, number of users and quantity of documents in network respectively,Indicate that user k connects
Enter the probability distribution of base station i,Indicate the probability distribution of user's k request contents j, bijIndicate that base station i has cached the general of content j
Rate is distributed.
3) calculate will all users of access base station i understand the probability of request content jDistribution, use vijIt indicatesJust convert optimization problem to
4) it is directed to each base station i,Under conditions of, it maximizesTo vij, (j=1,2 ..., F) into
Row sequence obtains vi'={ vi(1),vi(2),…,vi(F), it willIn with v 'iThe corresponding element of preceding S element is set as 1, other
Element is set as 0, allows in this wayObtain maximum value；
5) method for using first two steps to each base station makes the cache hit probability of all base stations reach maximum, that
The cache hit rate of small cell network entirety has just reached maximum value.
Claims (8)
1. a kind of active cache method based on user content popularity and movement rule, it is characterised in that：Include the following steps：
Step 1) carries out community's division to the user in small cell network, according to zipf law, analysis each community after dividing
Content popularit is distributed；
Step 2), by space-time Mining Algorithms of Frequent Patterns, user's movement rule in small cell network is excavated, according to
Regular confidence level establishes the migration probability matrix that user's subsequent time period accesses each base station；
Step 3), the probability distribution that each base station is distributed and accessed according to the content requests of user, in the limited buffer memory capacity in base station
Under, maximize cache hit rate by optimizing cache policy.
2. a kind of active cache method based on user content popularity and movement rule according to claim 1, special
Sign is：The step 1) includes the following steps：
1.1) user characteristics vector is established according to the personal information of user, using cosine similarity computational methods, according to feature to
Amount carrys out the similarity between calculate node, obtains the customer relationship figure of undirected weighting；
1.2) the customer relationship figure by the undirected weighting that step 1.1) obtains is converted into edge graph, based on node similarity calculation side with
Similarity between side uses bottom-up Hierarchical clustering methods to weighting side, obtains dividing about the non-overlapping community on side,
And then it converts the non-overlapping community on side to and is divided about an overlapping community；
1.3) community after the division obtained according to step 1.2), obtains the number of user's request content in each community, in turn
Determine the distribution of each community content popularity.
3. a kind of active cache method based on user content popularity and movement rule according to claim 2, special
Sign is：The step 1.3) includes the following steps：
(1) assume that a contents directory contains F content, these contents are ranked up by popularity descending, and first is most
Popular, the last item is most unfashionable, and content popularit distribution follows zipf law；User asks the general of f-th content
Rate formulaIt indicates, for the wherein value range of δ between 0 to 1.0, which determines the peak of distribution, influences point
The inclined degree of cloth, referred to as tilt parameters；
(2) step 1.2) as a result, obtaining the user of cc ∈ { 1,2,3 } class to content library { f in being divided according to communities of users1,
f2..., fFIn content sequence vector sc=[sc1, sc2..., scF] and each content of request probability value；
(3) probability value that overlapping community asks each content is obtained according to step (2), community content is calculated using following formula
Popularity：
4. a kind of active cache method based on user content popularity and movement rule according to claim 1, special
Sign is：The step 2) includes the following steps：
2.1) processing is carried out to the motion track of user and generates higher-dimension blurring trajectorie sequence training set；
2.2) to the higher-dimension blurring trajectorie sequence training set generated in step 2.1), use is excavated using Mining Algorithms of Frequent Patterns
The frequent migration series in family generate migration rules according to migration series；
2.3) migration rules that step 2.2) generates are based on, confidence level is set, establish the migration probability matrix of user.
5. a kind of active cache method based on user content popularity and movement rule according to claim 4, special
Sign is：The step 2) is as follows：
(1) candidate pattern of migration series is generated according to Mining Algorithms of Frequent Patterns；
(2) stop of user's all mobile sequence respective base stations in the residence time of each base station and library in candidate pattern is calculated
The sum of the intersection of time；
(3) calculate candidate pattern in user rest on each base station time and；
(4) ratio of step (2) and step (3) result is calculated, as time support support_Temporal, space is supported
Degree support_Spacial is defined as sequenceThe ratio of sequence sum in space support counting and library in sequence library BS
Value, the support of sequence entirety are expressed as formula：
Support (α)=k × support_Spacial+ (1-k) × support_Temporal；
(5) all frequent migration sequence patterns for meeting minimum support are excavated；
(6) it is directed to the frequent migration pattern excavated and obtained, the mobility rule of user is generated, constitutes movement rule collection, it is assumed that have
User's frequent migration sequence patternIt so can be according to this movement
The mobility rule of the user of schema creation is as follows：
(＜ b1,t1,t1' ＞) → (＜ b2,t2,t2' ＞ ..., ＜ bk,tk,tk' ＞)
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞) → (＜ b3,t3,t3' ＞ ..., ＜ bk,tk,tk' ＞) ...
(＜ b1,t1,t1' ＞, ＜ b2,t2,t2' ＞ ..., ＜ bk-1,tk-1,tk′-1＞) → (＜ bk,tk,tk' ＞)
The confidence calculations formula of movement rule is as follows：
(7) by migration rules come the migration probability matrix of structuring user's, migration probability is determined by the confidence level of mobility rule,
Migration probability matrix is expressed as formula：
Wherein, M, N indicate number of users and base station number, P respectivelyijIndicate that user i moves to the probability of base station j, the probability value
Calculation formula it is as follows：
According to formulaMigration probability in matrix is normalized；
8) in forecast period, a current mobile sequence of user is given, movement the most matched is found out in movement rule concentration
Property rule, to predict the base station that the next timestamp of user will access.
6. a kind of active cache method based on user content popularity and movement rule according to claim 1, special
Sign is：The step 3) includes the following steps：
3.1) user obtained according to step 1) is subordinate to community and the migration probability matrix of user that step 2) obtains, establishes
The mathematical model of active cache；
3.2) according to mobility prediction as a result, true according to the content popularit of community corresponding to the user for being under the jurisdiction of each base station
Determine cache contents；
3.3) cache hit rate is defined, cache hit rate is then maximized according to mathematical model, the active for obtaining total optimization is slow
Deposit strategy.
7. a kind of active cache method based on user content popularity and movement rule according to claim 1, special
Sign is：The step 3.1) includes the following steps：
(1) there is N in small cell network, the memory capacity of a base station (i=1,2 ..., N), each small base station is S, it is assumed that there are F
File, file directory are C={ f1,f2,…,fF, the length of each file is L, and a vector table is corresponded to for each base station
It is shown asWherein bij∈ { 0,1 }, (j=1,2 ..., F), bijDeng
Indicate that base station i has cached file f in 1i, equal to 0 expression does not cache, and the file total size that meet storage cannot
More than the memory capacity S of base station, i.e.,
(2) assume that current network has accessed M, (k=1,2 ..., M) a mobile subscriber, the probability point of some user's k access base station i
Cloth is expressed asWhereinIndicate number be k user subsequent period access base station i probability,
We can be obtained the value by the method used in abovementioned steps (two), which meets
(3) user k asks the probability distribution formula of F file in subsequent periodIt indicates, whereinIt indicates to number the probability that the user for being k accesses content j in subsequent period, which can pass through the content in abovementioned steps (two)
Popularity setting method obtains, which meets
(4) probability that user k can be obtained to request content in base station i is expressed asAll users can obtain in base station i and ask
The probability of content is asked to be exactlySo, in entire small cell network, all users can respectively be subordinate to
The probability that respective request content is obtained in base station is represented by
