CN113742603B - Object recommendation method, device and system and electronic equipment - Google Patents
Object recommendation method, device and system and electronic equipment Download PDFInfo
- Publication number
- CN113742603B CN113742603B CN202110422812.5A CN202110422812A CN113742603B CN 113742603 B CN113742603 B CN 113742603B CN 202110422812 A CN202110422812 A CN 202110422812A CN 113742603 B CN113742603 B CN 113742603B
- Authority
- CN
- China
- Prior art keywords
- information
- personnel
- person
- close
- relationship
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000004927 fusion Effects 0.000 claims abstract description 73
- 238000004364 calculation method Methods 0.000 claims abstract description 24
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 238000011176 pooling Methods 0.000 claims abstract description 12
- 230000006403 short-term memory Effects 0.000 claims abstract description 6
- 230000007774 longterm Effects 0.000 claims abstract description 3
- 230000015654 memory Effects 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 3
- 239000008280 blood Substances 0.000 claims description 3
- 210000004369 blood Anatomy 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 230000001815 facial effect Effects 0.000 claims description 3
- 230000007787 long-term memory Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 7
- 230000001149 cognitive effect Effects 0.000 abstract description 2
- 230000010365 information processing Effects 0.000 abstract description 2
- 238000005070 sampling Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 238000009877 rendering Methods 0.000 description 6
- 238000007405 data analysis Methods 0.000 description 4
- 238000007499 fusion processing Methods 0.000 description 4
- 230000019771 cognition Effects 0.000 description 3
- 238000012800 visualization Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000000547 structure data Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- KJONHKAYOJNZEC-UHFFFAOYSA-N nitrazepam Chemical compound C12=CC([N+](=O)[O-])=CC=C2NC(=O)CN=C1C1=CC=CC=C1 KJONHKAYOJNZEC-UHFFFAOYSA-N 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Tourism & Hospitality (AREA)
- Mathematical Physics (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the field of network information processing, and relates to an object recommendation method, an object recommendation device and electronic equipment; the method comprises the steps of collecting edge information and relation data of related personnel; fusing multisource edge attributes of related personnel in a variance weight mode; performing multi-granularity fusion expression learning after aligning the edge information and the relation data; information fusion of the two-way long-term short-term memory neural network to the peer relationship is utilized, and information fusion of the average pooling to the upper and lower relationship is utilized; calculating the importance degree of the close personnel contacted with the related personnel; calculating the similarity distance between the related person and the close person according to the low-dimensional vector representation learned by the related person, and recommending a plurality of close persons of the related person according to the similarity distance; the invention fuses the multisource data of related personnel, and removes noise information or clues influencing analysis based on a multiscale cognitive calculation theory; the accuracy of the recommendation result is improved.
Description
Technical Field
The invention belongs to the field of network information processing, and particularly relates to an object recommendation method and device and electronic equipment.
Background
Related personnel have higher and higher difficulty in processing and analyzing by combining various relation information and attribute information along with the influence of big data and informatization of the current society. In particular, in actual situations, the social relationship of a specific object is extremely complex, so that in the mining recommendation of the specific object, the time period is longer and the efficiency is lower.
In most of the existing research works aiming at social network analysis, the social network is regarded as a homogeneous network to be processed, and particularly in the job related network, some details in the person relationship are often ignored, and the recommendation direction is easily led to an incorrect investigation direction, so that the accuracy of a recommendation result is influenced, the time of object recommendation is delayed, and the efficiency of object recommendation is reduced.
Disclosure of Invention
Based on the problems existing in the prior art, the intelligent analysis is performed on the complex network based on the multi-granularity information fusion principle, the edge data and the relation structure data of related personnel are fused, the influence of irrelevant information on object analysis is reduced, the integration efficiency of effective clue information in object recommendation is accelerated, and the recommendation accuracy is improved.
The invention provides an object recommendation method, device and system and electronic equipment for realizing the technical scheme.
In a first aspect of the present invention, the present invention provides an object recommendation method, the recommendation method comprising:
collecting the edge information of related personnel and the relationship data between other close personnel;
the method comprises the steps of fusing multiple edge information of each related person in a variance weight mode to obtain embedded representation of multi-source edge attribute fusion of the related person;
carrying out data alignment processing on the edge information and the relationship data, taking various edge information of related personnel as fine granularity information, taking the relationship data between the related personnel and other close personnel as coarse granularity information, and carrying out multi-granularity fusion expression learning;
according to the frequent degree of contact between related personnel and different types of close personnel, dividing the relationship between the related personnel and the close personnel into a same-level relationship and a superior-subordinate relationship;
information fusion of the two-way long-term short-term memory neural network to the peer relationship is utilized, and information fusion of the average pooling to the upper and lower relationship is utilized;
according to the peer relationship and the upper and lower relationship after information fusion, calculating the importance degree of the close personnel contacted with the related personnel;
calculating a low-dimensional vector representation learned by the related person by using the importance degree of the close person contacted with the related person;
and calculating the similarity distance between the related person and the close person according to the learned low-dimensional vector representation, and recommending a plurality of close person sets of the related person according to the similarity distance.
In a second aspect of the present invention, the present invention further provides an object recommendation apparatus, the apparatus including:
the data acquisition module is used for acquiring the edge information of related personnel and the relationship data between other close personnel;
the data embedding module is used for fusing various edge information of each related person in a variance weight mode to obtain an embedded representation of multi-source edge attribute fusion of the related person;
the data alignment module is used for carrying out data alignment on the edge information and the relation data;
the multi-granularity fusion module is used for taking various edge information of related personnel as fine granularity information and relationship data between the related personnel and other close personnel as coarse granularity information to perform multi-granularity fusion expression learning;
the information fusion module is used for utilizing a two-way long-short-term memory neural network to fuse the information of the same-level relationship and utilizing average pooling to fuse the information of the upper-level relationship and the lower-level relationship;
the importance degree calculating module is used for calculating the importance degree of the close personnel contacted with the related personnel according to the same-level relationship and the upper-lower-level relationship after information fusion;
the low-dimensional vector calculation module is used for calculating a low-dimensional vector representation learned by the related person by utilizing the importance degree of the close person contacted with the related person;
the similarity distance calculation module is used for calculating the similarity distance between related personnel and close personnel according to the learned low-dimensional vector representation;
and the recommending module is used for recommending a plurality of close personnel sets of the related personnel according to the similarity distance.
In a third aspect of the present invention, the present invention provides an object recommendation system, including a data sampling server, a background data computing server, and a data rendering visualization server, which are sequentially connected;
the data sampling server comprises a data acquisition module according to the first aspect of the invention, and the background data calculation server comprises a data alignment module, a data embedding module, a multi-granularity fusion module, an information fusion module, an importance degree calculation module, a low-dimensional vector calculation module, a similarity distance calculation module and a recommendation module according to the first aspect of the invention; the data rendering and visualizing server is used for displaying the object recommendation result calculated by the background data calculation server.
In a fourth aspect of the invention, the invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method according to the first aspect of the invention when executing the program.
The invention has the beneficial effects that:
according to the invention, the multi-source data of the related personnel are fused by utilizing the big data intelligent analysis method, and the edge attribute characteristics of the related personnel are fused in a variance weight mode, so that noise information brought by embedding representation of the related personnel due to attribute drift can be reduced, and the effects of enhancing favorable information and weakening noise information are achieved; noise information or clues in which the object recommendation analysis is affected can be effectively removed. Based on a multi-granularity cognitive computing theory, the invention fuses the edge data and the relation structure data of related personnel, takes various edge information of the related personnel as fine granularity information and takes the relation data between the related personnel and other close personnel as coarse granularity information; characteristic representations of related personnel can be fully mined; the invention also divides the structure in the relation network into two types for classification modeling; the role information of the personnel in the relation network structure can be balanced better, so that the peer relation can interact and influence each other in the interaction level; whereas the upper and lower relationships can interact at the level to which they pertain. In addition, the information features of the different types of people closely related to the related people are fused, and the respective importance degrees of the different types of people closely related to the related people are independently learned through the attention mechanism, so that more potential relation of the people of which types have more potential relation to the related people can be accurately judged, better embedded representation of the related people is learned, further knowledge useful for object recommendation is found, the accuracy of object recommendation results is improved, and objects with higher accuracy are recommended to specific people in a terminal data presentation mode, and guidance information can be provided for the specific people.
Drawings
FIG. 1 is a flow chart of an object recommendation method in an embodiment of the invention;
FIG. 2 is a diagram of an object recommendation device architecture in an embodiment of the present invention;
FIG. 3 is a diagram of an object recommendation system architecture in an embodiment of the present invention;
fig. 4 is a schematic diagram of a related network embodiment in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
FIG. 1 is a flow chart of an object recommendation method in an embodiment of the invention; as shown in fig. 1, the object recommendation method includes:
101. collecting the edge information of related personnel and the relationship data between other close personnel;
the edge information of the related personnel comprises facial features, DNA, finger blood, fingerprints, palmprints, height weight, photos and the like.
