An information fusion management method based on an internet of things

Abstract

The application discloses an information fusion management method based on an Internet of Things, and relates to the technical field of information management, and comprises the following steps: S1, obtaining historical information of a user from a cloud platform, and constructing a knowledge graph according to the historical information of the user; S2, when the latest information of the user is generated on the cloud platform, generating a plurality of local characteristic coefficients, and based on the local characteristic coefficients, fusing the latest information and the historical information by using the knowledge graph to obtain a real-time portrait of the user; and S3, encrypting the real-time portrait of the user and uploading the real-time portrait to the cloud platform. The application can strengthen information fusion, solve the fragmentation defects of the existing information fusion technology, and generate a user interest portrait with high precision and high dynamics.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information management, and particularly relates to an information fusion management method based on Internet of Things. BACKGROUND

[0002] In the Internet of Things scenario, user browsing information (such as smart device browsing records and platform interaction content) has the characteristics of dynamic high-frequency update. However, in the fusion management of user browsing information, the prior art has the following key deficiencies: first, the local feature mining is insufficient. When the prior method constructs the knowledge graph of historical browsing information, only the global topology (such as the overall connection of the browsing relationship) is focused on, and the real-time information generated is not targeted, so that when the real-time browsing information (such as new browsing content) is associated with the historical graph, the dynamic continuity of user interest (such as the evolution of short-term interest to long-term preference) cannot be accurately captured. Second, the traditional information management method cannot efficiently fuse historical and real-time information, accurately depict the real-time state of the user, and deeply utilize the knowledge graph, resulting in poor fusion effect. Therefore, there is an urgent need for a complex information management method based on Internet of Things. SUMMARY

[0003] The present application is to solve the above problems, and proposes an information fusion management method based on Internet of Things.

[0004] The technical scheme of the present application is: an information fusion management method based on Internet of Things includes the following steps:

[0005] S1, obtaining the historical information of the user from the cloud platform, and constructing a knowledge graph according to the historical information of the user;

[0006] S2, when the latest information of the user is generated on the cloud platform, generating a plurality of local feature coefficients, and based on the local feature coefficients, using the knowledge graph to fuse the latest information and the historical information to obtain the real-time portrait of the user;

[0007] S3, encrypting the real-time portrait of the user and uploading it to the cloud platform.

[0008] Further, S2 includes the following sub-steps:

[0009] S21, extracting the global Laplacian matrix of the knowledge graph;

[0010] S22, calculating a plurality of local feature coefficients for the knowledge graph according to the global Laplacian matrix of the knowledge graph;

[0011] S23, when the latest information of the user is generated on the cloud platform, obtaining the real-time portrait of the user according to the plurality of local feature coefficients of the knowledge graph.

[0012] The beneficial effect of the further scheme is that in the application, the global Laplacian matrix is extracted to depict the global connectivity of historical information; the local feature coefficients are calculated based on the global matrix, the knowledge graph is associated with real-time browsing information through the local feature coefficients, and the portrait is generated. The application overcomes the limitation of the existing knowledge graph which only depends on the global topology, mines the fusion degree of historical information and real-time information, improves the fusion accuracy of real-time information and historical graph, and makes the portrait more consistent with user interest.

[0013] Further, S22 comprises the following sub-steps:

[0014] S221, split the global Laplacian matrix into a left singular vector matrix and a right singular vector matrix;

[0015] S222, extract a plurality of left singular vectors of the left singular vector matrix and a plurality of right singular vectors of the right singular vector matrix;

[0016] S223, calculate a plurality of local feature coefficients for the knowledge graph according to the left singular vector and the corresponding right singular vector.

[0017] The beneficial effect of the further scheme is that in the application, the global Laplacian matrix is split into left and right singular vector matrices by singular value decomposition, the core feature direction of the graph is decomposed, the singular vectors are extracted, and the norm and the extreme value of the elements of the vector are used for quantization. By using the dimension reduction and feature extraction capability of SVD, the complex graph matrix is converted into vector features to generate local feature coefficients participating in portrait updating.

[0018] Further, in S223, the local feature coefficient of the knowledge graph is expressed as:

[0019] ;

[0020] In the formula, L2 norm of the left singular vector, L2 norm of the corresponding right singular vector, maximum value of all elements in the left singular vector, average value of all elements of the left singular vector and the corresponding right singular vector, maximum value of all elements of the left singular vector and the corresponding right singular vector.

[0021] Further, S23 comprises the following sub-steps:

[0022] S231, when the latest information of the user is generated on the cloud platform, extract the keywords in the latest information as the respective to-be-inserted entities;

[0023] S232, acquire a plurality of entity pairs connected by a relationship edge in the knowledge graph;

[0024] S233, construct a knowledge graph update rule, and determine the best relationship edge corresponding to each to-be-inserted entity pair by using the knowledge graph update rule and the plurality of entity pairs;

[0025] S234, calculate the fusion weight of the latest information according to the best relationship edge corresponding to each to-be-inserted entity pair and a plurality of local feature coefficients of the knowledge graph;

[0026] S235, obtain the real-time portrait of the user according to the fusion weight of the latest information.

[0027] The beneficial effects of the above further scheme are: in the present application, the keywords (to-be-inserted entities) of real-time browsing information are extracted, the existing entity pairs in the knowledge graph are associated, the knowledge graph update rule is constructed, the best relationship edge of the to-be-inserted entity is screened, the local feature coefficients are combined, and the fusion weight of real-time information (balance the contribution of historical interest and real-time behavior) is calculated. The above operation can determine the association logic of real-time browsing information and the knowledge graph, and solve the problem of how to accurately integrate new browsing information into the historical interest graph through the best relationship edge screening (relationship type that fits the browsing behavior) and weight calculation, thereby improving the real-time and rationality of the portrait.

[0028] Further, in S233, the expression of the knowledge graph update rule is:

[0029] ;

[0030] In the formula, denotes the to-be-inserted entity, denotes one of the entities of the to-be-inserted entity inserted into the relationship edge, denotes the other entity of the to-be-inserted entity inserted into the relationship edge, denotes the weight of the relationship edge after the to-be-inserted entity is inserted into the relationship edge, denotes the weight of the entity in the knowledge graph, denotes the maximum value.

[0031] Further, in S234, the expression of the fusion weight of the latest information is:

[0032]

[0033] In the formula, denotes the i-th local feature coefficient in the knowledge graph, denotes the number of all local feature coefficients, denotes the number of all local feature coefficients, ​Indicates the entity to be inserted is inserted at the first... Relationship edge The weight of the relation edges after the relationship. This indicates the number of relation edges of the inserted entity.

[0034] The beneficial effects of the above-mentioned further solution are as follows: In this invention, the mean value of the local feature coefficients of the initial knowledge graph is used to further correct the mean value of the relation edge weights of the updated knowledge graph, thereby determining the fusion weight of the latest information, taking into account both the local patterns of the initial graph and the intensity characteristics of real-time behavior.

[0035] Furthermore, in S235, the difference between the fusion weight of 1 and the fusion weight of the latest information is used as the fusion weight of the historical information. Based on the fusion weight of the historical information and the fusion weight of the latest information, a Bayesian network is used to fuse the user's historical information and the latest information to obtain the user's real-time profile.

[0036] The beneficial effects of this invention are as follows: This invention constructs a global topology of historical browsing relationships through a knowledge graph, and then calculates local feature coefficients using singular value decomposition to accurately determine the fusion ratio of user historical information and latest information; it integrates local feature coefficients (historical interest structure) and real-time relationship edge weights (current behavior intensity) to balance the contributions of historical browsing and real-time browsing, breaking through the limitations of single-dimensional weighting; this invention can strengthen information fusion, solve the fragmentation defects of existing information fusion technologies, and generate user interest profiles with high accuracy and high dynamics. Attached Figure Description

[0037] Figure 1 This is a flowchart of an information fusion management method based on the Internet of Things. Detailed Implementation

[0038] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0039] like Figure 1 As shown, the present invention provides an information fusion management method based on the Internet of Things, comprising the following steps:

[0040] S1. Obtain users' historical information from the cloud platform and construct a knowledge graph based on the users' historical information;

[0041] S2. When the latest information of a user is generated on the cloud platform, several local feature coefficients are generated, and based on the local feature coefficients, the latest information and historical information are fused using a knowledge graph to obtain a real-time profile of the user.

[0042] S3. Encrypt the user's real-time profile and upload it to the cloud platform.

[0043] In this embodiment of the invention, S2 includes the following sub-steps:

[0044] S21. Extract the global Laplacian matrix of the knowledge graph;

[0045] S22. Based on the global Laplacian matrix of the knowledge graph, calculate several local feature coefficients for the knowledge graph;

[0046] S23. When the latest information about a user is generated on the cloud platform, a real-time profile of the user is obtained based on several local feature coefficients of the knowledge graph.

[0047] In this invention, a global Laplacian matrix is ​​extracted to characterize the global connectivity of historical information. Local feature coefficients are calculated based on the global matrix, and these coefficients are used to link the knowledge graph with real-time browsing information to generate a profile. This invention overcomes the limitations of existing knowledge graphs that rely solely on global topology, explores the fusion degree of historical and real-time information, improves the fusion accuracy of real-time information and historical graphs, and makes the profile more aligned with user interests.

[0048] In this embodiment of the invention, S22 includes the following sub-steps:

[0049] S221. Split the global Laplacian matrix into a left singular vector matrix and a right singular vector matrix;

[0050] S222. Extract several left singular vectors from the left singular vector matrix and several right singular vectors from the right singular vector matrix;

[0051] S223. Calculate several local feature coefficients for the knowledge graph based on the left singular vector and the corresponding right singular vector.

[0052] In this invention, singular value decomposition (SVD) is used to split the global Laplacian matrix into left and right singular vector matrices, decomposing the core feature directions of the graph and extracting singular vectors. These vectors are then quantified using their norm and the extreme values ​​of their elements. SVD's dimensionality reduction and feature extraction capabilities are leveraged to transform the complex graph matrix into vector features, generating local feature coefficients that participate in image updating.

[0053] In this embodiment of the invention, in S223, the local feature coefficients of the knowledge graph The expression is:

[0054] ;

[0055] In the formula, Denotes the L2 norm of the left singular vector. This represents the L2 norm of the corresponding right singular vector. This represents the maximum value of all elements in the left singular vector. Let represent the average of all elements of the left singular vector and the corresponding right singular vector. This represents the maximum value of all elements of the left singular vector and its corresponding right singular vector.

[0056] In this embodiment of the invention, S23 includes the following sub-steps:

[0057] S231. When the latest information about a user is generated on the cloud platform, extract the keywords from the latest information and use them as entities to be inserted.

[0058] S232. Obtain several entity pairs connected by relational edges in the knowledge graph;

[0059] S233. Construct knowledge graph update rules, and use the knowledge graph update rules and several entity pairs to determine the best relation edge corresponding to the entity to be inserted.

[0060] S234. Calculate the fusion weight of the latest information based on the best relation edges corresponding to each entity to be inserted and several local feature coefficients of the knowledge graph.

[0061] S235. Based on the fusion weight of the latest information, obtain the user's real-time profile.

[0062] In this invention, keywords (entities to be inserted) of real-time browsing information are extracted, associated with existing entity pairs in the knowledge graph, knowledge graph update rules are constructed, the best relation edges for the entities to be inserted are selected, and the fusion weight of real-time information is calculated by combining local feature coefficients (balancing the contributions of historical interests and real-time behavior). The above operations can determine the association logic between real-time browsing information and the knowledge graph. By selecting the best relation edges (relationship types that fit browsing behavior) and calculating weights, the problem of how to accurately integrate new browsing information into the historical interest graph is solved, improving the real-time performance and rationality of the profile.

[0063] In this embodiment of the invention, in S233, the knowledge graph update rules... The expression is:

[0064] ;

[0065] In the formula, Indicates the entity to be inserted. Indicates the entity to be inserted Insert one of the entities in the relation edge. Indicates the entity to be inserted Insert another entity with the relation edge. Indicates the entity to be inserted The weight of the relation edge after insertion. This represents the weight of an entity in the knowledge graph. This indicates taking the maximum value.

[0066] In this embodiment of the invention, in S234, the expression for the fusion weight of the latest information is:

[0067]

[0068] In the formula, Represents the first in a knowledge graph Local characteristic coefficients, This represents the number of all local characteristic coefficients. Indicates the entity to be inserted is inserted at the first... Relationship edge The weight of the relation edges after the relationship. This indicates the number of relation edges of the inserted entity.

[0069] In this invention, the mean value of the local feature coefficients of the initial knowledge graph is used to further correct the mean value of the relation edge weights of the updated knowledge graph, thereby determining the fusion weight of the latest information, taking into account both the local patterns of the initial graph and the intensity characteristics of real-time behavior.

[0070] In this embodiment of the invention, in S235, the difference between the fusion weight of 1 and the fusion weight of the latest information is used as the fusion weight of the historical information. Based on the fusion weight of the historical information and the fusion weight of the latest information, a Bayesian network is used to fuse the user's historical information and the latest information to obtain the user's real-time profile.

[0071] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. An information fusion management method based on the Internet of Things, characterized in that, Includes the following steps: S1. Obtain users' historical information from the cloud platform and construct a knowledge graph based on the users' historical information; S2. When the latest information of a user is generated on the cloud platform, several local feature coefficients are generated, and based on the local feature coefficients, the latest information and historical information are fused using a knowledge graph to obtain a real-time profile of the user. S3. Encrypt the user's real-time profile and upload it to the cloud platform; S2 includes the following sub-steps: S21. Extract the global Laplacian matrix of the knowledge graph; S22. Based on the global Laplacian matrix of the knowledge graph, calculate several local feature coefficients for the knowledge graph; S23. When the latest information about a user is generated on the cloud platform, a real-time profile of the user is obtained based on several local feature coefficients of the knowledge graph. S22 includes the following sub-steps: S221. Split the global Laplacian matrix into a left singular vector matrix and a right singular vector matrix; S222. Extract several left singular vectors from the left singular vector matrix and several right singular vectors from the right singular vector matrix; S223. Calculate several local feature coefficients for the knowledge graph based on the left singular vector and the corresponding right singular vector. In S223, the local feature coefficients of the knowledge graph The expression is: ； In the formula, Denotes the L2 norm of the left singular vector. This represents the L2 norm of the corresponding right singular vector. This represents the maximum value of all elements in the left singular vector. Let represent the average of all elements of the left singular vector and the corresponding right singular vector. This represents the maximum value of all elements of the left singular vector and its corresponding right singular vector; S23 includes the following sub-steps: S231. When the latest information about a user is generated on the cloud platform, extract the keywords from the latest information and use them as entities to be inserted. S232. Obtain several entity pairs connected by relational edges in the knowledge graph; S233. Construct knowledge graph update rules, and use the knowledge graph update rules and several entity pairs to determine the best relation edge corresponding to the entity to be inserted. S234. Calculate the fusion weight of the latest information based on the best relation edges corresponding to each entity to be inserted and several local feature coefficients of the knowledge graph. S235. Based on the fusion weight of the latest information, obtain the user's real-time profile; In step S235, the difference between the fusion weight of 1 and the fusion weight of the latest information is used as the fusion weight of the historical information. Based on the fusion weight of the historical information and the fusion weight of the latest information, a Bayesian network is used to fuse the user's historical information and the latest information to obtain the user's real-time profile.

2. The information fusion management method based on the Internet of Things according to claim 1, characterized in that, In S233, the knowledge graph update rules The expression is: ； In the formula, Indicates the entity to be inserted. Indicates the entity to be inserted Insert one of the entities in the relation edge. Indicates the entity to be inserted Insert another entity with the relation edge. Indicates the entity to be inserted The weight of the relation edge after insertion. This represents the weight of an entity in the knowledge graph. This indicates taking the maximum value.

3. The information fusion management method based on the Internet of Things according to claim 1, characterized in that, In step S234, the expression for the fusion weight of the latest information is: ； In the formula, Represents the first in a knowledge graph Local characteristic coefficients, This represents the number of all local characteristic coefficients. Indicates the entity to be inserted is inserted at the first... Relationship edge The weight of the relation edges after the relationship. This indicates the number of relation edges of the inserted entity.

Images