8. a kind of active cache method based on user content popularity and movement rule according to claim 1, special
Sign is：The step 3.2) includes the following steps：
(1) for whole network,Value reaches maximum, means that all users all corresponding
Base station in obtain the sum of the probability of respective request content maximum, i.e., user asks the maximum probability of hit, therefore is based on above formula
Cache hit rate is defined, hit rate is expressed asThe value value range is 0 to arrive
1；
(2) optimization aim is set, since the storage capacity of small base station is limited, under the content that cannot ask all users all caches
Come, therefore the final goal of cache policy is exactly to make user that the probability of hit be asked to reach maximum using limited spatial cache,
The design optimization aim that such as following formula indicates thus：
N, M and F therein indicate base station number, number of users and quantity of documents in network respectively,Indicate that user k accesses base
It stands the probability distribution of i,Indicate the probability distribution of user's k request contents j, bijIndicate that base station i has cached the probability point of content j
Cloth；
(3) calculate will all users of access base station i understand the probability of request content jDistribution, use vijIt indicatesJust convert optimization problem to
(4) it is directed to each base station i,Under conditions of, it maximizesTo vij, (j=1,2 ..., F) it carries out
Sequence obtains vi'={ vi(1),vi(2),…,vi(F), it willIn with v 'iThe corresponding element of preceding S element is set as 1, other yuan
Element is set as 0, allows in this wayObtain maximum value；
(5) method for using first two steps to each base station makes the cache hit probability of all base stations reach maximum, then small
The cache hit rate of cellular network entirety has just reached maximum value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810354251.8A CN108668287B (en) | 2018-04-19 | 2018-04-19 | Active caching method based on user content popularity and mobile rule |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810354251.8A CN108668287B (en) | 2018-04-19 | 2018-04-19 | Active caching method based on user content popularity and mobile rule |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108668287A true CN108668287A (en) | 2018-10-16 |
| CN108668287B CN108668287B (en) | 2021-01-19 |
Family
ID=63780794
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810354251.8A Active CN108668287B (en) | 2018-04-19 | 2018-04-19 | Active caching method based on user content popularity and mobile rule |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108668287B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109617962A (en) * | 2018-12-11 | 2019-04-12 | 电子科技大学 | A kind of car networking mist node content caching method based on the content degree of association |
| CN109714790A (en) * | 2019-01-23 | 2019-05-03 | 南京邮电大学 | A kind of edge cooperation caching optimization method based on user mobility prediction |
| CN109831801A (en) * | 2019-01-04 | 2019-05-31 | 东南大学 | The node B cache algorithm of user's behavior prediction based on deep learning neural network |
| CN112468597A (en) * | 2020-12-11 | 2021-03-09 | 深圳市知小兵科技有限公司 | Data center resource allocation method and device based on artificial intelligence |
| CN112911614A (en) * | 2021-01-12 | 2021-06-04 | 重庆邮电大学 | Cooperative coding caching method based on dynamic request D2D network |
| CN114866610A (en) * | 2022-05-23 | 2022-08-05 | 电子科技大学 | Satellite-ground network caching method based on CCN |
| CN115604130A (en) * | 2022-12-01 | 2023-01-13 | 中南大学（Cn） | APP popularity prediction model construction method, prediction method, device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105049326A (en) * | 2015-06-19 | 2015-11-11 | 清华大学深圳研究生院 | Social content caching method in edge network area |
| CN107105043A (en) * | 2017-04-28 | 2017-08-29 | 西安交通大学 | A kind of content center network caching method based on software defined network |
| CN107659946A (en) * | 2017-09-19 | 2018-02-02 | 北京工业大学 | A kind of mobile communications network model building method based on edge cache |
| CN107733949A (en) * | 2016-08-11 | 2018-02-23 | 北京邮电大学 | A kind of wireless access network caching method and system |
-
2018
- 2018-04-19 CN CN201810354251.8A patent/CN108668287B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105049326A (en) * | 2015-06-19 | 2015-11-11 | 清华大学深圳研究生院 | Social content caching method in edge network area |
| CN107733949A (en) * | 2016-08-11 | 2018-02-23 | 北京邮电大学 | A kind of wireless access network caching method and system |
| CN107105043A (en) * | 2017-04-28 | 2017-08-29 | 西安交通大学 | A kind of content center network caching method based on software defined network |
| CN107659946A (en) * | 2017-09-19 | 2018-02-02 | 北京工业大学 | A kind of mobile communications network model building method based on edge cache |
Non-Patent Citations (3)
| Title |
|---|
| GONGYE REN: "A Distributed User Association and Resource Allocation Method in Cache-Enabled Small Cell Networks", 《COMMUNICATIONS THEORIES & SYSTEMS》 * |
| 唐红: "基于内容流行度的移动CCN缓存策略研究", 《重庆邮电大学学报》 * |
| 曲桦: "内容中心网络中一种改进型缓存机制", 《计算机工程》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109617962A (en) * | 2018-12-11 | 2019-04-12 | 电子科技大学 | A kind of car networking mist node content caching method based on the content degree of association |
| CN109831801A (en) * | 2019-01-04 | 2019-05-31 | 东南大学 | The node B cache algorithm of user's behavior prediction based on deep learning neural network |
| CN109831801B (en) * | 2019-01-04 | 2021-09-28 | 东南大学 | Base station caching method for user behavior prediction based on deep learning neural network |
| CN109714790A (en) * | 2019-01-23 | 2019-05-03 | 南京邮电大学 | A kind of edge cooperation caching optimization method based on user mobility prediction |
| CN112468597A (en) * | 2020-12-11 | 2021-03-09 | 深圳市知小兵科技有限公司 | Data center resource allocation method and device based on artificial intelligence |
| CN112468597B (en) * | 2020-12-11 | 2021-05-28 | 深圳市知小兵科技有限公司 | Data center resource allocation method and device based on artificial intelligence |
| CN112911614A (en) * | 2021-01-12 | 2021-06-04 | 重庆邮电大学 | Cooperative coding caching method based on dynamic request D2D network |
| CN112911614B (en) * | 2021-01-12 | 2022-05-03 | 重庆邮电大学 | Cooperative coding caching method based on dynamic request D2D network |
| CN114866610A (en) * | 2022-05-23 | 2022-08-05 | 电子科技大学 | Satellite-ground network caching method based on CCN |
| CN115604130A (en) * | 2022-12-01 | 2023-01-13 | 中南大学（Cn） | APP popularity prediction model construction method, prediction method, device and storage medium |
| CN115604130B (en) * | 2022-12-01 | 2023-03-14 | 中南大学 | APP popularity prediction model construction method, prediction method, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108668287B (en) | 2021-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108668287A (en) | A kind of active cache method based on user content popularity and movement rule | |
| Yu et al. | An intelligent game-based offloading scheme for maximizing benefits of IoT-edge-cloud ecosystems | |
| CN110730471B (en) | Mobile edge caching method based on regional user interest matching | |
| CN112020103B (en) | Content cache deployment method in mobile edge cloud | |
| Yan et al. | PECS: Towards personalized edge caching for future service-centric networks | |
| CN110213627A (en) | Flow medium buffer distributor and its working method based on multiple cell user mobility | |
| CN107592656B (en) | Caching method based on base station clustering | |
| Wang et al. | A flexible and generalized framework for access network selection in heterogeneous wireless networks | |
| Yan et al. | Distributed edge caching with content recommendation in fog-rans via deep reinforcement learning | |
| Tan et al. | Femtocaching in wireless video networks: Distributed framework based on exact potential game | |
| Tang et al. | EICache: A learning-based intelligent caching strategy in mobile edge computing | |
| Ye et al. | Processing capability and QoE driven optimized computation offloading scheme in vehicular fog based F-RAN | |
| Li et al. | Edge caching for D2D enabled hierarchical wireless networks with deep reinforcement learning | |
| CN113993168A (en) | Multi-agent reinforcement learning-based cooperative caching method in fog wireless access network | |
| CN109673015B (en) | Tracking area planning method based on spectral clustering | |
| CN112004265B (en) | Social network resource allocation method based on SRM algorithm | |
| CN112911614A (en) | Cooperative coding caching method based on dynamic request D2D network | |
| Sultan et al. | Where big data meets 5G? | |
| CN111885113A (en) | Self-adaptive selection and resource allocation method for anchor nodes in social network | |
| Zhang et al. | Learning-to-rank based strategy for caching in wireless small cell networks | |
| CN114245422A (en) | Edge active caching method based on intelligent sharing in cluster | |
| Chakraborty et al. | R2-d2d: A novel deep learning based content-caching framework for d2d networks | |
| Swedha et al. | LSTM network for hotspot prediction in traffic density of cellular network | |
| Han et al. | A novel handover detection model via frequent trajectory patterns mining | |
| Chen et al. | Hierarchical cooperative caching in fog radio access networks: a brain storm optimization approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |