Mobile video service resource supply method based on demand prediction

Abstract

The invention relates to a demand prediction-based mobile video service resource supply method. The method comprises the following steps of S1, dividing users into uninfected persons, infected persons, infected persons and immunized persons; s2, according to the user preference and the user social relation, measuring the probability that the user obtains the video; s3, establishing a video spreading model based on the infectious disease model, and predicting the spreading state of the video; s4, dividing the time period of starting spreading and ending spreading of the video into a plurality of spreading rounds with the same time length, and predicting the basic regeneration number of the video spreading model in each spreading round based on the probability of acquiring the video by the user; s5, making a mobile video service resource supply strategy; and S6, executing a supply strategy based on the numerical value of the basic regeneration number in the current spreading round. According to the method, the spreading model of the video in the social network is established by investigating the user preference and the social relationship between the users, so that the video resource demand can be accurately predicted, the video resource can be effectively configured, and the supply and demand can be balanced.

Description

Technical Field

[0001] The present invention relates to the technical field of video resource allocation, and in particular to a mobile video service resource supply method based on demand prediction. Background Art

[0002] The rapid development of wireless communication technologies represented by 5G provides mobile users with sufficient air interfaces and network bandwidth, thus enabling mobile users to "ubiquitously" access Internet application services and resources. Video services provide users with rich visual content, greatly enhancing the user experience quality, thereby attracting a large number of users, generating huge video traffic, and bringing a heavy traffic burden to the backbone network; convenient "ubiquitous" access not only increases the frequency of user use, but also further promotes the rapid growth of the number of users using video services, thereby further exacerbating the traffic burden of the backbone network. In addition to the growth of video traffic brought by the huge user base, the continuous improvement of video quality (such as high-definition video, virtual reality and augmented reality technologies) has also become an important factor in the growth of video traffic. The application of peer-to-peer (P2P) network technology provides a feasible solution for the large-scale deployment of video services. By using the remaining bandwidth of the client as the service upload bandwidth of the video system, the service scale of the video system is increased and the scalability of the video system is improved. The dissemination of video resources can expand the distribution of video resources and increase the supply scale of upload bandwidth, becoming the core of improving the service scale of the video system.

[0003] In order to improve the dissemination performance of video resources, video systems have been deeply integrated with social networks, such as Tencent Video and Douyin. Different from the requests and pushes between users and video servers in traditional video systems, social networks allow users to use their social connections to implement video resource requests and pushes, thereby improving the scale and speed of video resource dissemination. For example, the views of popular videos on Douyin can reach 30 million within 24 hours. In addition, the "small world phenomenon" of social networks allows users to share video resources using shorter social distances, and higher social communication distances further improve the scale and speed of video resource dissemination. In the process of video resource dissemination, the existing upload bandwidth supply scale of the video system should meet the upload bandwidth scale required for video dissemination, so as to achieve a balance between supply and demand. However, the rapid increase in video dissemination speed greatly increases the risk of sudden intensive requests, seriously destroys the balance of supply and demand of video system resources, increases the startup delay of video request users, reduces the quality of experience (QoE), and limits the scope and scale of video dissemination. Before sudden intensive requests occur, the demand for video resources should be predicted and the distribution of video resources should be adjusted in advance to reduce the degree to which sudden intensive requests damage the balance of video resource supply and ensure the quality of user experience.

[0004] Demand forecasting can reflect the dynamic development of user demand for video resources. The distribution characteristics of video resources and user preferences determine the scale and distribution speed of video resource demand. Through video resource demand forecasting, the supply capacity of upload bandwidth can be improved to predict the demand quantity in advance. However, existing video resource demand forecasting methods usually only consider the prediction of traffic demand, and do not consider the factors that affect the prediction results when the traffic demand changes. It is difficult to realize resource configuration and adjustment, and it is impossible to ensure the accuracy of video demand forecasting. Summary of the invention

[0005] In view of the above problems, the present invention proposes a mobile video service resource supply method based on demand prediction, establishes a video dissemination model based on epidemics, accurately describes the process of video dissemination in social networks, and uses the basic regeneration number to predict the upload bandwidth scale required for video resources, so as to formulate a video resource allocation strategy to balance the supply and demand of video resources in bursty intensive requests.

[0006] The specific technical solutions of the present invention are as follows:

[0007] The method for supplying mobile video service resources based on demand prediction includes the following steps:

[0008] S1, dividing users into uninfected people who have not obtained video information, susceptible people who have obtained video information, infected people who are watching the video, and immune people who have completed watching the video;

[0009] S2, measure the probability of users obtaining videos based on user preferences and user social relationships;

[0010] S3. Establish a video diffusion model based on the infectious disease model to predict the diffusion status of the video in the social network composed of users;

[0011] S4, dividing the time period from the start of video dissemination to the end of dissemination into multiple dissemination rounds of equal length, predicting the number of newly infected people in each dissemination round based on the probability of users acquiring the video, and predicting the basic reproduction number of the video dissemination model in each dissemination round;

[0012] S5. Formulate a mobile video service resource supply strategy, and judge whether the stored video data needs to provide additional video copies according to the relationship between the basic reproduction number in the current dissemination round and the amount of video data supply service that can be provided to other users by the stored video data;

[0013] S6. Execute a mobile video service resource supply strategy based on the value of the basic regeneration number in the current broadcast round.

[0014] Specifically, in S2, the probability of a user in the social network obtaining video data is calculated by formula (1):

[0015]

[0016] in, is the probability that the user does not obtain video data with preference, The probability that the user does not obtain video data through social push.

[0017] Specifically, in S3, the video diffusion model is defined as formula (2):

[0018]

[0019] Among them, N(S) is the number of susceptible people in the social network; N(I) is the number of infected people in the social network; N(R) is the number of immune people in the social network; ε is the rate of new infected people; λ is the rate of new immune people; μ is the rate of new susceptible people.

[0020] Specifically, in S5, when the basic reproduction number in the current dissemination round is greater than or equal to 1 and greater than the number of stored video data that can provide video data supply services to other users, storing the video data requires additional video copies.

[0021] Specifically, the basic reproduction number R0(r i ) is greater than or equal to 1 and greater than the number N of stored video data that can provide video data supply services to other users s When the social network needs an additional copy of the video (R0(r i )-N s ) / N s .

[0022] The beneficial effects of the present invention are:

[0023] 1. By examining user preferences and social relationships between users, an epidemic-based video dissemination model is established to accurately describe the process of video dissemination in social networks. The basic regeneration number is used to predict the upload bandwidth required for video resources, so as to formulate a video resource allocation strategy to balance the supply and demand of video resources in sudden intensive requests. It can effectively configure video resources and balance supply and demand, thereby improving the scale and speed of video dissemination.

[0024] 2. Divide the dissemination into multiple dissemination rounds. By predicting the basic regeneration number in each dissemination round, the video resource supply strategy can be adjusted in time according to the changes in the traffic demand of video dissemination. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0026] Figure 1 It is a flow chart of the method for supplying mobile video service resources based on demand prediction according to the present invention. DETAILED DESCRIPTION

[0027] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0028] The present invention provides the following specific embodiments:

[0029] like Figure 1 As shown, the present invention provides a method for supplying mobile video service resources based on demand prediction, comprising the following steps:

[0030] S1. Users need to understand and obtain video information before obtaining the video. According to the infectious disease model and the state transition of users during the video dissemination process, users experience four states during the video dissemination process: "uninfected", "susceptible to infection", "infected" and "immune". Users are divided into uninfected people who have not obtained video information, susceptible people who have obtained video information, infected people who are watching videos, and immune people who have completed video viewing.

[0031] Among them, "uninfected": if the user does not obtain any information about the video, they will not request the video (for example, the physical isolation between "uninfected" and the virus); "susceptible to infection": the user knows the information of the video and has the possibility of requesting the video. "Susceptible to infection" users can also spread the information of the video, and continuously expand the scale of "susceptible to infection" users. "Infected": after the user knows the information of the video, if he wants to watch the video, he will send a request message to the video provider who caches the video, or receive the push data of the video from the video push user, and become an infected person. The infected person will push the information and data of the video to other users. "Immune": if the user has watched the video, he will become immune, and the immune person will continue to push video information.

[0032] S2. The probability of a user obtaining a video is measured based on user preferences and user social relationships. The probability of a user in a social network obtaining video data is calculated by formula (1):

[0033]

[0034] in, The probability that the user does not request video data, The probability that the user is not pushed to obtain video data.

[0035] Specifically, in order to clearly indicate that user u i The interest preference range can be i The video collection that has been watched can be clustered by evaluating the content-based similarity between videos and then using clustering methods to cluster the video collection. Among them, the content-based similarity between videos can represent the video as multiple attributes such as name, actor, category, plot summary, etc. The content-based similarity between videos can be calculated using the cosine of the vector angle. This method is widely used in calculating video similarity. After calculating the content-based similarity between videos, popular clustering methods such as spectral clustering can be used to cluster the video collection to generate u i Clustering results of watched video collection IR i =(c1,c2,…,c n ), IR i Each element in contains u i Similar videos that have been watched can also be considered as u i The video category and range of interest, that is, if the video v i With IR i Medium k Class Video p Similar, then v i Can be considered as c k Class potential u i Interested in the video. kEach element in the class can represent u i About c k The boundary of the video of interest. In addition, each video cluster c k There is a cluster head element, which has the minimum average distance to the videos in other clusters. i Watched the video j , and u i Stored and watched v j , then v j Belong to u i The range of interest is calculated by v j With IR i The video similarity of the head element of each video cluster in , and select to join the video cluster to which the cluster head element with the maximum similarity belongs. i For video v j If you are interested, i Request v j The probability of can be defined as:

[0036]

[0037] Among them, s hj v h With v j The similarity value based on video content (v h for u i IR i Medium video cluster c h MAX[s aj ,s bj ,…,s nj ] for all video cluster head nodes and v j The maximum value of similarity. MAX[s aj ,s bj ,…,s nj ]Median hj v j Belongs to the head node s h Video cluster c h . N m / N t for u i Request belongs to video cluster c h The probability of any video in P. ij Indicates u i Request video j probability.

[0038] The dissemination of video information or data through social networks is essentially to replicate the video information or data layer by layer using the social connections between users. A social network can be defined as G = (V, E), where V is the set of vertices in G, and all users in the social network are regarded as vertices; E is the set of edges in G, and the social connections between users are regarded as edges between vertices in G. If there is no edge between two vertices in G, the video information or data cannot be directly transmitted between the two vertices without an edge; if there is an edge between two vertices in G, the two vertices are regarded as social neighbors, and the video information or data is shared with each other through the edge between the vertices. The user who obtains the initial video data is regarded as the source user, and the social neighbors of the source user obtain the video information or data in the form of "request" and "push", and continue to spread the video information or data layer by layer to other social network users in the form of "request" and "push". If any user stops spreading video information or data in the form of "request" and "push", the social neighbors of the user can only obtain the video information or data through other users; if all users in the network stop spreading video information or data, the video is in a dissemination termination state.

[0039] The probability that a user obtains video information in “request” mode:

[0040] Users rely on social neighbors to obtain video information. i Get video v j The probability of information is:

[0041]

[0042] Among them, N si for u i the number of social neighbors; is u i Social Neighbors c Forward v j Probability of information; for u c Not to u c Social Neighbor Forwarding v j Probability of information; for u i All social neighbors do not forward v j The probability of information. If u i Any social neighbor forwards v j Information, u i You can get v j If u g is u i Social neighbors, u g Forward v j The probability of information can be defined as:

[0043]

[0044] in, for u g Forward contains v j The number of all videos in the video cluster; for u g The total number of times all videos have been reposted. for u g From u g Get v from your social neighbors j The probability of information. Obviously, if u g Social neighbors do not forward v j Information, u g You will not get v j information, that is and is u g Get v j The probability of information. and There is a chain relationship and mutual influence, but and The value of is easy to calculate. For example, path hi Indicates that from the source user u h to u i Shortest path. The shortest path length between any two users in a social network is always kept low. According to the six degrees of separation, path hi The length of path should be in the range [2,6]. hg The number of relay nodes in should be in the range [0,4]. h Forward v j The probability of information is 1, then u h Any social neighbor u a Get v j The probability of information is also 1, u h Forward v to its social neighbors j The probability of information is:

[0045]

[0046] in, for u a Forward contains v j The number of all videos in the video cluster; for u a The total number of times all videos have been forwarded. a The forwarding probability of the next social neighbor can be calculated cumulatively according to formula (2) and formula (3).

[0047] The probability of users receiving video information via “push”:

[0048] In the two stages of “susceptible to infection” and “infectious”, the dissemination of video resources in the “push” mode depends on the spread of video information and data in social networks. i Browse and forward a video j After the information, u i You can use v j Information is pushed to u i Users with social connections. The use of the "push" mode to spread video information between two users requires that the two users have an edge in the graph G, that is, social connection. The form of video information push is the same as the form of video data push. In the process of video data push, push user u i Push video links only to other users k , when u k After accepting the push video, u k Send video request message to u i ,u i Transfer video data to u k .

[0049] In the process of video information push, the motivations of video information push mainly include social interaction and video content. i with u i There is a high frequency of video information and data sharing among social neighbors, indicating that u i with u i The social neighbors of u have strong social ties, and u i The social neighbors will also receive u with a higher probability i Push video information and data. i to u i Before social neighbors push video information or data, u i According to u i The strength of social relations between social neighbors and the pushed video content are related to u i The decision of whether to push is made based on the degree of conformity of the preferences of the social neighbors. In fact, the user's push motivation is very complex. For example, when u i and u k Have strong social connections and i and u k Frequent sharing of video information and data, when u k to u k Push video v j When v j The content is not in u i Within the scope of interest, u iwill also accept v with a certain probability j If u k Evaluation j Contents in u i Within the scope of interest, even if u i Not considering u k The social relationship between them will also accept u with a certain probability k Push; if u i and u k Have strong social connections and j Contents in u i Within the scope of interest, then u i will accept v with a higher probability j .u i Accept k Push video v j The probability of information can be defined as:

[0050]

[0051] Among them, β is the weight parameter, which is used to adjust u i and u k The estimated social relationship value SR ki and u i The proportion of interest preference; SR ki is u k to u i Estimated value of the social relationship strength for pushing video information:

[0052]

[0053] Among them, F ki for u i Receive k The frequency of pushing video information; F kt for u k The u k The total frequency of video information pushed by social neighbors; Indicates that in u k All push targets (u k Social Neighbors i The importance of u k Evaluation of video information push i and u k the strength of social relationships; for u i Among all social neighbors of u i The sum of the importance of i Evaluation of video information reception i The strength of social ties with all social neighbors; SR kiIndicates u k In all u i Push video information i The proportion of social relationship strength among social neighbors. i Receive push video v j The probability of information is:

[0054]

[0055] in, Indicates u i A social neighbor did not i Push video v j Probability of information; Indicates u i All social neighbors have not i Push video v j The probability of information.

[0056] Comprehensive evaluation of the probability of users obtaining video information through "request" and "push":

[0057] u i The social neighbors "forwarding" the video information can be regarded as u i Get video information by "request" mode; if u i If u's social neighbors do not "forward" the video information, then i Unable to obtain video information, the video information transmission path is interrupted. In addition to the source user, u i From u i Any social neighbor of u can obtain video information by "request" or "push". i Get video v j The probability of information can be defined as:

[0058]

[0059] in, Indicates u i No video received with "request" j Probability of information; Indicates u i No video received via "push" j Probability of information; Indicates u i No video was received via both "request" and "push" j The probability of information. and The value of depends on the number of social neighbors: the more social neighbors there are, the more channels there are to obtain video information, and the higher the probability of obtaining video information. On the other hand, according to the six degrees of separation theory, the shortest path length between two nodes in a social network should be in the range of [1,7] (i.e., the number of edges). The number of relay nodes contained in the shortest path between nodes in a social network is relatively small, so and The number of cascade layers is relatively small and will not affect and has a greater impact on the value of .

[0060] The probability of users obtaining video data by "request" and "push":

[0061] In the infection stage of the "request" mode, users who have obtained video information are considered to be in the "susceptible to infection" state, and decide whether to request video data based on the consistency between the video content and interest preferences. As mentioned above, in addition to the consistency between the video content and interest preferences, the social relationship between users plays a positive role in making the decision whether to request video data. For example, user u i Received video v j information, and for v j However, most of the u i of social neighbors have watched v j , and forwarded v j At this time, u i Will be affected by u i The influence of social neighbors, actively request v j In the “request” mode, social influence mainly depends on the scale and evaluation of the video data request. i Before requesting and watching a video, u i Refer to the number of video views and video ratings. If the video has a large number of views and good ratings, u i The social relationship will actively request the video. i The impact of making a request for video data can be defined as:

[0062]

[0063] in, v j Number of times it has been viewed; The total number of times all videos have been viewed; Relative to all videos v j The popularity of x max and x min Respectively represent the maximum and minimum values ​​of the video score; For any user u c v j Normalized value of the score; Represents all user pairs v j The mean of the normalized values ​​of the ratings; i Get v j After the information, u i Request v j The probability of the data is:

[0064]

[0065] Among them, α is the weight parameter, which means SI ij and Interaction degree; SI ij and Indicates u i Request v j According to the two-stage video spreading process of pre-infection and infection in the "request" mode, u i Request v j The probability (u i Get v j Information and request v j Data) is:

[0066]

[0067] In the "push" mode, video requests not only rely on the layer-by-layer forwarding of nodes in the social network, but are also affected by the social influence of other users (views and comments). Obviously, social networks can have a positive impact on the performance of video dissemination in the "push" mode, thereby increasing the speed of video information forwarding and pushing, and promoting the scale of video requests. On the other hand, social networks are dynamically changing. The entry and exit of nodes into and out of social networks and the addition and deletion of edges between nodes will affect the social network, which will further affect video dissemination. For example, the dissemination of videos in a "request" manner in social networks can be regarded as the continuous generation of video copies along the edges between vertices in the graph G. If two users u i and u k In a social network, there is no social connection (not a social neighbor), when u i and u k After sharing video information and data, you can establish social connections and become social neighbors. i and u kAfter a new social connection is generated, the social network G will change, and the new social connection will have an impact on the dissemination of video information and data. Although the increase in the number of edges between vertices in the graph G will expand the channels for users to obtain video data, the social strength between users is still the main factor that determines whether users actively request video data or passively accept pushed video data. Therefore, the strength of social relationships between users and the degree of matching between video content and interest preferences are the key factors that determine the success of video push. i Accept k Push v j The probability of video data can be defined as:

[0068]

[0069] Among them, γ is a weight parameter used to adjust and SD ki The value of The proportion of SD ki For user u i and u k The strength of social relationship between people regarding video push, SD ki The value of can be defined as:

[0070]

[0071] in, for u i Accept k The frequency of pushing video data; for u k The u k The total frequency of video data pushed by social neighbors; Indicates that in u k All push targets (u k Social Neighbors i The importance of u k Evaluation of video data push i and u k the strength of social relationships; for u i Among all social neighbors of u i The sum of the importance of i Evaluation of video data reception i The strength of social ties with all social neighbors; SD ki Indicates u k In all u i u that pushes video data i The proportion of social relationship strength among social neighbors. i Receive push v jThe probability of video data can be defined as:

[0072]

[0073] in, Indicates u i A social neighbor did not i Push v j The probability of the data; Indicates u i All social neighbors have not sent any i Push v j The probability of the data;

[0074] Finally, u i Get v j The probability of the data is:

[0075]

[0076] in, No video requested j The probability of the data; for u i No video was received via "push" j The probability of the data; for u i No video is obtained by "request" and "push" at the same time j The probability of the data.

[0077] S3. Establish a video diffusion model based on the infectious disease model to predict the diffusion status of the video in the social network composed of users. The video diffusion model is defined as formula (2):

[0078]

[0079] Among them, N(S) is the number of susceptible people in the social network; N(I) is the number of infected people in the social network; N(R) is the number of immune people in the social network; ε is the rate of new infected people; λ is the rate of new immune people; μ is the rate of new susceptible people.

[0080] The basic reproduction number of the model can be defined as:

[0081]

[0082] When any user u i Get video v j When the data is i As an infected user and u i The duration of the infection state is (0,L]; where L is v jThe length of time. i When in the infected state, u i You can i Social neighbors spread v by "pushing" j information and data, and u i Can be v j The basic reproduction number R0 is the number of other users that an infected user can make infected during the time in the infected state. R0≥1 means that one infected user can surpass one other user to become infected, the replication rate of the infected state is greater than 1, and the scale of video dissemination is gradually increasing; R0<1 means that one infected user can become infected less than one other user, the replication rate of the infected state is less than 1, and the scale of video dissemination is gradually declining.

[0083] S4. Divide the time period from the start of video dissemination to the end of video dissemination into multiple dissemination rounds of equal length, predict the number of newly infected people in each dissemination round based on the probability of users acquiring the video, and predict the basic regeneration number of the video dissemination model in each dissemination round.

[0084] Specifically, the value of the basic reproduction number R0 depends on the values ​​of parameters λ, μ, and ε. λ is the rate of new immune recipients; i Complete the video v j After watching, I lost my interest in the video j If the video is not requested again, j Data and receive video v j Push, that is, u i The state of SS changes from infected state to immune state. j Indicated in the video v j The set of users in the vulnerable state during the spreading process; IS j Indicated in the video v j The set of infected users during the spreading process; j All users in the video completed j After watching, IS j All users in the group leave IS j IS j All users in IS j The residence time range is (0,L]. i The recovery rate can be defined as l k and L j The ratio of k for u i Watch v j The length of time and l k∈(0,L], the predicted value of λ can be defined as IS j The average viewing time for all users in .

[0085]

[0086] Among them, NV c For any user u c The number of videos watched; e For any video v e Length of time; ce For any user u c Watch the video e μ represents the length of time for a new user to join the social network and has completed the video v j Users who viewed and rejected videosv j The total number of users spreading the message.

[0087] ε is the rate of new infections:

[0088]

[0089] Among them, NI j (t i ) and NI j (t j )$ is the i to j Get the video v within the time period j Number of users of the data; NI j (t j )-NI j (t i ) indicates that from t i to j Get the video v within the time period j The increase in the number of users of the data.

[0090] Own video v j The user of the source data is regarded as the source user. Initially, the source user constructs a user set SU(t0); |SU(t0)| represents the number of members in the set SU(t0) and |SU(t0)|=NI j (t0); members in the set SU(t0) spread the video v using social connections j Information and data can be used to obtain video v j The probability of the data is used to infer the video dissemination process and calculate the number of infected users. Drawing on the linear threshold theorem of social networks, the probability threshold T p Measuring the transition of user status: If user u i The probability pd ijGreater than or equal to T p ,u i Get v j Data, u i Become an infected user. For example, at the initial t0, The social neighbors constitute the set SU i If SU i All users in the store have no videos v j data, then u i Get v j The probability of the data is pd ij The value is 0; if SU i The source user is included and SU i The source users in the set SSU i ∈SU(t0),SSU i ∈SU i ,u i Get v j The probability of the data is pd ij The value can be calculated by formula 17.

[0091] pd ij is the predictable value, pd ij The predicted value depends on and To calculate Members (two sets SU i and The difference of) does not store video v j Data, then The value of is 0; The value of u can be determined according to i and Historical data and u of video data push between members i Preference Calculate (for example, and ). For calculation The value of (Two sets of SU i and The difference of) has no members with video v j Information, and The values ​​of are all 0; and pr kj The value of u can be determined according to i and Historical data of video data push between members (for example, SI kj ,F kand F kt ) and u i Preference Calculated. ij The predicted value can be calculated and compared with T p For comparison: pd kj ≥T p , then u i Able to switch from susceptible to infected state and be added to the set IS j and from the set SS j If pd kj <T p ,u i Maintain the current vulnerable status. i PD ij After the predicted value of is measured, other users obtain the video v j The probability prediction value of the data can be calculated, and whether the user status has been transferred can be measured based on the prediction value and the threshold. After the status of all users is measured, the infected users are added to the set IS j Medium; vulnerable users are added to SS j IS j The status of all users in the IS can be continuously measured. j The member increment of 0 can be regarded as the convergence condition of the iterative process. j The members are users who are infected during the entire video dissemination process.

[0092]

[0093] Among them, |IS j | for the set IS j The number of members in; t0 and t e For video v j The start and end time of the dissemination. |IS j The value of | can be obtained according to the video v j Data probability measures are calculated. j Each member u i is defined as a triple u i =(u,r x ,sn), where u is user information; r x is the prediction iteration round; sn is u in r x Round Join Collection IS j The sequence number of the time interval t e -t0 can be divided into multiple time segments of equal length, which can be regarded as spreading rounds, namely TS j =t e-t0=(r1,r2,...,r m ).

[0094] t e -t0 can be regarded as video v j The cumulative sum of the duration of all dissemination rounds of:

[0095]

[0096] Among them, m is the video v j The total number of dissemination rounds; For video v j The number of rounds of dissemination r c Duration of SIS j (r h ) is defined as IS j A subset of SIS j (r h ) is a member in the dissemination round r h The set of users who transition from the susceptible state to the infected state. The value of can be calculated using the following formula:

[0097]

[0098] in, SIS for collection j (r h ) For r h The round start time is r h The time interval between the round end time and the user u c ∈SIS j (r h ) is the average length of time it takes for a person to transition from a susceptible state to an infected state. The value of can be calculated based on historical statistics:

[0099]

[0100] Among them, m is the c Participate in the video dissemination process (during the video dissemination process c The number of videos that were spread during the transition from the susceptible state to the infected state; ce For the video v e Spread the start time to user u c The time interval of the state transition from the susceptible state to the infected state. The duration of the spreading round and the number of users who transitioned from the susceptible state to the infected state during the spreading round can be calculated.

[0101] Finally, in rh The transmission round conversion rate ε(r h ) can be defined as:

[0102]

[0103] The value of μ can be calculated using historical statistical information for video diffusion prediction. For example, the time interval TS j =t e -t0 can be divided into spreading rounds TS j =t e -t0=(r1,r2,...,r m ) set. a (r i ) and N r (r i ) represent the newly joined social network and the newly spread round r i The number of users who do not participate in video dissemination. The value of μ can be defined as:

[0104]

[0105] Among them, |TS j | for the set TS j The number of all members in ; and They represent the growth rate of the number of new users joining the social network and the number of new users in the dissemination round r. c The growth rate of the number of users who refuse to disseminate video information and data. a (r c ) and N r (r c ) can be obtained by spreading round r c The historical statistical information is calculated.

[0106] According to the above calculation method of the values ​​of μ, ε and λ in the spreading round, the value of the basic reproduction number in each spreading round can be calculated. h Number of users who transitioned from vulnerable to infected state in |SIS j (r h )|You can get the video v according to the measurement user j The probability of the data is calculated. Therefore, the video v j Any dissemination round r h The internal basic reproduction number R0(r h ) can be defined as:

[0107]

[0108] S5. Formulate a mobile video service resource supply strategy, and judge whether the stored video data needs to provide additional video copies based on the relationship between the basic reproduction number in the current dissemination round and the amount of video data supply service that can be provided to other users by the stored video data. i )≥1 and N s ≥R0(r i ), social networks do not require additional copies of the video (R0(r i )-N s ) / N s , if R0(r i )≥1 and N s <R0(r i ), social networks need additional videos v j The copy (R0(r i )-N s ) / N s , to meet the user's demand for bandwidth; if R0(r i )<1, it means that the video v j The spread of video is declining, and social networks do not need additional video v j A copy of .

[0109] S6. Execute a mobile video service resource supply strategy based on the value of the basic regeneration number in the current broadcast round.

[0110] By examining user preferences and social relationships between users, an epidemic-based video dissemination model is established to accurately describe the process of video dissemination in social networks. The basic regeneration number is used to predict the upload bandwidth required for video resources, so as to formulate a video resource allocation strategy to balance the supply and demand of video resources in bursty intensive requests. It can effectively configure video resources and balance supply and demand, thereby improving the scale and speed of video dissemination.

[0111] The dissemination is divided into multiple dissemination rounds. By predicting the basic regeneration number in each dissemination round, the video resource supply strategy can be adjusted in time according to the changes in the traffic demand of video dissemination.

[0112] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention to be protected.

Claims

1. A method for providing mobile video service resources based on demand prediction, characterized in that: The steps include: S1, dividing users into uninfected people who have not obtained video information, susceptible people who have obtained video information, infected people who are watching the video, and immune people who have completed watching the video; S2, measure the probability of users obtaining videos based on user preferences and user social relationships; S3. Establish a video diffusion model based on the infectious disease model to predict the diffusion status of the video in the social network composed of users; S4, dividing the time period from the start of video dissemination to the end of dissemination into multiple dissemination rounds of equal length, predicting the number of newly infected people in each dissemination round based on the probability of users acquiring the video, and predicting the basic reproduction number of the video dissemination model in each dissemination round; S5. Formulate a mobile video service resource supply strategy, and judge whether the stored video data needs to provide additional video copies according to the relationship between the basic reproduction number in the current dissemination round and the amount of video data supply service that can be provided to other users by the stored video data; S6. Execute a mobile video service resource supply strategy based on the value of the basic regeneration number in the current broadcast round.

2. The method for providing mobile video service resources based on demand prediction according to claim 1, characterized in that: Specifically, in S2, the probability of a user in the social network obtaining video data is calculated by formula (1); in, is the probability that the user does not obtain video data with preference, The probability that the user does not obtain video data through social push.

3. The method for providing mobile video service resources based on demand prediction according to claim 1, characterized in that: In S3, the video diffusion model is defined as formula (2); Among them, N(S) is the number of susceptible people in the social network; N(I) is the number of infected people in the social network; N(R) is the number of immune people in the social network; ε is the rate of new infected people; λ is the rate of new immune people; μ is the rate of new susceptible people.

4. The method for providing mobile video service resources based on demand prediction according to claim 1, characterized in that: Specifically, in S5, when the basic reproduction number in the current dissemination round is greater than or equal to 1 and greater than the number of stored video data that can provide video data supply services to other users, storing the video data requires additional video copies.

5. The method for providing mobile video service resources based on demand prediction according to claim 4, characterized in that: Specifically, the basic reproduction number R0(r i ) is greater than or equal to 1 and greater than the number N of stored video data that can provide video data supply services to other users s When the social network needs an additional copy of the video (R0(r i )-N s ) / N s .

Images