The related personnel and other close personnel are subjected to social analysis, wherein the related personnel comprise mobile phone terminal information, judicial terminal information, office work business information, financial flow information and the like.
102. The method comprises the steps of fusing multiple edge information of each related person in a variance weight mode to obtain embedded representation of multi-source edge attribute fusion of the related person;
in order to acquire the fusion feature, the variance weight of the edge attribute of the related person is firstly required to be constructed:
β i ＝Softmax(1-|c i -ε|·|c i -ε| T )
wherein ,βi Variance weights representing the ith edge attribute of the associated person; softmax represents the normalized exponential function; c i A low-dimensional vector representation of the ith edge attribute after the relevant person is encoded, e representing the result after averaging pooling of all edge attributes of the relevant person, which represents the most accurate representation of the attribute; and (c) i - ∈) represents c i Degree of deviation from E, therefore, (c) i - ∈) is larger, beta i The smaller the size; n represents the number of edge attributes.
The nodes obtained after fusion of all edge attributes related to the relevant person are expressed as:
Cont v representing the person v concernedAn embedded representation of multi-source edge attribute fusion; the embedded representation fuses the multi-source edge attribute of the related personnel v, and can fully mine the edge attribute characteristics of the related personnel v.
103. Carrying out data alignment processing on the edge information and the relationship data, taking various edge information of related personnel as fine granularity information, taking the relationship data between the related personnel and other close personnel as coarse granularity information, and carrying out multi-granularity fusion expression learning;
in the embodiment of the invention, the edge data and the relationship data of the related personnel can be aligned according to the serial numbers of the related personnel, namely, the edge information and the social relationship information can be processed in a one-to-one correspondence manner according to the acquired edge information and social relationship information of each person in an id number or renumbering manner; for example, assuming that the number of the relevant person is 1001, the edge data of the number 1001 is aligned with the relationship data of the number 1001, and social relationship structure information between the relevant person and other close persons is also included in the relationship data besides social relationship data of the relevant person, so that the relevant person and the close person can be used as node users, and a relevant relationship network can be constructed according to the relationship.
After the initial data of each person is prepared, the edge data and the related coefficient data are transmitted to a background data calculation server, and according to the data cognition principle, the edge data are regarded as fine granularity information by adopting a multi-granularity fusion method, the relation data are regarded as coarse granularity information, and multi-source data fusion processing is carried out. For fine-granularity data, the invention adopts the idea of variance to carry out mean value processing on all relevant personnel attribute information, and weights are distributed by judging the distance between the result after mean value and each attribute before mean value, the greater the distance is, the greater the possibility of noise is, the smaller the weight is, and vice versa; aiming at coarse-granularity data, the invention divides the whole job related network into a peer relationship and an upper-lower relationship, respectively models and fuses, and finally obtains related personnel vector representation in a low-dimensional space. In addition, the feature value of each person in the processed data carries some potential features of the surrounding environment, and the person with potential association is easier to find in a large network.
And performing multi-granularity fusion representation learning, namely combining basic relationship structure information of related personnel with edge attribute information:
emd v ＝BiLSTM{b v ||cont v }，(v∈V)
wherein ,emdv Representing a multi-granularity fusion of all multi-source edge attribute information and basic structure information features related to related personnel v, b v Basic structural information features representing related personnel v, namely relationship data with other close personnel; biLSTM means performing a two-way long and short term memory neural network process; the i indicates a stitching action.
104. According to the frequent degree of contact between related personnel and different types of close personnel, dividing the relationship between the related personnel and the close personnel into a same-level relationship and a superior-subordinate relationship;
the invention can divide the whole related network relationship into a peer relationship and a superior-inferior relationship according to the frequent degree of the contact of related personnel and different types of people, for example, the peer relationship of colleagues and the superior-inferior relationship of leadership subordinates.
105. Information fusion of the two-way long-short-term memory neural network to the same-level relationship is utilized, and information fusion of the average pooling to the upper-level relationship and the lower-level relationship is utilized;
the information fusion of the same-level relationship by using the two-way long-short-term memory neural network comprises the following steps:
wherein ,information fusion is carried out on information features of close persons of the same class type t with the relevant person v, namely information features belonging to the same class type t in the close persons contacted with the relevant person v are subjected to information fusion; mean { } represents the Mean operation; />Representing that the close person v' of the related person v peer type t is processed by the two-way long-short-term memory neural network; n (N) P (v) A set of close persons representing peer relationships directly connected to the associated person v; emd v' The representation merges all edge attribute information and basic structure information features associated with the affinity v'.
