Intelligent information pushing method and system based on cloud computing and big data

Abstract

The invention discloses an intelligent information pushing method and system based on cloud computing and big data and a storage medium, and the method comprises the steps: obtaining registration information of a user on a page and an input interest label, building an initialized user interest model, obtaining behavior record information of the user on the page according to the registration information and the user interest model, and pushing the behavior record information to the user. User interest feature words are obtained through classification processing, a relation neighborhood model of the user is constructed based on the user interest feature words and the behavior record information, and hotspot analysis and neighborhood authoritative user analysis are conducted on the relation neighborhood model to obtain a neighborhood chain; the method comprises the steps of obtaining trust relationships among users in a relationship neighborhood, establishing a trust model, determining an optimal push channel among the users based on a neighborhood chain and the trust model to complete intelligent information push, classifying social circles of the users in combination with the neighborhood model to obtain a propagation mode of information in the social circles, improving the push accuracy of a push system, and improving the user experience. And the timeliness and the reliability of information pushing are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of information push, and particularly relates to an intelligent information push method and system based on cloud computing and big data. Background Art

[0002] With the continuous development of network technology and the explosion of information volume, information mining has attracted more and more attention. Information mining technology can help people effectively obtain the information they really need from the Internet. Its mining process is a process of continuously searching for the most relevant information according to the query information input by users. The core part of information mining technology is the matching and selection of information query and information set. Due to the huge computational amount of the push system, the generation efficiency of push information is restricted, which is also the reason why most recommendation systems cannot ensure both accuracy and timeliness at the same time. However, since the complexity of the information mining process is closely related to the user's ability to query information, that is, whether the information input by the user can accurately express their needs determines the length of the information mining process. Therefore, it is necessary to provide a method for users to accurately and quickly mine the information they really need from a large amount of uncertain information. Summary of the Invention

[0003] In view of this, the present invention provides an intelligent information push method, system and storage medium based on cloud computing and big data to improve the accuracy of information push and the best channel, and specifically adopts the following technical solutions to achieve.

[0004] In a first aspect, the present invention provides an intelligent information push method based on cloud computing and big data, including the following steps:

[0005] Obtain the registration information of the user on the page and the input interest tags, and establish an initial user interest model;

[0006] Obtain the behavior record information of the user on the page according to the registration information and the user interest model, convert the behavior record information into text information and perform classification processing to obtain user interest feature words;

[0007] Construct a relationship neighborhood model of the user based on the user interest feature words and the behavior record information, and perform hotspot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the neighborhood chain of the user;

[0008] Obtain the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push.

[0009] As a preference of the above technical solution, obtaining the registration information of the user on the page and the input interest tags, and establishing an initial user interest model includes:

[0010] The model representation algorithm based on the vector space is adopted, and an interest status parameter is added to determine the changes in user interests over time. The corresponding expression is:

[0011] User = {<I1, W 1, ω1>, <I2, W2, ω2>...... <I n , W n , ω n >}

[0012] Among them, I i represents the interest word i of the user, W i represents the weight of the interest word i, and ω i represents the user interest status parameter; the user interest model will combine the feedback of user display information and implicit information;

[0013] An initial user interest model is established according to the user's registration information and the input interest tags, and then the user interest model is updated by tracking and recording the user's historical behavior. Specifically, it includes:

[0014] Establish an initial user interest model U0 = {<I1, W1, ω1>, <I2, W2, ω2>... <I n , W n , ω n >};

[0015] The text page vector space obtained by mining the pages that the user is interested in is T0 = {<I1, W t1 , ω1>, <I2, W t2 , ω2>... <I n , β, ω n >}(W t1 , W t2 ... > β);

[0016] Compare U0 and T0. If U0 and T0 contain the same interest words, update the weights of the keywords in U0 to the weights of the keywords in T0; if there are interest words in T0 that are not in U0, add the interest words to U0 and delete the interest words in U0 that are lower than the average weight of the interest words in the text page vector to complete the first update of the user interest model. The obtained expression is:

[0017]

[0018] The subsequent model updates repeat the above process.

[0019] As an optimization of the above technical solution, a relationship neighborhood model of the user is constructed based on the user interest feature words and behavior record information, including:

[0020] The default set X is a domain, and R is a mapping from X×X to [0,1], that is, It means that the mapping R determines a fuzzy relation neighborhood on X, namely the fuzzy neighborhood;

[0021] Assume Q is a set of relations, use r and s to represent different relations, where r, s∈Q, and use N r (x) represents a fuzzy neighborhood of object x under relation r.

[0022] As a preferred embodiment of the above technical solution, K is preset as an interest set, k0, k i ∈K,g(t,k i ) is time t k i The frequency of occurrence of the word in H(k0,Δt) represents the heat of k0 at time t0+Δ;

[0023] If H(k0,Δt)>δ1, then k0 is called a hot interest that meets the δ1 heat level, where δ1 is used to determine the appropriate threshold given by the hot interest. Whether the object interest has changed is determined by comparing the changes in the object search keywords and the changes in the keyword weights between the previous period and the current period;

[0024] If a certain type of keywords appears in the current time period, the interests corresponding to this type of keyword set are the user's newly added interests;

[0025] If a certain type of keyword appeared in the previous period but not in the current period, the interest corresponding to this type of keyword set will fade;

[0026] The user's interest level is calculated based on the number of views, searches, favorites, forwardings, and behavioral records of the market used. The time the user spends on the interest is regarded as the user's interest level. k(x,t) represents the interest level of interest k of x at time t.

[0027] As a preferred embodiment of the above technical solution, the interest updating process includes:

[0028] N(k(t)≥a) is introduced to represent the neighborhood of users whose interest in interest k is not less than a at time t. It is assumed that at time t, all interests of user x are K(x,t,α)={k1,k2...k n}, at time t+Δt, all interests of user x are K(x,t+Δt,a)={k1,k2...k m};

[0029] By comparing K(x, t + Δt, a) at time t + Δt with K(x, t, α) at time t, it is determined whether the interest of x changes within the time period Δt. That is, if K(x, t + Δt, a) / K(x, t, α) ≠ Φ, then there exists k i ∈K(x, t + Δt, a) and then x has added a new interest k i , k i ∈K(x, t + Δt, a), and x has a new neighborhood N(k i (t) ≥ a). Otherwise, x has not added a new interest within the time period Δt;

[0030] If K(x, t, a) / K(x, t + Δt, a) ≠ Φ, then there exists k j ∈K(x, t, a), and It is determined whether x no longer pays attention to the interest k j ∈K(x, t, a) according to the size of Δt. If x no longer pays attention to the interest k j ∈K(x, t, a), then the relationship between x and the objects in the neighborhood corresponding to the interest changes;

[0031] If K(x, t, α) ∩ K(x, t + Δt, a) ≠ Φ, and k j (x, t + Δt) - k j (x, t) > δ2, k j ∈K(x, t, a) ∩ K(x, t + Δt, a), then x's interest in k j increases, where δ2 is the degree of increase in interest.

[0032] As an optimization of the above technical solution, hotspot analysis and neighborhood authoritative user analysis are performed on the relationship neighborhood model to obtain the neighborhood chain of users, including:

[0033] It is preset that K is the interest set, X is the user set, x ∈ X, and within Δt, the interests added by x are:

[0034] ΔK(x, Δt, a) = {k i | i = 1, 2... n, k i ∈K};

[0035] The interests added by all users in X are:

[0036] ΔK(Δt, a) = ∪ x∈X ΔK(x, Δt, a) = {k i | i = 1, 2... m, k i ∈K};

[0037] All neighborhoods formed around the new interests are:

[0038] ΔN(Δt,a) = {N(k i (t)≥a)|k i ∈ΔK(Δt,a)};

[0039] Let

[0040] Then the interests in the set ΔK(H>δ1) are the popular interests that meet the δ1 heat level, and the neighborhoods in the set ΔN(H>δ1) are the neighborhoods formed around each popular interest;

[0041] Preset N(k i (t)≥a) represents the neighborhood formed around the interest k i , k i ∈ΔK(H>δ1), x∈N(k i (t)≥a), B(x,k i ,Δt) represents the set of search terms of x regarding k i within Δt time. Then the set of search terms of all objects in N(k i (t)≥a) regarding k i within Δt time is:

[0042]

[0043] The interest k i contains different sub - interests k ij , then

[0044] where B(k ij ,Δt) represents the set of search terms of all objects in N(k i (t)≥a) regarding k ij within Δt time. Let

[0045] Then ΔK(k i ,H>δ1) represents the true popular interests in k i , and ΔN(K i ,H>δ1) represents all the neighborhoods formed around the interest k ij .

[0046] As an optimization of the above technical solution, represent the social network between users as G=(V,E), where V = {v1,V2...V n} represents the set of nodes, represents the set of edges between nodes. The edges between nodes represent the interactions between nodes. Use the Gaussian potential function of short - range field interaction to describe the interactions between nodes. In the social network G=(V,E), v i ∈V, let Among them, m j is the mass of v j , v j ∈V, j = 1, 2... n, d ij represents the distance between v i and v j , and σ represents the influence factor and is used to control the influence range of each node.

[0047] As an optimization of the above technical solution, obtain the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push, including:

[0048] Preset N9k9t)≥α) is the neighborhood formed around interest k, x, y ∈ N9k9t)≥α), and e(t, x, y, k) represents the number of times y approves of x during the time period [t - Δt, t]. Let E(t, x, k) = ∑ y∈N(k(t)≥α) e(t, x, y, k), and E(t, x, k) is the neighborhood energy of x in N(k(t)≥α) at time t;

[0049] The initial trust degree between users is expressed as Among them, E(t, y, k) represents the neighborhood energy of y in N(k(t)≥α) at time t;

[0050] Preset l is a neighborhood chain formed by several neighborhoods. Let R i+1 (t, x i , x i+1 ) is the energy resistance between two adjacent objects x i and x i+1 on l at time t. Among them, x i , x i+1 ∈l, i = 0, 1... n, D(t, x i , x i+1 ) is the horizontal comprehensive difference degree between x i and x i+1 at time t, is the topological potential of x i at time t, F(t, x i , x i+1 ) is the trust degree of x i+1 in x i at time t, and F(t, x i , x i+1 )≠0;

[0051] Then let R(t, x0, x n ) is the connection between x0 and x at time tn The energy resistance of chain l The operation is calculated according to the following expression:

[0052]

[0053] The energy loss of information transmission along the neighborhood chain can be characterized by energy resistance. R(t, x0, x n ) represents using the energy resistance between adjacent users on the neighborhood chain to measure the energy resistance of the neighborhood chain, and R'(t, x0, x n ) represents using the maximum value of the energy resistance between adjacent users on the neighborhood chain, that is, the energy resistance of the neighborhood chain.

[0054] In a second aspect, the present invention also provides an intelligent information push system based on cloud computing and big data, including:

[0055] An information acquisition unit, configured to acquire the registration information of the user on the page and the input interest tags, and establish an initial user interest model;

[0056] A feature extraction unit, configured to obtain the behavior record information of the user on the page according to the registration information and the user interest model, convert the behavior record information into text information and perform classification processing to obtain user interest feature words;

[0057] A model analysis unit, configured to construct a relationship neighborhood model of the user based on the user interest feature words and the behavior record information, and perform hotspot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the neighborhood chain of the user;

[0058] An information push unit, configured to obtain the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push.

[0059] In a third aspect, the present invention also provides a storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned intelligent information push method based on cloud computing and big data are implemented.

[0060] The present invention provides an intelligent information push method, system and storage medium based on cloud computing and big data. By obtaining the registration information of users on the page and the input interest tags, and establishing an initial user interest model, the behavior record information of users on the page is obtained according to the registration information and the user interest model. The behavior record information is converted into text information and classified to obtain user interest feature words. A relationship neighborhood model of users is constructed based on the user interest feature words and the behavior record information, and hotspot analysis and neighborhood authoritative user analysis are performed on the relationship neighborhood model to obtain the neighborhood chain of users. The trust relationship between users in the relationship neighborhood is obtained and a trust model is established. Based on the neighborhood chain and the trust model, the best push channel between users is determined to complete intelligent information push. Combining the neighborhood model to classify the social circle of users to obtain the dissemination mode of information in the social circle, the push accuracy of the push system is improved, and the timeliness and reliability of information push are effectively enhanced. Brief Description of the Drawings

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0062] Figure 1 It is a flowchart of the intelligent information push method based on cloud computing and big data provided by the present invention;

[0063] Figure 2 It is a structural block diagram of the intelligent information push system based on cloud computing and big data provided by the present invention. Detailed Embodiments

[0064] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

[0065] Refer to Figure 1 , the present invention provides an intelligent information push method based on cloud computing and big data, including the following steps:

[0066] S1: Obtain the registration information of users on the page and the input interest tags, and establish an initial user interest model;

[0067] S2: Obtain the user's behavior record information on the page according to the registration information and the user interest model, convert the behavior record information into text information and perform classification processing to obtain the user interest feature words;

[0068] S3: Build a user's relationship neighborhood model based on the user interest feature words and the behavior record information, and perform hot spot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the user's neighborhood chain;

[0069] S4: Obtain the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push.

[0070] In this embodiment, obtain the user's registration information on the page and the input interest tags, and establish an initial user interest model, including: using a model representation algorithm based on vector space and adding an interest status parameter to determine the change of the user interest over time, and the corresponding expression is: User = {<I1, W 1, ω1>, <I2, W2, ω2>...... <I n , W n , ω n >}, where, I i represents the user's interest word i, W i represents the weight of the interest word i, ω i represents the user interest status parameter; the user interest model will combine the feedback of the user's display information and the implicit information; establish an initial user interest model according to the user's registration information and the input interest tags, and then update the user interest model by tracking and recording the user's historical behavior, specifically including: establishing an initial user interest model U0 = {<I1, W1, ω1>, <I2, W2, ω2>... <I n , W n , ω n >};

[0071] The text page vector space obtained by mining the pages that the user is interested in is T0 = {<I1, W t1 , ω1>, <I2, W t2 , ω2>... <I n , β, ω n >}(W t1 , W t2 ... > β);

[0072] Compare U0 and T0. If U0 and T0 contain the same interesting words, update the weight of the keywords in U0 to the weight of the keywords in T0. If there are interesting words in T0 that are not in U0, add the interesting words to U0 and delete the interesting words in U0 that are lower than the average weight of the interesting words in the text page vector to complete the first update of the user interest model, and the obtained expression is:

[0073]

[0074] Repeat the above process for subsequent model updates.

[0075] It should be noted that the user's interests change at all times. The user may suddenly be very interested in a new thing or gradually lose interest in a once-loved commodity. Therefore, when building a user interest model, it is necessary to consider the impact of time factors on the user's interests to fully ensure the accuracy and effectiveness of the user interest model. According to the characteristics of time, the user's interests are divided into long-term user interests and short-term user interests. Determine whether it is a long-term interest or a short-term interest based on the user's browsing time and realize the mutual conversion between long-term interests and short-term interests. At the current moment, the user's own understanding of their interests and hobbies is the most accurate. When the user registers an account, the system will require the user to select three interest tags in the order of interest degree from among many interest tags (this requirement only exists when the user registers for the first time) and store the information in the database, and establish an initial user interest model based on this information. The mining of user interests is the basis for building a user interest model. Two methods are adopted to obtain user information. Explicit information acquisition: When the user registers an account, the system will require the user to select the tag words they are interested in. Implicit information acquisition: Obtain the user's implicit interests and hobbies by mining web logs and recording the user's historical behaviors. The method of implicit information acquisition is relatively complex compared to explicit acquisition. The keywords that the user is interested in are obtained through the content mining of web text pages.

[0076] It should be understood that the construction of the user interest model will combine the feedback of explicit user information and the feedback of implicit information. An initial user interest model is established based on the user's registration information and input interest tags, and then the user interest model is updated by tracking and recording the user's historical behavior. No one knows the user's interests and hobbies better than the user himself. Therefore, the initial user interest model established based on the interest tags selected when the user registers the account can best reflect the user's current interests and hobbies. Over time, hobbies will change, and the model is updated according to the user's behavior using the vector space representation method. By obtaining the user's registration information on the page and the input interest tags, and establishing an initial user interest model, the user's behavior record information on the page is obtained based on the registration information and the user interest model. The behavior record information is converted into text information and classified to obtain user interest feature words. A relational neighborhood model of the user is constructed based on the user interest feature words and the behavior record information, and hot spot analysis and neighborhood authoritative user analysis are performed on the relational neighborhood model to obtain the user's neighborhood chain. The trust relationship between users in the relational neighborhood is obtained and a trust model is established. The best push channel between users is determined based on the neighborhood chain and the trust model to complete intelligent information push. The social circle of the user is classified in combination with the neighborhood model to obtain the dissemination method of information in the social circle, improving the push accuracy of the push system and effectively enhancing the timeliness and reliability of information push.

[0077] Optionally, constructing a relational neighborhood model of the user based on the user interest feature words and the behavior record information includes:

[0078] Preset set X is a universe of discourse, and R is a mapping from X×X to [0,1], that is It means that the mapping R determines a fuzzy relational neighborhood on X, that is, a fuzzy neighborhood;

[0079] Preset Q is a set of relationships, and r and s are used to represent different relationships, where r, s ∈ Q, and N r (x) represents a fuzzy neighborhood of object x under the r relationship.

[0080] In this embodiment, preset K is a set of interests, k0, k i ∈ K, g(t, k i ) is the frequency of occurrence of the word in k i at time t, let H(k0, Δt) represents the popularity of k0 at the moment t0 + Δt; if H(k0, Δt) > δ1, then k0 is called a popular interest that meets the δ1 popularity level, where δ1 is used to determine an appropriate threshold for popular interests. It is judged whether the object interest has changed by comparing the changes in the object search keywords and the keyword weights in the previous period and the current period; if a certain type of keyword appears in the current period, the interest corresponding to the keyword set of this type is the newly added interest of the user; if a certain type of keyword appeared in the previous period and does not appear in the current period, the interest corresponding to the keyword set of this type fades; among them, the degree of the user's interest is calculated based on the behavior record information such as the number of views, search times, collections, forwards, and the used market. The duration of time the user spends on the interest is regarded as the degree of their interest, and k(x, t) is denoted to represent the degree of interest of the interest k of x at the moment t.

[0081] It should be noted that the process of interest update includes:

[0082] Introduce N(k(t) ≥ a) to represent the neighborhood composed of users whose degree of interest in the interest k at the moment t is not less than a. It is preset that at the moment t, all the interests of the user x are K(x, t, α) = {k1, k2... k n}, and at the moment t + Δt, all the interests of the user x are K{x, t + Δt, a} = {k1, k2... k m};

[0083] By comparing K(x, t + Δt, a) at the moment t + Δt and K(x, t, α) at the moment t, it is judged whether the interest of x has changed within the Δt time period, that is, if K(x, t + Δt, a) / K(x, t, α) ≠ Φ, then there exists k i ∈K(x, t + Δt, a) and then x has added a new interest k i , k i ∈K(x, t + Δt, a), and x has a new neighborhood N(k i (t) ≥ a), otherwise, x has not added a new interest within the Δt time period;

[0084] If K(x, t, a) / K(x, t + Δt, a) ≠ Φ, then there exists k j ∈K(x, t, a), and Judge whether x no longer pays attention to the interest k j ∈K(x, t, a) according to the size of Δt. If x no longer pays attention to the interest k j ∈K(x, t, a), then the relationship between x and the objects in the neighborhood corresponding to the interest changes;

[0085] If \(K(x,t,\alpha)\cap K(x,t + \Delta t,\alpha)\neq\varnothing\) and \(k\) j (x,t + \Delta t)-k j (x,t)>\delta^2\), \(k\) j \(\in K(x,t,\alpha)\cap K(x,t + \Delta t,\alpha)\), then the interest of \(x\) in \(k\) j increases, where \(\delta^2\) is the degree of increase in interest.

[0086] First, analyze the time-varying process of the number of object neighborhoods. Usually, an object will search or consult on relevant websites according to its needs. By analyzing the content of the information consulted by the object on the Internet, it is judged whether the interest of the object has changed. In the big data era, the storage types of information are diverse, and so are the ways for objects to obtain information. To study the change of object interest using big data, it is first necessary to obtain the online behavior records of the object over a period of time, then analyze its content, then represent the content of this information in words, and finally determine the interest of the object according to the word set. Moreover, the interest degree of the object in each interest can also be calculated according to the weights of the words in the word set. The online behavior record information of the object includes all the websites it browses, clicks, and collects, as well as the specific information within the websites and the corresponding browsing time, search times, etc.

[0087] Optionally, perform hotspot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the neighborhood chain of users, including:

[0088] Preset \(K\) as the interest set and \(X\) as the user set, \(x\in X\). Within \(\Delta t\), the increased interest of \(x\) is:

[0089] \(\Delta K(x,\Delta t,\alpha)=\{k\) i |i = 1,2...n,k\) i \(\in K\}\);

[0090] The increased interests of all users in \(X\) are:

[0091] \(\Delta K(\Delta t,\alpha)=\cup\) x∈X \(\Delta K(x,\Delta t,\alpha)=\{k\) i |i = 1,2...m,k\) i \(\in K\}\);

[0092] All neighborhoods formed around the new interest are:

[0093] \(\Delta N(\Delta t,\alpha)=\{N(k\) i (t)\geq\alpha)|k\) i \(\in\Delta K(\Delta t,\alpha)\}\);

[0094] Let

[0095] Then the interests in the set ΔK(H>δ1) are the popular interests that meet the δ1 heat level, and the neighborhoods in the set ΔN(H>δ1) are the neighborhoods formed around each popular interest;

[0096] Preset N(k i (t)≥a) represents the neighborhood formed around the interest k i where k i ∈ΔK(H>δ1), x∈N(k i (t)≥a), and B(x,k i ,Δt) represents the set of search terms of x with respect to k within Δt time. i Then the set of search terms of all objects in N(k i (t)≥a) with respect to k within Δt time is: i

[0097]

[0098] The interest k i contains different sub-interests k ij , then

[0099] where B(k ij ,Δt) represents the set of search terms of all objects in N(k i (t)≥a) with respect to k within Δt time. Let ij

[0100] Then ΔK(k i ,H>δ1) represents the true popular interests in k i , and ΔN(k i ,H>δ1) represents all the neighborhoods formed around the interest k ij .

[0101] In this embodiment, the social network between users is represented as G=(V,E), where V = v1, V2...V n} represents the set of nodes, represents the set of edges between nodes. The edges between nodes represent the interactions between nodes. The Gaussian potential function of short-range field interaction is used to describe the interactions between nodes. In the social network G=(V,E), v i ∈V, let where m j is the mass of v j , v j ∈V, j = 1,2...n, d ij represents the distance between v i and v j , and σ represents the influence factor and is used to control the influence range of each node.

[0102] It should be noted that obtaining the trust relationships among users in the relationship neighborhood and establishing a trust model, and determining the best push channels among users based on the neighborhood chain and the trust model to complete intelligent information push, including:

[0103] Preset N(k(t)≥α) is the neighborhood formed around the interest k. For x,y∈N(k(t)≥α), e(t,x,y,k) represents the number of times y approves of x during the time period [t - Δt,t]. Let E(t,x,k) = ∑ y∈N(k(t)≥α) e(t,x,y,k). E(t,x,k) is the neighborhood energy of x in N(k(t)≥α) at time t;

[0104] The initial trust degree among users is expressed as where E(t,y,k) represents the neighborhood energy of y in N(k(t)≥α) at time t;

[0105] Preset l is a neighborhood chain formed by several neighborhoods. Let R i+1 (t,x i ,x i+1 ) is the energy resistance between two adjacent objects x i and x i+1 on l at time t. Among them, x i ,x i+1 ∈l, i = 0,1...n, D(t,x i ,x i+1 ) is the horizontal comprehensive difference degree between x i and x i+1 at time t, is the topological potential of x i at time t, F(t,x i ,x i+1 ) is the trust degree of x i+1 in x i at time t, and F(t,x i ,x i+1 )≠0;

[0106] Then let R(t,x0,x n ) be the energy resistance of the chain l connecting x0 and x n at time t, The operation is calculated according to the following expression:

[0107]

[0108] The energy loss when information is transmitted along the neighborhood chain can be characterized by the energy resistance. R(t,x0,x n) It represents using the energy resistance between adjacent users on the neighborhood chain to measure the energy resistance of the neighborhood chain, R'(t,x0,x n ) It represents using the maximum value of the energy resistance between adjacent users on the neighborhood chain, that is, the energy resistance of the neighborhood chain.

[0109] In order to more accurately understand people's daily lives and improve people's quality of life, it is necessary to determine the hot events that objects are most concerned about and interested in at a certain time and place. By analyzing the degree of attention of objects in the neighborhood formed around the event to the event, find out those hot events that receive the most attention. First, determine the increased interests of each object within a period of time, and then find out those interests that receive the most attention from these newly increased interests. In this way, it is possible to clarify what objects are most concerned about and what they need at what time. Those interests that receive the most attention are the current hot topics, thus improving the accuracy of information push.

[0110] Refer to Figure 2 , the present invention also provides an intelligent information push system based on cloud computing and big data, including:

[0111] An information acquisition unit, configured to acquire the registration information of the user on the page and the input interest tags, and establish an initial user interest model;

[0112] A feature extraction unit, configured to acquire the behavior record information of the user on the page according to the registration information and the user interest model, convert the behavior record information into text information and perform classification processing to obtain user interest feature words;

[0113] A model analysis unit, configured to construct a relationship neighborhood model of the user based on the user interest feature words and the behavior record information, and perform hot spot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the neighborhood chain of the user;

[0114] An information push unit, configured to acquire the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push.

[0115] In this embodiment, the process of establishing the initial trust relationship between objects, i.e., between users, by the recommendation system is analyzed. The establishment process is divided into direct and indirect establishment processes. The indirect establishment process is that an object combines the opinions of its surrounding objects to obtain the initial trust degree of a certain object. The direct establishment process is that objects establish initial trust through interactions, transactions, etc. When different push channels are selected, the push effects are different. When a website pushes information, there are different requirements for the information dissemination effect. Some websites require the information to expand the influence range during dissemination, while some only need to quickly disseminate it to specific objects. The present invention adopts the best push channel with the smallest energy resistance. The energy resistance between adjacent objects on the neighborhood chain mainly considers the relationship between two objects. The stronger the relationship between two adjacent objects on the chain, the smaller the energy resistance of the neighborhood chain. Therefore, when selecting a channel with a smaller energy resistance to push information, the loss of information during transmission is the smallest. The channel with the smallest energy resistance is called the best push channel with the smallest energy resistance.

[0116] It should be noted that by obtaining the registration information of users on the page and the input interest tags, and establishing an initial user interest model, the behavior record information of users on the page is obtained according to the registration information and the user interest model. The behavior record information is converted into text information and classified to obtain user interest feature words. A relationship neighborhood model of users is constructed based on the user interest feature words and the behavior record information, and a hot spot analysis and a neighborhood authoritative user analysis are performed on the relationship neighborhood model to obtain the neighborhood chain of users. The trust relationship between users in the relationship neighborhood is obtained and a trust model is established. The best push channel between users is determined based on the neighborhood chain and the trust model to complete intelligent information push. By classifying the social circle of users in combination with the neighborhood model, the dissemination method of information in the social circle is obtained, improving the push accuracy of the push system, and also effectively enhancing the timeliness and reliability of information push.

[0117] In a feasible embodiment, the present invention also provides a storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned intelligent information push method based on cloud computing and big data are implemented.

[0118] In all the examples shown and described here, any specific value should be construed as merely exemplary, not as a limitation. Therefore, other examples of the exemplary embodiments may have different values.

[0119] It should be noted that similar reference numerals and letters denote similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0120] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention.

Claims

1. An intelligent information push method based on cloud computing and big data, characterized in that: The steps include: Obtain the user's registration information and input interest tags on the page, and establish an initial user interest model; Obtain the user's behavior record information on the page according to the registration information and the user interest model, convert the behavior record information into text information and perform classification processing to obtain user interest feature words; Construct a relationship neighborhood model of the user based on the user interest feature words and behavior record information, and perform hotspot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the user's neighborhood chain; Obtain the trust relationship among users in the relationship neighborhood and establish a trust model, and determine the best push channel among users based on the neighborhood chain and the trust model to complete intelligent information push.

2. The intelligent information push method based on cloud computing and big data according to claim 1 is characterized in that: Obtain the user's registration information and input interest tags on the page, and establish an initial user interest model, including: Adopt a model representation algorithm based on vector space and add an interest status parameter to determine the changes in user interests over time. The corresponding expression is: User={ <I1,W 1, ω1>,<I2,W2,ω2> ...... n ,W n ,oh n >}​ Among them, I i represents the user's interest word i, W i represents the weight of interest word i, ω i Represents the user interest state parameter; the user interest model will combine the user's feedback on displayed information and implicit information; Establish an initial user interest model according to the user's registration information and input interest tags, and then update the user interest model by tracking and recording the user's historical behaviors, specifically including: Establish the initial user interest model U0 = {<I1,W1,ω1> ,<I2,W2,ω2> ... n ,W n ,ω n >};​ The text page vector space obtained by mining the pages that users are interested in is T0 = { <I1,W t1 ,ω1>, <I2,W t2 ,ω2>... n ,β,ω n >}(W t1 ,W t2 ...>β);​ Compare U0 and T0. If U0 and T0 contain the same interest words, update the weight of the keywords in U0 to the weight of the keywords in T0; if there are interest words in T0 that do not exist in U0, add the interest words to U0 and delete the interest words in U0 that are lower than the average weight of the interest words in the text page vector to complete the first update of the user interest model, and the obtained expression is: Repeat the above process for subsequent model updates.

3. The intelligent information push method based on cloud computing and big data according to claim 2 is characterized in that: Construct a relationship neighborhood model of the user based on the user interest feature words and behavior record information, including: The default set X is a domain, and R is a mapping from X×X to [0,1], that is, It means that the mapping R determines a fuzzy relation neighborhood on X, namely the fuzzy neighborhood; Assume Q is a set of relations, use r and s to represent different relations, where r, s∈Q, and use N r (x) represents a fuzzy neighborhood of object x under relation r.

4. The intelligent information push method based on cloud computing and big data according to claim 3 is characterized in that: It also includes: Preset K as an interest set, k0,k i ∈K,g(t,k i ) is time t k i The frequency of occurrence of the word in H(k0,Δt) represents the heat of k0 at time t0+Δ; If H(k0,Δt)>δ1, then k0 is called a popular interest that satisfies the δ1 heat level, where δ1 is used to determine an appropriate threshold for the popular interest, and it is judged whether the object's interest has changed by comparing the changes in the object search keywords and the keyword weights in the previous period and the current period; If a certain type of keyword appears in the current period, the interest corresponding to the keyword set of this type is the newly added interest of the user; If a certain type of keyword appeared in the previous period and did not appear in the current period, the interest corresponding to the keyword set of this type fades; Among them, the degree of the user's interest is calculated according to the behavior record information such as the number of views, search times, collections, forwards, and the used market. The duration of the user's interest in the interest is regarded as the degree of his interest. Denote k(x,t) as the degree of the interest k of x at time t.

5. The intelligent information push method based on cloud computing and big data according to claim 4 is characterized in that: The update process of the interest includes: N(k(t)≥a) is introduced to represent the neighborhood of users whose interest in interest k is not less than a at time t. It is assumed that at time t, all interests of user x are K(x,t,α)={k1,k2...k n }, at time t+Δt, all interests of user x are K(x,t+Δt,a)={k1,k2...k m }; By comparing K(x,t+Δt,a) at time t+Δt and K(x,t,α) at ​​time t, we can determine whether x’s interest has changed during the time period Δt. That is, if K(x,t+Δt,a) / K(x,t,α)≠Φ, then there exists k i ∈K(x,t+Δt,a) and Then x adds new interest k i , k i ∈K(x,t+Δt,a), and x has a new neighborhood N(k i (t) ≥ a), otherwise, x has not gained new interest in the period Δt; If K(x,t,a) / K(x,t+Δt,a)≠Φ, then there exists k j ∈K(x,t,a), and Determine whether x is of interest k based on the size of Δt j ∈K(x,t,a) no longer cares about x, if x is interested in k j ∈K(x,t,a) is no longer concerned, then the relationship between x and the objects in the neighborhood corresponding to the interest changes; If K(x,t,α)∩K(x,t+Δt,a)≠Φ, and k j (x,t+Δt)-k j (x,t)>δ2,k j ∈K(x,t,a)∩K(x,t+Δt,a), then x is k j The interest level increases, where δ2 is the degree of increase in interest.

6. The intelligent information push method based on cloud computing and big data according to claim 1 is characterized in that: Perform hotspot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the user's neighborhood chain, including: Preset K as the interest set, X as the user set, x∈X, and within Δt, the interest added by x is: ΔK(x,Δt,a)={k i |i=1,2...n,k i ∈K}; The interests added by all users in X are: ΔK(Δt,a)=∪ x∈X ΔK(x,Δt,a)={k i |i=1,2...m,k i ∈K}; All neighborhoods formed around the new interest are: ΔN(Δt,a)={N(k i (t)≥a)|k i ∈ΔK(Δt,a)}; make Then the interests in the set ΔK(H>δ1) are popular interests that satisfy the δ1 heat level, and the neighborhoods in the set ΔN(H>δ1) are neighborhoods formed around each popular interest; Preset N(k i (t)≥a) represents the interest k i The neighborhood formed, k i ∈ΔK(H>δ1), x∈N(k i (t)≥a), B(x,k i ,Δt) represents the relationship between x and k in Δt time i The search word set, then N(k i (t)≥a) with respect to k i The search terms are: Interest i Contains different sub-interests k ij ,but Where B(k ij ,Δt) represents N(k i (t)≥a) with respect to k ij The search term set is Then ΔK(k i ,H>δ1) means k i The real hot interest in ΔN(k i ,H>δ1) represents the interest k ij All neighborhoods formed.

7. The intelligent information push method based on cloud computing and big data according to claim 1 is characterized in that: It also includes: The social network between users is represented as G = (V, E), where V = {v1, V2...V n Represents a collection of nodes. represents the set of edges between nodes. The edges between nodes represent the interactions between nodes. The Gaussian potential function of the short-range field is used to describe the interactions between nodes. In the social network G = (V, E), v i ∈V, let Among them, m j Yes j The mass, v j ∈V, j = 1, 2...n, d ij Indicates v i and v j , σ represents the influence factor and is used to control the influence range of each node.

8. The intelligent information push method based on cloud computing and big data according to claim 1 is characterized in that: Obtain the trust relationship between users in the relationship neighborhood and establish a trust model. Determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push, including: Assume that N(k(t)≥α) is the neighborhood around interest k, x,y∈N(k(t)≥α), e(t,x,y,k) represents the number of times y recognizes x in the time period [t-Δt,t], let E(t,x,k)=Σ y∈N(k(t)≥α )e(t,x,y,k), E(t,x,k) is the neighborhood energy of x at time t in N(k(t)≥α); The initial trust between users can be expressed as: Among them, E(t,y,k) represents the neighborhood energy of y in N(k(t)≥α) at ​​time t; Assume that l is a neighborhood chain formed by several neighborhoods. R i+1 (t,x i ,x i+1 ) are two adjacent objects x on l at time t i 、x i+1 The energy resistance of i ,x i+1 ∈l, i=0,1...n, D(t,x i ,x i+1 ) is x at time t i With x i+1 The comprehensive difference of the level of is x at time t i The topological potential, F(t,x i ,x i+1 ) is x at time t i+1 x i The trust level of F(t,x i ,x i+1 )≠0; Then let R(t,x0,x n )=R1(t,x0,x1) o R2(t,x1,x2) o ...o Rn (t,x n-1 ,x n ), R(t,x0,x n ) is the connection between x0 and x at time t n The energy resistance of chain l, the "o" operation is calculated according to the following expression: Energy resistance can be used to characterize the energy loss when information is transmitted along the neighborhood chain, R(t,x0,x n ) represents the energy resistance between adjacent users on the neighborhood chain to measure the energy resistance of the neighborhood chain, R'(t,x0,x n ) represents the maximum value of the block between adjacent users on the neighborhood chain, that is, the block that the neighborhood chain can block.

9. An intelligent information push system based on cloud computing and big data according to the intelligent information push method based on cloud computing and big data according to any one of claims 1 to 8, characterized in that: include: An information acquisition unit, used to acquire the user's registration information and input interest tags on the page, and to establish an initial user interest model; The feature extraction unit obtains the user's behavior record information on the page according to the registration information and the user interest model, converts the behavior record information into text information and performs classification processing to obtain the user's interest feature words; A model analysis unit is used to construct a user's relationship neighborhood model based on the user's interest feature words and behavior record information, and perform hot spot analysis and neighborhood authoritative user analysis on the relationship neighborhood model to obtain the user's neighborhood chain; The information push unit is used to obtain the trust relationship between users in the relationship neighborhood and establish a trust model, and determine the best push channel between users based on the neighborhood chain and the trust model to complete intelligent information push.

10. A storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the intelligent information push method based on cloud computing and big data as described in any one of claims 1 to 8 are implemented.

Images