The information fusion of the upper and lower level relations by using the average pooling comprises the following steps:
wherein ,information fusion is carried out on information features of close personnel of the upper and lower class type t of the related personnel v, namely, information features belonging to the upper and lower class type t in the close personnel contacted with the related personnel v are subjected to information fusion; mean { } represents the Mean operation; />Representing that all persons with the upper and lower class type t are subjected to average pooling treatment; n (N) S (v) A close personnel set for representing the upper and lower level relations directly connected with the related personnel v; emd v' The representation merges all edge attribute information and basic structure information features associated with the affinity v'.
In an embodiment of the present invention, in the present invention,the fusion of the close person of type t in all sibling neighbors (close persons) belonging to the relevant person v represents the result, because there are multiple types of close persons in all sibling neighbors, the close persons of the same type are fused individually, and the result after the fusion of the various types can be fused again by using the attention mechanism。
106. According to the peer relationship and the upper and lower relationship after information fusion, calculating the importance degree of the close personnel contacted with the related personnel;
the importance of the close personnel of the related personnel includes:
wherein ,τv,i Representing the importance degree of the ith type to the relevant person v among the adjacent close persons in contact with the relevant person v, σ represents the activation function, α represents the importance degree of the close person to the relevant person, |represents the splicing operation; emd v Representing that all edge attribute information and basic structure information characteristics related to related personnel are fused;information fusion is carried out on information features belonging to the type i in close personnel of related personnel v; o (o) j Represents o v Each element of (a) is a member of the group; o (o) v Representing a set of character types, ++> T v Representing a set of role types throughout the relevant network.
In some preferred embodiments, the importance of the type of closely related people can be learned from the main in a method of attention weight aggregation; therefore, the crowd of which types have more potential relations to related people can be accurately judged, better embedded representation to the related people is learned, knowledge useful for object recommendation is found, and accuracy of object recommendation results is improved.
107. Calculating a low-dimensional vector representation learned by the related person by using the importance degree of the close person contacted with the related person;
the low-dimensional vector representation learned by the relevant person includes:
z v a low-dimensional vector representation learned by the relevant person v; sigma represents an activation function; τ v,v Representing the importance degree of the related personnel v; emd v Representing that all edge attribute information and basic structure information characteristics related to related personnel are fused; τ v,k Representing the importance degree of the kth type on the related personnel v in the adjacent close personnel of the related personnel v; t (T) v Representing a set of role types in the overall relevant network;information fusion of information features belonging to the type i in close personnel contacted with the related personnel v is indicated.
108. And calculating the similarity distance between the related person and the close person according to the learned low-dimensional vector representation, and recommending a plurality of close person sets of the related person according to the similarity distance.
Under the low-dimensional vector representation of the related personnel, the similarity distance between the related personnel and the close personnel can be calculated by adopting various similarity distance calculation modes, for example, the similarity between the related personnel and the other close personnel can be measured by adopting the cosine similarity mode in the embodiment of the invention in the modes of pearson correlation coefficient, jaccard similarity coefficient, manhattan distance and the like:
in actual object recommendation, a processor can deeply investigate social relations and some associated data of each person, and the embedded representation of each related person can be calculated by adopting the related person embedded representation learning method in the job related network. For example, K persons having the smallest similarity distance to the person a are determined within a certain threshold, i.e. the K persons defined as the most closely related to the person a.
Fig. 2 is a diagram of an object recommendation apparatus according to an embodiment of the present invention, as shown in fig. 2, including:
100. the data acquisition module is used for acquiring the edge information of related personnel and the relationship data between other close personnel;
the data acquisition module comprises an edge data sampling module and a relation data sampling module; the edge data sampling module is used for collecting edge information of related personnel, wherein the edge information comprises: facial features, DNA, finger blood, fingerprint, palmprint, height weight, photograph, etc.; and carrying out social relationship analysis on related personnel through a relationship data sampling module, wherein the relationship data comprises: mobile phone terminal information, judicial terminal information, office work business information, financial flow information and the like;
200. the data alignment module is used for carrying out data alignment on the edge information and the relation data;
and the data alignment module performs alignment processing on the edge data and the relation data according to the numbers of related personnel.
300. The multi-granularity fusion module is used for carrying out data fusion in a multi-granularity hierarchical mode;
the multi-granularity fusion module comprises an edge data fusion module and a relation data fusion module; the edge data fusion module fuses edge data; the relational data fusion data module fuses relational data; the multi-granularity level data fusion module.
And according to the data cognition principle, transmitting the aligned data into a multi-granularity-level data fusion module for data fusion processing.
400. The information fusion module is used for utilizing a two-way long-short-term memory neural network to fuse the information of the same-level relationship and utilizing average pooling to fuse the information of the upper-level relationship and the lower-level relationship;
the information fusion module is used as a multi-granularity level data fusion module, utilizes a two-way long-short-term memory neural network to fuse information of the same-level relationship, and utilizes average pooling to fuse information of the upper-level relationship and the lower-level relationship.
500. The importance degree calculating module is used for calculating the importance degree of the close personnel contacted with the related personnel according to the same-level relationship and the upper-lower-level relationship after information fusion;
the importance of the close personnel of the related personnel includes:
wherein ,τv,i Representing the importance degree of the ith type to the relevant person v among the adjacent close persons in contact with the relevant person v, σ represents the activation function, α represents the importance degree of the close person to the relevant person, |represents the splicing operation; emd v Representing that all edge attribute information and basic structure information characteristics related to related personnel are fused;information fusion is carried out on information features belonging to the type i in close personnel of related personnel v; o (o) j Represents o v Each element of (a) is a member of the group; o (o) v Representing a set of character types, ++> T v Representing a set of role types throughout the relevant network.
600. The low-dimensional vector calculation module is used for calculating a low-dimensional vector representation learned by the related person by utilizing the importance degree of the close person contacted with the related person;
the low-dimensional vector representation learned by the relevant person includes:
z v a low-dimensional vector representation learned by the relevant person v; sigma represents an activation function; τ v,v Representing the importance degree of the related personnel v; emd v Representing that all edge attribute information and basic structure information characteristics related to related personnel are fused; τ v,k Representing the importance degree of the kth type on the related personnel v in the adjacent close personnel of the related personnel v; t (T) v Representing a set of role types in the overall relevant network;information fusion is carried out on information features belonging to the type i in close personnel of related personnel v.
700. The similarity distance calculation module is used for calculating the similarity distance between related personnel and close personnel according to the learned low-dimensional vector representation;
the invention can calculate the similarity distance between related personnel and other close personnel by adopting various similarity distance calculation modes, for example, the similarity between related personnel and other close personnel can be calculated by the modes of Pelson correlation coefficient, jaccard similarity coefficient, manhattan distance and the like.
800. And the recommending module is used for recommending a plurality of close personnel sets of the related personnel according to the similarity distance.
And sorting the similarity distances, and selecting a plurality of close people in front as a close person set of related people.
FIG. 3 is a diagram of an object recommendation system architecture in an embodiment of the present invention; as shown in fig. 3, the system comprises a data sampling server, a background data calculation server and a data rendering visualization server which are connected in sequence;
the data sampling server comprises a data acquisition module, and the background data calculation server comprises a data alignment module, a multi-granularity fusion module, an information fusion module, an importance degree calculation module, a low-dimensional vector calculation module, a similarity distance calculation module and a recommendation module; the data rendering and visualizing server is used for displaying the object recommendation result calculated by the background data calculation server.
The data sampling server stores the collected multi-source data information in a classified mode; the background data calculation server performs alignment processing on the edge data and the relation data according to the serial numbers of related personnel by receiving the data transmitted by the data sampling server; according to the data cognition principle, the aligned data are transmitted into a multi-granularity level data fusion module to be subjected to data fusion processing, and finally the fused data are transmitted into a data analysis and representation module; the data analysis and representation module is used for recommending M close personnel with nearest similarity through calculation; the data communication module is used for uploading the data analysis result of related personnel to the data rendering and visualization server; the data rendering and visualizing server is used for visualizing the received data analysis results about related personnel, or inquiring specific related personnel information and multi-source data information of related personnel from the data acquisition server and presenting the information to related specific personnel of the recommendation department.
The embodiment of the invention also provides electronic equipment, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the object recommendation method when executing the program.
Fig. 4 is a schematic diagram of an embodiment of an object recommendation network in the embodiment of the present invention, as shown in fig. 4, after multi-granularity data fusion processing, it may be found that a and B in an original network are not directly connected, but a and B have similar network structures, which indicates that a and B have very similar social relationship networking or have the same vein flow; from the perspective of the relationship structure, a and B are indirectly linked by one person, so by modeling the classification of A, B and man-in-the-middle relationships, respectively, it can be exploited that a and B have some degree of potential relevance. Then, according to similar conditions, K related persons with potential association are screened out and stored; and feeding the K related personnel recommended in the background data calculation server back to terminal equipment of related specific personnel of the recommendation department, wherein the specific personnel refer to the recommendation result.
In the description of the present invention, it should be understood that the terms "coaxial," "bottom," "one end," "top," "middle," "another end," "upper," "one side," "top," "inner," "outer," "front," "center," "two ends," etc. indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the invention.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "configured," "connected," "secured," "rotated," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; either directly or indirectly through intermediaries, or in communication with each other or in interaction with each other, unless explicitly defined otherwise, the meaning of the terms described above in this application will be understood by those of ordinary skill in the art in view of the specific circumstances.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. An object recommendation method, characterized in that the recommendation method comprises:
collecting the edge information of related personnel and the relationship data between other close personnel;
the method comprises the steps of fusing multiple edge information of each related person in a variance weight mode to obtain embedded representation of multi-source edge attribute fusion of the related person;
the embedding of the multisource edge attribute fusion of the related personnel is expressed asWherein Cont v An embedded representation representing a multisource edge attribute fusion of the relevant person v; beta i Variance weight, beta, representing the ith edge attribute of the associated person i ＝Softmax(1-|c i -ε|·|c i -ε| T ) The method comprises the steps of carrying out a first treatment on the surface of the Soft max represents the normalized exponential function; c i A low-dimensional vector representation representing the ith edge attribute of the person concerned after it has been encoded; epsilon represents the result after all edge attributes of related personnel are averaged and pooled; n represents the number of edge attributes;
carrying out data alignment processing on the edge information and the relationship data, taking various edge information of related personnel as fine granularity information, taking the relationship data between the related personnel and other close personnel as coarse granularity information, and carrying out multi-granularity fusion expression learning;
the multi-granularity fusion representation learning includes emd v ＝BiLSTM{b v ||Cont v }, therein emd v Representing that the multi-granularity fuses all multi-source edge attribute information and basic structure information characteristics related to related personnel v; b v Basic structural information features representing related personnel v, namely relationship data with other close personnel; biLSTM means performing a two-way long and short term memory neural network process; the I represents a stitching action;
according to the frequent degree of contact between related personnel and different types of close personnel, dividing the relationship between the related personnel and the close personnel into a same-level relationship and a superior-subordinate relationship;
information fusion of the two-way long-term short-term memory neural network to the peer relationship is utilized, and information fusion of the average pooling to the upper and lower relationship is utilized;
according to the peer relationship and the upper and lower relationship after information fusion, calculating the importance degree of the close personnel contacted with the related personnel;
calculating a low-dimensional vector representation learned by the related person by using the importance degree of the close person contacted with the related person;
and calculating the similarity distance between the related person and the close person according to the learned low-dimensional vector representation, and recommending a plurality of close person sets of the related person according to the similarity distance.
2. The method according to claim 1, wherein the edge information of the related person includes facial features, DNA, finger blood, fingerprint and palmprint, height weight and photo; the relationship data between the related personnel and other close personnel comprises mobile terminal information, judicial terminal information, job passing information and financial flow information.
3. The method of claim 1, wherein the information fusion comprises:
wherein ,information fusion is carried out on information features belonging to the type t in close personnel contacted with related personnel v; mean { } represents the Mean operation; />Representing that the close person v' of the related person v peer type t is processed by the two-way long-short-term memory neural network; n (N) P (v) A set of close persons representing peer relationships directly connected to the associated person v;representing all up and downCarrying out average pooling treatment on close personnel with the class type of t; n (N) S (v) A close personnel set for representing the upper and lower level relations directly connected with the related personnel v; emd v' The representation merges all edge attribute information and basic structure information features associated with the affinity v'.
4. The object recommendation method according to claim 1, wherein the importance level of the person-who is closely related to the person comprises:
wherein ,τv,i Representing the importance degree of the ith type to the relevant person v among the adjacent close persons in contact with the relevant person v, σ represents the activation function, α represents the importance degree of the close person to the relevant person, |represents the splicing operation;information fusion is carried out on information features belonging to the type i in close personnel contacted with related personnel v; o (o) j Represents o v Each element of (a) is a member of the group; o (o) v Representing a set of character types, ++>T v Representing a set of role types throughout the relevant network.
5. An object recommendation method in accordance with claim 1, wherein said low-dimensional vector representation learned by said associated person comprises:
wherein ,zv A low-dimensional vector representation learned by the relevant person v; sigma representsActivating a function; τ v,v Representing the importance degree of the related personnel v; τ v,k Representing the importance degree of the kth type to the relevant person v among the adjacent close persons in contact with the relevant person v; t (T) v Representing a set of role types in the overall relevant network;information fusion of information features belonging to the type i in close personnel contacted with the related personnel v is indicated.
6. An object recommendation device, the device comprising:
the data acquisition module is used for acquiring the edge information of related personnel and the relationship data between other close personnel;
the data embedding module is used for fusing various edge information of each related person in a variance weight mode to obtain an embedded representation of multi-source edge attribute fusion of the related person; the embedding of the multisource edge attribute fusion of the related personnel is expressed asWherein Cont v An embedded representation representing a multisource edge attribute fusion of the relevant person v; beta i Variance weight, beta, representing the ith edge attribute of the associated person i ＝Softmax(1-|c i -ε|·|c i -ε| T ) The method comprises the steps of carrying out a first treatment on the surface of the Soft max represents the normalized exponential function; c i A low-dimensional vector representation representing the ith edge attribute of the person concerned after it has been encoded; epsilon represents the result after all edge attributes of related personnel are averaged and pooled; n represents the number of edge attributes;
the data alignment module is used for carrying out data alignment on the edge information and the relation data;
the multi-granularity fusion module is used for taking various edge information of related personnel as fine granularity information and relationship data between the related personnel and other close personnel as coarse granularity information to perform multi-granularity fusion expression learning; the multiple particle sizesFusion representation learning includes emd v ＝BiLSTM{b v ||Cont v }, therein emd v Representing that the multi-granularity fuses all multi-source edge attribute information and basic structure information characteristics related to related personnel v; b v Basic structural information features representing related personnel v, namely relationship data with other close personnel; biLSTM means performing a two-way long and short term memory neural network process; the I represents a stitching action;
the information fusion module is used for dividing the relationship between related personnel and the close personnel into a same-level relationship and a superior-subordinate relationship according to the frequent degree of the contact between the related personnel and the close personnel of different types, utilizing a two-way long-short-term memory neural network to fuse the information of the same-level relationship, and utilizing average pooling to fuse the information of the superior-subordinate relationship;
the importance degree calculating module is used for calculating the importance degree of the close personnel contacted with the related personnel according to the same-level relationship and the upper-lower-level relationship after information fusion;
the low-dimensional vector calculation module is used for calculating a low-dimensional vector representation learned by the related person by utilizing the importance degree of the close person contacted with the related person;
the similarity distance calculation module is used for calculating the similarity distance between related personnel and close personnel according to the learned low-dimensional vector representation;
and the recommending module is used for recommending a plurality of close personnel sets of the related personnel according to the similarity distance.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 5 when the program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110422812.5A CN113742603B (en) | 2021-04-19 | 2021-04-19 | Object recommendation method, device and system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110422812.5A CN113742603B (en) | 2021-04-19 | 2021-04-19 | Object recommendation method, device and system and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113742603A CN113742603A (en) | 2021-12-03 |
| CN113742603B true CN113742603B (en) | 2023-09-05 |
Family
ID=78728278
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110422812.5A Active CN113742603B (en) | 2021-04-19 | 2021-04-19 | Object recommendation method, device and system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113742603B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103391302A (en) * | 2012-05-08 | 2013-11-13 | 阿里巴巴集团控股有限公司 | Information sending method and system |
| CN107318014A (en) * | 2017-07-25 | 2017-11-03 | 西安电子科技大学 | The video quality evaluation method of view-based access control model marking area and space-time characterisation |
| CN109637225A (en) * | 2018-12-20 | 2019-04-16 | 广东小天才科技有限公司 | A kind of interdynamic studying method and system |
| CN110765350A (en) * | 2019-09-29 | 2020-02-07 | 深圳市云积分科技有限公司 | Data fusion method and device for member points |
| CN111079027A (en) * | 2019-11-29 | 2020-04-28 | 东软集团股份有限公司 | User recommendation method and device, readable storage medium and electronic equipment |
| CN111966787A (en) * | 2020-08-18 | 2020-11-20 | 上海海洋大学 | Intelligent fishery question-answering robot construction method based on knowledge graph |
| CN111985561A (en) * | 2020-08-19 | 2020-11-24 | 安徽蓝杰鑫信息科技有限公司 | Fault diagnosis method and system for intelligent electric meter and electronic device |
| CN112016004A (en) * | 2020-08-21 | 2020-12-01 | 重庆邮电大学 | Multi-granularity information fusion-based job crime screening system and method |
| CN112115367A (en) * | 2020-09-28 | 2020-12-22 | 北京百度网讯科技有限公司 | Information recommendation method, device, equipment and medium based on converged relationship network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050005266A1 (en) * | 1997-05-01 | 2005-01-06 | Datig William E. | Method of and apparatus for realizing synthetic knowledge processes in devices for useful applications |
| US9773272B2 (en) * | 2014-11-10 | 2017-09-26 | 0934781 B.C. Ltd. | Recommendation engine |
-
2021
- 2021-04-19 CN CN202110422812.5A patent/CN113742603B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103391302A (en) * | 2012-05-08 | 2013-11-13 | 阿里巴巴集团控股有限公司 | Information sending method and system |
| CN107318014A (en) * | 2017-07-25 | 2017-11-03 | 西安电子科技大学 | The video quality evaluation method of view-based access control model marking area and space-time characterisation |
| CN109637225A (en) * | 2018-12-20 | 2019-04-16 | 广东小天才科技有限公司 | A kind of interdynamic studying method and system |
| CN110765350A (en) * | 2019-09-29 | 2020-02-07 | 深圳市云积分科技有限公司 | Data fusion method and device for member points |
| CN111079027A (en) * | 2019-11-29 | 2020-04-28 | 东软集团股份有限公司 | User recommendation method and device, readable storage medium and electronic equipment |
| CN111966787A (en) * | 2020-08-18 | 2020-11-20 | 上海海洋大学 | Intelligent fishery question-answering robot construction method based on knowledge graph |
| CN111985561A (en) * | 2020-08-19 | 2020-11-24 | 安徽蓝杰鑫信息科技有限公司 | Fault diagnosis method and system for intelligent electric meter and electronic device |
| CN112016004A (en) * | 2020-08-21 | 2020-12-01 | 重庆邮电大学 | Multi-granularity information fusion-based job crime screening system and method |
| CN112115367A (en) * | 2020-09-28 | 2020-12-22 | 北京百度网讯科技有限公司 | Information recommendation method, device, equipment and medium based on converged relationship network |
Non-Patent Citations (1)
| Title |
|---|
| "基于用户兴趣－ 标签的混合推荐方法研究";李兴华 等;《情报学报》;第466-470页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113742603A (en) | 2021-12-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Balaanand et al. | An enhanced graph-based semi-supervised learning algorithm to detect fake users on Twitter | |
| Zhou et al. | Structure based user identification across social networks | |
| Sun et al. | Near real-time twitter spam detection with machine learning techniques | |
| CN110188208A (en) | A kind of the information resources inquiry recommended method and system of knowledge based map | |
| Jiang et al. | Fema: flexible evolutionary multi-faceted analysis for dynamic behavioral pattern discovery | |
| Kumar et al. | Rough set-based meta-heuristic clustering approach for social e-learning systems | |
| Chen et al. | Home location profiling for users in social media | |
| Govindasamy et al. | Analysis of student academic performance using clustering techniques | |
| Qiao et al. | Siamese neural networks for user identity linkage through web browsing | |
| Gao | Stability analysis of rock slope based on an abstraction ant colony clustering algorithm | |
| Khan et al. | DVAEGMM: Dual variational autoencoder with gaussian mixture model for anomaly detection on attributed networks | |
| Bashabsheh et al. | Big data analysis using hybrid meta-heuristic optimization algorithm and MapReduce framework | |
| Wu et al. | Topological machine learning for multivariate time series | |
| JP7092194B2 (en) | Information processing equipment, judgment method, and program | |
| Liu et al. | Digger: Detect similar groups in heterogeneous social networks | |
| Manasa et al. | Tweet Spam Detection Using Machine Learning and Swarm Optimization Techniques | |
| CN113742603B (en) | Object recommendation method, device and system and electronic equipment | |
| CN111538859B (en) | Method and device for dynamically updating video tag and electronic equipment | |
| Wahid et al. | Human Micro-Expression: A Novel Social Behavioral Biometric for Person Identification | |
| Reddy et al. | A new clustering algorithm based on Voronoi diagram | |
| Maheswari et al. | Kernelized Spectral Clustering based Conditional MapReduce function with big data | |
| Washha et al. | Information quality in social networks: Predicting spammy naming patterns for retrieving twitter spam accounts | |
| CN111159543B (en) | Personalized tourist place recommendation method based on multi-level visual similarity | |
| Saxena et al. | Evolving efficient clustering patterns in liver patient data through data mining techniques | |
| CN115964478A (en) | Network attack detection method, model training method and device, equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |