CN117370650A - Cloud computing data recommendation method based on service combination hypergraph convolutional network - Google Patents
Cloud computing data recommendation method based on service combination hypergraph convolutional network Download PDFInfo
- Publication number
- CN117370650A CN117370650A CN202311271535.8A CN202311271535A CN117370650A CN 117370650 A CN117370650 A CN 117370650A CN 202311271535 A CN202311271535 A CN 202311271535A CN 117370650 A CN117370650 A CN 117370650A
- Authority
- CN
- China
- Prior art keywords
- service
- hypergraph
- api
- combination
- matrix
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 87
- 239000013598 vector Substances 0.000 claims abstract description 57
- 238000011176 pooling Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000004364 calculation method Methods 0.000 claims abstract description 11
- 239000011159 matrix material Substances 0.000 claims description 92
- 230000006870 function Effects 0.000 claims description 14
- 238000013135 deep learning Methods 0.000 claims description 11
- 238000005070 sampling Methods 0.000 claims description 10
- 238000005065 mining Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 230000009191 jumping Effects 0.000 claims description 6
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012733 comparative method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 238000012360 testing method Methods 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
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Pure & Applied Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Algebra (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A cloud computing data recommendation method based on service combination hypergraph convolutional network excavates potential service combination relation in cloud computing data and constructs a sequence combination set; constructing a service combination hypergraph based on the sequence combination set, and realizing effective modeling of the combination characteristics of the API service; according to the thought of Chebyshev approximate convolution, a hypergraph convolution network is designed to extract hypergraph signals on a service combination hypergraph; then, performing dimension reduction treatment on the hypergraph signal by using an Hg-Pool pooling method; carrying out semantic coding on the API service by utilizing a pre-training language model to obtain a semantic embedded vector, and fusing the semantic embedded vector and a hypergraph signal to obtain a combined embedded vector; and finally, calculating the recommendation probability of the API service by utilizing the combined embedded vector and the hypergraph signal to obtain a recommendation result. The method has higher association degree, reduces the calculation complexity and the risk of overfitting, and improves the accuracy of the recommended result.
Description
Technical Field
The invention relates to a cloud computing data recommendation method based on a service combination hypergraph convolutional network.
Background
The development of fusion of service computing, distributed computing and containerization technology has driven the advent of the cloud-grown era. As an underlying implementation of the core of the cloud native architecture, cloud computing provides a flexible, unified infrastructure service. Cloud computing may enable on-demand, flexible retrieval of resources (e.g., storage, computing power, networks, services, data, etc.) from a shared pool of configurable computing resources. Cloud computing is becoming a strategic focus of information technology industry development, enterprises and organizations disputely deploy business processes such as product management, personnel systems, financial management and the like to the cloud under the promotion of cloud computing trend, and massive cloud computing data is generated. The cloud computing data is deeply mined by utilizing a reasonable algorithm and a model, and the acquired information and knowledge are widely applied to the fields of service computing, recommendation systems, financial analysis, engineering design and the like.
In the field of recommendation systems, analysis and mining of cloud computing data are beneficial to realizing personalized recommendation and improving relevance and accuracy of recommendation results. For example, patent number 201911411071.X, named as a science popularization content personalized recommendation system based on cloud computing, applies cloud computing data to personalized recommendation of science popularization content through cooperation of a search playing module, a history browsing information recording module and an automatic filling module. The patent number is 201710742043.0, the name is that an intelligent service recommendation method based on cloud computing utilizes a logistic regression algorithm to conduct accurate service selection decision, and the technical problems of large calculated amount and low efficiency of the existing service recommendation algorithm are solved. The patent number is 201510072895.4, and the method and the device for recommending the cold start project based on cloud computing calculate the probability of containing cold start data in the project liked by the user according to the scoring information and the Bayesian algorithm, so that the recommending accuracy of the cold start project is improved.
In the application scene of service calculation, the cloud calculation data contains rich service function, service call and service combination information, and has higher analysis value. The service combination means that two or more API services are combined and mixed to construct the API service with new functions. The service composition relationship laterally reflects the relevance and complementarity between API services. If the two API services contain a plurality of service combination relations, the two API services have stronger relevance. Through reasonable modeling of the service combination relationship and application of the service combination relationship in a recommendation system, the relevance and accuracy of recommendation results are improved.
Disclosure of Invention
In order to overcome the defects of low relevancy, low accuracy and the like of a recommendation result in the conventional cloud computing data recommendation method, the invention provides a service combination hypergraph convolutional network-based cloud computing data recommendation method which has the advantages of higher relevancy, reduced computation complexity and overfitting risk and improved accuracy of the recommendation result, and the method comprises the steps of firstly, carrying out deep mining on cloud computing data to obtain a sequence combination set for representing a potential service combination relation; secondly, constructing a service combination hypergraph by utilizing the sequence combination set, and realizing effective modeling of combination features; then, based on chebyshev approximate convolution thought, designing a hypergraph convolution network to extract hypergraph signals on the service combination hypergraph; considering that the hypergraph convolution operation has higher overfitting risk, reducing the characteristic dimension of the hypergraph signal by using an Hg-Pool (Hypergraph Pool) hypergraph pooling method; and finally, fusing the semantic coding result of the API service and the hypergraph signal after dimension reduction, calculating the recommendation probability of the API service and recommending.
The technical scheme adopted by the invention is as follows:
a cloud computing data recommendation method based on a service combination hypergraph convolutional network comprises the following steps:
step one: mining potential service combination relations in cloud computing data, and constructing a sequence combination set;
step two: constructing a service combination hypergraph based on the sequence combination set in the step one, realizing effective modeling of the combination characteristics of the API service, wherein the hypergraph is a special graph data structure, and the edges of the hypergraph can be connected with any number of nodes;
step three: according to the thought of Chebyshev approximate convolution, a hypergraph convolution network is designed to extract hypergraph signals on a service combination hypergraph; then, performing dimension reduction treatment on the hypergraph signal by using an Hg-Pool pooling method;
step four: carrying out semantic coding on the API service by utilizing a pre-training language model to obtain a semantic embedded vector, fusing the semantic embedded vector and a hypergraph signal to obtain a combined embedded vector, and finally calculating the recommendation probability of the API service by utilizing the combined embedded vector and the hypergraph signal to obtain a recommendation result; the pre-training language model is a natural language processing method in deep learning and has a word vector coding function.
Further, the procedure of the first step is as follows:
1.1 cloud computing data: in a cloud computing environment, call data between a user and an API service comprises the API service, a service call sequence and a service combination;
1.2API service: the application program interface (ApplicationProgrammingInterface, API) is denoted with the symbol a;
1.3 service call sequence: sequencing API service call records of users according to time sequence, wherein the sequencing result is called a service call sequence and is represented by a symbol S;
1.4 service composition: combining two or more API services, mixing and lapping to construct an API service with a new function, which is represented by a symbol mu;
1.5 mining service combination relations, constructing a sequence combination set: traversing the service call sequence, and extracting the API service according to the service combination relation. The API service set with service combination relation is extracted from the service call sequence and is defined as sequence combination item, which is represented by symbol M, and the set formed by all sequence combination items is defined as sequence combination set, which is represented by symbol M.
Preferably, in 1.5, the process of constructing the sequence combination set is as follows:
1.5.1 defining a sequence combination set M and initializing to be an empty set;
1.5.2 traversing all service combinations in cloud computing data, recording the service call record fetched the ith time as mu i ；
1.5.3 defining a sequence combination item m and initializing to an empty set;
1.5.4 traversing mu i The API service in the (k) th acquisition is denoted as a) k ；
1.5.5 traversing service call sequence S, and recording the service call record fetched by the jth time as S j ；
1.5.6 sign a for API service in service call record j A representation;
1.5.7 if a j And a k Equal, then a k Incorporating the sequence combination term m; otherwise, jumping to the step 1.5.5;
1.5.8 when s j When the last service call record in the service call sequence is recorded, finishing the traversal;
1.5.9 when a k Is mu i Ending the traversal when the last API service in the (a) is provided;
1.5.10 incorporating M into M;
1.5.11 when mu i When the last service in the cloud computing data is combined, finishing the traversal;
1.5.12 outputs a sequence combination set M.
Still further, the procedure of the second step is as follows:
2.1 building a super point set of a service composition super graph: the service composition and service call sequence are aggregated,
taking the intersection of the two, converting the intersection into a superpoint in the service combination supergraph, and adding the superpoint into the superpoint set;
the super point set construction process of the service combination super graph is as follows:
2.1.1 creating a super point set V, and initializing the super point set V into an empty set;
2.1.2 traversing all service combinations in the cloud computing data, recording the service call record fetched the ith time as mu i ；
2.1.3 mu i All API services in (1) are marked as a set mSet i ；
2.1.4 record all API services in the service call sequence as set sSet:
2.1.5 pairs sSet and mSet i Performing set operation, taking intersection of the two, and using symbol nSet i A representation;
2.1.6 traversing nSet i The j-th API service in the collection is marked as a j ；
2.1.7 if there is no superpoint V with subscript j in the superpoint set V j Then create superpoint v j Adding it to V;
2.1.8 if V is present in V j Then jump to step 2.1.6;
2.1.9 when a j Is nSet i Ending the traversal when the last API service in the (a) is provided;
2.1.10 when mu i And when the last service in the cloud computing data is combined, finishing the creation of the super point set and finishing the traversal.
2.2 building a hyperedge set of a service composition hypergraph: creating a superpoint set, defining a connection relation between the description superpoint of the incidence matrix and the superedge, traversing the sequence combination set and the superpoint set, and constructing the connection relation according to the inclusion relation of the API service corresponding to the sequence combination item and the superpoint; the creation process of the hyperedge set of the service composition hypergraph is as follows:
2.2.1 creating a superside set E and initializing the superside set E as an empty set;
2.2.2 defining an association matrix H for describing the connection relation between the superpoints and the supersides, and initializing the association matrix H into a null matrix;
2.2.3 traversing sequence combination set M, and taking out sequence combination item using symbol M k A representation;
2.2.4 creation of hyperedge e k ；
2.2.5 traversing the super-point set V, and marking the super-point fetched by the ith time as V j ；
2.2.6 find v j Corresponding API service a j ；
2.2.7 if the sequence combination term m k Contains a j Will exceed edge e k And superpoint v j Connecting, otherwise, jumping to the step 2.2.5;
2.2.8 in the association matrix H, assigning the j-th row and k-th column matrix elements to be 1;
2.2.9 when v j When the point is the last super point in V, finishing the traversal;
2.2.10 if e k At least two different superpoints are connected, e k Incorporate E, otherwise delete E k ；
2.2.11 when m k And when the last item in the M sequence combination items is the last item, finishing the creation of the superside set and finishing the traversal.
2.3 constructing a service composition hypergraph and rewriting the hypergraph into a matrix form: the service composition hypergraph is modeling of service call relation, and the structure of the service composition hypergraph can be represented by a symbol G= (V, E), wherein G represents the service composition hypergraph, V represents a superpoint set, E represents a superedge set, and the service composition hypergraph is rewritten into a matrix form for extracting hypergraph signals; the process of rewriting the hypergraph into matrix form is as follows:
2.3.1 definition of the hyperedge matrix D e The method is used for recording the connection relation between any superside and different superpoints;
2.3.2 pair D e Initializing, D e The number of rows and columns are the number of superedges in E, traversing D e The diagonal element of row j and column j is assigned e j The number of all the connected super points, and the other matrix elements are assigned to 0;
2.3.3 definition of adjacency matrix as matrix form of service composition hypergraph, its expression is θ=h T ·D e H, wherein θ represents the adjacency momentMatrix H is the associated matrix defined in 2.2.2, "T" is the transposed symbol of the matrix, "" is the multiplication symbol of the matrix, "=" is the assignment symbol of the matrix.
Still further, the procedure of the third step is as follows:
3.1 approximate convolution on service composition hypergraph: firstly, defining a Laplacian matrix of a hypergraph and a hypergraph convolution process, and then, according to the Chebyshev approximate convolution thought, performing expansion solution on the hypergraph convolution process to obtain a hypergraph signal, wherein the extraction process of the hypergraph signal is as follows:
3.1.1 defining a Laplacian matrix of the service composition hypergraph, denoted by the symbol L;
3.1.2 initializing a laplace matrix, which expression is l=i- θ. Wherein I is an identity matrix, and the symbol "-" represents matrix subtraction;
3.1.3 defining a convolution kernel of the hypergraph convolution network, denoted by the symbol ρ, the hypergraph convolution kernel being essentially a filter in deep learning, the mathematical form of which is a two-dimensional matrix;
3.1.4 defining a hypergraph signal on the service composition hypergraph, which is represented by a symbol x and is a vector representation of the service composition relationship;
3.1.5 defining a hypergraph convolution operation on the service composition hypergraph, denoted by a symbol, the whole hypergraph convolution process being definable as x p;
3.1.6 according to Chebyshev approximate convolution idea, the hypergraph convolution process defined in step 3.1.5 is solved by using the chebyshev polynomial expansion, and the expression is thatIn the formula, sigma represents summation operation, the upper index N represents approximate maximum order, the lower index N represents order of summation, mu n Is the n-order chebyshev polynomial coefficient,/>Is the scaled n-th order polynomial of the Laplace matrix L, the symbol "≡" representing approximately equal;
3.1.7, a first-order chebyshev convolution operator is selected for solving, and the hypergraph signal extraction process can be further rewritten into x, rho, and tau, theta, and x. Wherein, tau is the parameter of the convolution kernel rho;
3.1.8 outputs a hypergraph signal x.
3.2 pooling the dimension reduction hypergraph signals by utilizing the Hg-Pool method: firstly, sampling an adjacent matrix to obtain an assignment matrix; then, pooling the adjacent matrixes by using the assignment matrix, wherein the pooling result is represented by a symbol theta'; finally, replacing theta in the steps 3.1.2 and 3.1.7 with theta ', and carrying out the step 3.1 again to obtain a hypergraph signal x' after dimension reduction; the pooling dimension reduction process of the hypergraph signal is as follows:
3.2.1 defining a sampler HG, wherein the sampler is a deep learning component and is used for acquiring neighbor node information around a designated node in a graph neural network model;
3.2.2 sampling theta by utilizing HG, wherein the sampling result is defined as an assignment matrix, and is represented by a symbol A;
3.2.3 pooling of adjacency matrices with assignment matrices using the expression θ' =a T θ·a, where θ' is the pooling result;
3.2.4 replacing the theta in the step 3.1.2 with theta', and re-initializing the Laplace matrix;
3.2.5 replace θ in step 3.1.7 with θ', and re-solve the hypergraph signal;
3.2.6 the reduced dimension hypergraph signal is obtained by using step 3.1 and is denoted by the symbol x'.
The process of the fourth step is as follows:
4.1 semantically encoding the API service: inputting the description document of the API service into a pre-training language model to obtain a semantic embedded vector;
4.2, fusing semantic embedded vectors and hypergraph signals to obtain combined embedded vectors: splicing the semantic embedded vector in the step 4.1 with the hypergraph signal pooled in the step 3.2 to obtain a combined embedded vector, wherein the combined embedded vector is represented by a symbol z;
4.3, calculating recommendation probability to obtain a recommendation result: calculating the recommendation score of each API service by using the semantic embedded vector and the combined embedded vector; and then, generating recommendation probability by using the recommendation score, and selecting the API service with the highest probability as a recommendation result after obtaining the recommendation probability.
Preferably, in the 4.1, the encoding process of the semantic embedded vector is as follows:
4.1.1 defining a super parameter h for controlling the quantity of recommended results;
4.1.2 traversing all API services in the cloud computing data, and marking the API service fetched by the ith time as a i ；
4.1.3 will a i Is input into a pre-training language model to obtain a semantic embedded vector, and is marked by a symbol y i Representing that the function description document is a piece of function introduction text information about the API service;
4.1.4 when a i When the last API in the cloud computing data is served, the traversal is ended.
Still preferably, in the 4.3, the process of calculating the recommendation probability and making the recommendation is as follows:
4.3.1 traversing the API service in the cloud computing data, and marking the API service acquired at the ith time as a i ；
4.3.2 definition a i Is signed by the symbol se i A representation;
4.3.3 defining an exponentially formed recommendation score, symbolized byA representation;
4.3.4 pair a is based on step 4.1 i Semantic coding is carried out to obtain a semantic embedded vector y i ；
4.3.5 obtaining a combined embedding vector z based on step 4.2;
4.3.6 calculating the transpose of z and y i And normalized by the symbol y i ' indicating that normalization is a preprocessing method in deep learning, which can map the original data value to [0,1 ]]Between them;
4.3.7 y is i ' assign to se i ；
4.3.8 the natural constant eNumber operation with index y i ' assign the calculation result to
4.3.9 when a i When servicing the last API in the cloud computing data, finishing the traversal;
4.3.10 accumulate all exponentially related recommendation scores, the result being signed Σe se A representation;
4.3.11 traversing API service in cloud computing data and marking the API service fetched by the jth time as a j ；
4.3.12 definition a j Recommended probability rm of (1) j ；
4.3.13 use a j Recommendation score of (2)Divided by Σe se Assigning the calculation result to rm j ；
4.3.14 when a j When servicing the last API in the cloud computing data, finishing the traversal;
4.3.15 the API services are ordered according to the size of the recommendation probability;
4.3.16 the first h services with higher recommendation probability are selected as recommendation results.
The invention has the beneficial effects that: (1) The potential service combination relation in the cloud computing data is mined and utilized, so that the defect of low association degree of the existing recommendation method is overcome; (2) Designing a service combination hypergraph to realize effective modeling of service combination relation; (3) The approximate convolution and hypergraph pooling are combined, so that the calculation complexity and the overfitting risk in the hypergraph convolution process are reduced; (4) And semantic coding information and hypergraph signals are fused, so that the accuracy of the recommendation result is improved.
Drawings
FIG. 1 is a flow chart of constructing a combined set of sequences.
FIG. 2 is a schematic diagram of hypergraphic signal dimension reduction using the Hg-Pool method.
FIG. 3 shows the results of experiments comparing the proposed models of the invention and HyperRec, DHCN.
Detailed Description
The invention is further described below with reference to the drawings and examples of the specification.
Referring to fig. 1 to 3, a cloud computing data recommendation method based on a service combination hypergraph convolutional network includes the following steps:
step one: the potential service combination relation in the cloud computing data is mined, and a sequence combination set is constructed, wherein the process is as follows:
1.1 cloud computing data: in a cloud computing environment, call data between a user and an API service comprises the API service, a service call sequence and a service combination;
1.2API service: an application program interface (Application Programming Interface, API), denoted with the symbol a;
1.3 service call sequence: sequencing API service call records of users according to time sequence, wherein the sequencing result is called a service call sequence and is represented by a symbol S;
1.4 service composition: combining two or more API services, mixing and lapping to construct an API service with a new function, which is represented by a symbol mu;
1.5 mining service combination relations, constructing a sequence combination set: traversing the service call sequence, and extracting the API service according to the service combination relation. The API service set with service combination relation extracted from the service call sequence is defined as sequence combination item, which is represented by symbol M, the set formed by all sequence combination items is defined as sequence combination set, which is represented by symbol M, fig. 1 shows the construction process of sequence combination set in the form of flow chart, and the process of constructing sequence combination set is as follows:
1.5.1 defining a sequence combination set M and initializing to be an empty set;
1.5.2 traversing all service combinations in cloud computing data, recording the service call record fetched the ith time as mu i ；
1.5.3 defining a sequence combination item m and initializing to an empty set;
1.5.4 traversing mu i The API service in the process takes the kth time toThe API service is denoted as a k ；
1.5.5 traversing service call sequence S, and recording the service call record fetched by the jth time as S j ；
1.5.6 sign a for API service in service call record j A representation;
1.5.7 if a j And a k Equal, then a k Incorporating the sequence combination term m; otherwise, jumping to the step 1.5.5;
1.5.8 when s j When the last service call record in the service call sequence is recorded, finishing the traversal;
1.5.9 when a k Is mu i Ending the traversal when the last API service in the (a) is provided;
1.5.10 incorporating M into M;
1.5.11 when mu i When the last service in the cloud computing data is combined, finishing the traversal;
1.5.12 outputs a sequence combination set M.
Step two: constructing a service combination hypergraph based on the sequence combination set in the step one, realizing effective modeling of the combination characteristics of the API service, wherein the hypergraph is a special graph data structure, and the edges of the hypergraph can be connected with any number of nodes, and the process is as follows:
2.1 building a super point set of a service composition super graph: performing set operation on the service combination and the service call sequence, taking an intersection of the service combination and the service call sequence, converting the intersection into a superpoint in the service combination supergraph, adding the superpoint into the superpoint set, and constructing the superpoint set of the service combination supergraph, wherein the construction process of the superpoint set of the service combination supergraph is as follows:
2.1.1 creating a super point set V, and initializing the super point set V into an empty set;
2.1.2 traversing all service combinations in the cloud computing data, recording the service call record fetched the ith time as mu i ；
2.1.3 mu i All API services in (1) are marked as a set mSet i ；
2.1.4, marking all API services in the service call sequence as a set sSet;
2.1.5 pairs sSet and mSet i Performing set operation, taking the intersection of the two,by the symbol nSet i A representation;
2.1.6 traversing nSet i The j-th API service in the collection is marked as a j ；
2.1.7 if there is no superpoint V with subscript j in the superpoint set V j Then create superpoint v j Adding it to V;
2.1.8 if V is present in V j Then jump to step 2.1.6;
2.1.9 when a j Is nSet i Ending the traversal when the last API service in the (a) is provided;
2.1.10 when mu i And when the last service in the cloud computing data is combined, finishing the creation of the super point set and finishing the traversal.
2.2 building a hyperedge set of a service composition hypergraph: creating a superpoint set, defining a connection relation between a description superpoint of an incidence matrix and a superedge, traversing the sequence combination set and the superpoint set, and constructing the connection relation according to the inclusion relation of the API service corresponding to the sequence combination item and the superpoint, wherein the creation process of the superedge set of the service combination supergraph is as follows:
2.2.1 creating a superside set E and initializing the superside set E as an empty set;
2.2.2 defining an association matrix H for describing the connection relation between the superpoints and the supersides, and initializing the association matrix H into a null matrix;
2.2.3 traversing sequence combination set M, and taking out sequence combination item using symbol M k A representation;
2.2.4 creation of hyperedge e k ；
2.2.5 traversing the super-point set V, and marking the super-point fetched by the ith time as V j ；
2.2.6 find v j Corresponding API service a j ；
2.2.7 if the sequence combination term m k Contains a j Will exceed edge e k And superpoint v j Connecting, otherwise, jumping to the step 2.2.5;
2.2.8 in the association matrix H, assigning the j-th row and k-th column matrix elements to be 1;
2.2.9 when v j When the point is the last super point in V, finishing the traversal;
2.2.10 if e k At least two different superpoints are connected, e k And E is incorporated. Otherwise, delete e k ；
2.2.11 when m k And when the last item in the M sequence combination items is the last item, finishing the creation of the superside set and finishing the traversal.
2.3 constructing a service composition hypergraph and rewriting the hypergraph into a matrix form: the service composition hypergraph is modeling of service call relation, and the structure of the service composition hypergraph can be represented by the symbol G= (V, E), wherein G represents the service composition hypergraph, V represents the superpoint set, E represents the superedge set, and in order to extract a hypergraph signal, the service composition hypergraph is rewritten into a matrix form, and the process of rewriting the hypergraph into the matrix form is as follows:
2.3.1 definition of the hyperedge matrix D e The method is used for recording the connection relation between any superside and different superpoints;
2.3.2 pair D e Initializing, D e The number of rows and columns are the number of superedges in E. Traversal D e The diagonal element of row j and column j is assigned e j The number of all the connected super points, and the other matrix elements are assigned to 0;
2.3.3 definition of adjacency matrix as matrix form of service composition hypergraph, its expression is θ=h T ·D e H. Where θ represents the adjacency matrix, H is the associated matrix defined in 2.2.2, "T" is the transpose symbol of the matrix, "" is the multiplication symbol of the matrix, "=" is the assignment symbol of the matrix.
Step three: according to the ideas of Chebyshev approximate convolution, a hypergraph convolution network is designed to extract hypergraph signals on service combination hypergraphs, and then the hypergraph signals are subjected to dimension reduction by using an Hg-Pppl pooling method, wherein the process is as follows:
3.1 approximate convolution on service composition hypergraph: firstly, defining a Laplacian matrix of a hypergraph and a hypergraph convolution process, and then, according to the Chebyshev approximate convolution thought, performing expansion solution on the hypergraph convolution process to obtain a hypergraph signal, wherein the extraction process of the hypergraph signal is as follows:
3.1.1 defining a Laplacian matrix of the service composition hypergraph, denoted by the symbol L;
3.1.2 initializing a Laplace matrix, wherein the expression is L=I-theta', I is an identity matrix, and the symbol "-" represents matrix subtraction;
3.1.3 defining a convolution kernel of the hypergraph convolution network, denoted by the symbol ρ, the hypergraph convolution kernel being essentially a filter in deep learning, the mathematical form of which is a two-dimensional matrix;
3.1.4 define hypergraph signals on the service composition hypergraph, denoted by the symbol x. Hypergraph signals are vector representations of service composition relationships;
3.1.5 define the hypergraph convolution operation on the service composition hypergraph, denoted by the symbol. The overall hypergraph convolution process may be defined as x ρ;
3.1.6 according to Chebyshev approximate convolution idea, the hypergraph convolution process defined in step 3.1.5 is solved by using the chebyshev polynomial expansion, and the expression is thatIn the formula, sigma represents summation operation, the upper index N represents approximate maximum order, the lower index N represents order of summation, mu n Is the n-order chebyshev polynomial coefficient,/>Is the scaled n-th order polynomial of the Laplace matrix L, the symbol "≡" representing approximately equal;
3.1.7, a first-order chebyshev convolution operator is selected for solving, N is assigned to be 1, and the hypergraph signal extraction process can be further rewritten into x, rho, tau, theta, x. Wherein, tau is the parameter of the convolution kernel rho;
3.1.8 outputs a hypergraph signal x.
3.2 pooling the dimension reduction hypergraph signals by utilizing the Hg-Pool method: firstly, sampling an adjacent matrix to obtain an assignment matrix, then, pooling the adjacent matrix by using the assignment matrix, wherein the pooling result is represented by a symbol theta ', finally, theta ' is used for replacing theta in the steps 3.1.2 and 3.1.7, and step 3.1 is carried out again to obtain a hypergraph signal x ' after dimension reduction, and the pooling dimension reduction process of the hypergraph signal is as follows:
3.2.1 define a sampler HG. The sampler is a deep learning component for acquiring neighbor node information around a specified node in the graph neural network model, for example, lightGCN may be used as the sampler;
3.2.2 sampling theta by utilizing HG, wherein the sampling result is defined as an assignment matrix, and is represented by a symbol A;
3.2.3 pooling of adjacency matrices with assignment matrices using the expression θ' =a T θ·a. Wherein θ' is the pooling result;
3.2.4 replacing the theta in the step 3.1.2 with theta', and re-initializing the Laplace matrix;
3.2.5 replace θ in step 3.1.7 with θ', and re-solve the hypergraph signal;
3.2.6 obtaining a hypergraph signal after dimension reduction by utilizing the step 3.1, wherein the hypergraph signal is represented by a symbol x';
as shown in FIG. 2, the constructed service composition hypergraph instance contains v 1 、v 2 、v 3 、v 4 、v 5 、v 6 、v 7 Seven super points e 1 、e 2 、e 3 Three supersides, wherein superside e 1 Connection superpoint v 1 、v 2 、v 3 、v 7 Superb e 2 Connection superpoint v 3 、v 4 、v 5 Superb e 2 Connection v 5 、v 6 And rewriting the service combination hypergraph instance into a matrix form theta, and then sampling by using a sampler HG to obtain an assignment matrix A. Performing matrix multiplication operation on the transposed matrix of A, theta and A to obtain a pooling result theta'; then, θ 'is used to replace θ in step 3.1.2 and step 3.1.7, and step 3.1 is performed again to obtain hypergraph signal x' after dimension reduction.
Step four: carrying out semantic coding on the API service by utilizing a pre-training language model to obtain a semantic embedded vector, and fusing the semantic embedded vector and a hypergraph signal to obtain a combined embedded vector; finally, calculating the recommendation probability of the API service by utilizing the combined embedded vector and the hypergraph signal to obtain a recommendation result; the pre-training language model is a natural language processing method in deep learning, has a word vector coding function, and comprises the following steps of:
4.1 semantically encoding the API service: inputting the description document of the API service into a pre-training language model to obtain a semantic embedded vector, wherein the encoding process of the semantic embedded vector is as follows:
4.1.1 defining a super parameter h for controlling the quantity of recommended results;
4.1.2 traversing all API services in the cloud computing data, and marking the API service fetched by the ith time as a i ；
4.1.3 will a i Is input into a pre-training language model to obtain a semantic embedded vector, and is marked by a symbol y i And (3) representing. The function description document is a piece of function introduction text information about the API service;
4.1.4 when a i When servicing the last API in the cloud computing data, ending the traversal;
4.2, fusing semantic embedded vectors and hypergraph signals to obtain combined embedded vectors: splicing the semantic embedded vector in the step 4.1 with the hypergraph signal pooled in the step 3.2 to obtain a combined embedded vector, wherein the combined embedded vector is represented by a symbol z;
4.3, calculating recommendation probability to obtain a recommendation result: calculating the recommendation score of each API service by using the semantic embedded vector and the combined embedded vector, generating recommendation probability by using the recommendation score, and selecting the API service with the highest probability as a recommendation result after obtaining the recommendation probability; the specific process of calculating the recommendation probability and recommending is as follows:
4.3.1 traversing the API service in the cloud computing data, and marking the API service acquired at the ith time as a i ；
4.3.2 definition a i Is signed by the symbol se i A representation;
4.3.3 defining an exponentially formed recommendation score, symbolized byA representation;
4.3.4 pair a is based on step 4.1 i The semantic coding is performed such that,obtaining a semantic embedded vector y i ；
4.3.5 obtaining a combined embedding vector z based on step 4.2;
4.3.6 calculating the transpose of z and y i And normalized by the symbol y i ' representation. Normalization is a preprocessing method in deep learning, which can map the original data value to [0,1 ]]For example, a standard deviation normalization method is used;
4.3.7 y is i ' assign to se i ；
4.3.8 performing exponential operation on the natural constant e, wherein the index is yi', and assigning the calculation result to
4.3.9 when a i When servicing the last API in the cloud computing data, finishing the traversal;
4.3.10 accumulate all exponentially related recommendation scores, the result being signed Σe se A representation;
4.3.11 traversing API service in cloud computing data and marking the API service fetched by the jth time as a j ；
4.3.12 definition a j Recommended probability rm of (1) j ；
4.3.13 use a j Recommendation score of (2)Divided by Σe se Assigning the calculation result to rm j ；
4.3.14 when a j When servicing the last API in the cloud computing data, finishing the traversal;
4.3.15 the API services are ordered according to the size of the recommendation probability;
4.3.16 the first h services with higher recommendation probability are selected as recommendation results.
The following analyzes the actual effects of the invention with specific service data:
1. cloud computing data is selected as an experimental data set, and specifically comprises 1032 API services, 1423 service combinations and 741 service call sequences.
2. A DHCN model and a HyperRec model based on hypergraph convolution are adopted as a comparison method. The DHCN, hyperRec model is a method which combines hypergraphs to recommend and lead performance in the field of the current recommendation system.
3. The recommendation effect is evaluated using Hit Ratio (HR) index. HR index is used to measure recommendation accuracy, with higher index indicating better recommendation quality. The calculation rule of HR index is as follows:
wherein Q represents the number of recommended results, and Q is the serial number of the recommended results. hit is a hit function, which is 1 when the recommended result with sequence number q is selected by the user, and is 0 otherwise.
3. And randomly selecting 80% of data in the cloud computing data as training samples, and using the rest data as test samples. 10 independent replicates were run and the effect of the experiment is shown in figure 3. Analysis of FIG. 3 shows that the HR recommendation index of the present invention was higher than that of the two comparative methods in 10 experiments. The HR index of the present invention is up to 0.594, while the HR index of the DHCN model is up to 0.527, and the HR index of the HyperRec model is up to 0.428. The method has great advantage in recommendation precision, so that the method can be considered to have higher service recommendation quality in a cloud computing scene.
Claims (8)
1. The cloud computing data recommendation method based on the service combination hypergraph convolutional network is characterized by comprising the following steps of:
step one: mining potential service combination relations in cloud computing data, and constructing a sequence combination set;
step two: constructing a service combination hypergraph based on the sequence combination set in the step one, and realizing effective modeling of the combination characteristics of the API service;
step three: according to the thought of Chebyshev approximate convolution, a hypergraph convolution network is designed to extract hypergraph signals on a service combination hypergraph; then, performing dimension reduction treatment on the hypergraph signal by using an Hg-Pool pooling method;
step four: carrying out semantic coding on the API service by utilizing a pre-training language model to obtain a semantic embedded vector, and fusing the semantic embedded vector and a hypergraph signal to obtain a combined embedded vector; and finally, calculating the recommendation probability of the API service by utilizing the combined embedded vector and the hypergraph signal to obtain a recommendation result.
2. The cloud computing data recommendation method based on the service composition hypergraph convolutional network as recited in claim 1, wherein the procedure of the first step is as follows:
1.1 cloud computing data: in a cloud computing environment, call data between a user and an API service comprises the API service, a service call sequence and a service combination;
1.2API service: the application program interface API is represented with the symbol a;
1.3 service call sequence: sequencing API service call records of users according to time sequence, wherein the sequencing result is called a service call sequence and is represented by a symbol S;
1.4 service composition: combining two or more API services, mixing and lapping to construct an API service with a new function, which is represented by a symbol mu;
1.5 mining service combination relations, constructing a sequence combination set: traversing the service calling sequence, extracting API service according to the service combination relation, extracting an API service set with the service combination relation from the service calling sequence to be defined as a sequence combination item, wherein the sequence combination item is denoted by a symbol M, and a set formed by all the sequence combination items is defined as a sequence combination set and denoted by a symbol M.
3. The cloud computing data recommendation method based on the service composition hypergraph convolutional network according to claim 2, wherein in the step 1.5, the process of constructing the sequence composition set is as follows:
1.5.1 defining a sequence combination set M and initializing to be an empty set;
1.5.2 traversing all service combinations in the cloud computing data, and recording the service call record fetched by the ith time as mu i ；
1.5.3 defining a sequence combination item m and initializing to an empty set;
1.5.4 traversing mu i The API service in the (k) th acquisition is denoted as a) k ；
1.5.5 traversing service call sequence S, and recording the service call record fetched by the jth time as S j ；
1.5.6 sign a for API service in service call record j A representation;
1.5.7 if a j And a k Equal, then a k Incorporating the sequence combination term m; otherwise, jumping to the step 1.5.5;
1.5.8 when s j When the last service call record in the service call sequence is recorded, finishing the traversal;
1.5.9 when a k Is mu i Ending the traversal when the last API service in the (a) is provided;
1.5.10 incorporating M into M;
1.5.11 when mu i When the last service in the cloud computing data is combined, finishing the traversal;
1.5.12 outputs a sequence combination set M.
4. A cloud computing data recommendation method based on service combination hypergraph convolutional network as described in one of claims 1-3, wherein the procedure of the second step is as follows:
2.1 building a super point set of a service composition super graph: performing set operation on the service combination and the service call sequence, taking the intersection of the service combination and the service call sequence, converting the intersection into a superpoint in the service combination supergraph, and adding the superpoint into the superpoint set; the super point set construction process of the service combination super graph is as follows:
2.1.1 creating a super point set V, and initializing the super point set V into an empty set;
2.1.2 traversing all service combinations in the cloud computing data, and marking the service call record fetched by the ith time as m i ；
2.1.3 mu i All of the API services in (a)Denoted as set mSet i ；
2.1.4, marking all API services in the service call sequence as a set sSet;
2.1.5 pairs sSet and mSet i Performing set operation, taking intersection of the two, and using symbol nSet i A representation;
2.1.6 traversing nSet i The j-th API service in the collection is marked as a j ；
2.1.7 if there is no superpoint V with subscript j in the superpoint set V j Then create superpoint v j Adding it to V;
2.1.8 if V is present in V j Then jump to step 2.1.6;
2.1.9 when a j Is nSet i Ending the traversal when the last API service in the (a) is provided;
2.1.10 when mu i And when the last service in the cloud computing data is combined, finishing the creation of the super point set and finishing the traversal.
2.2 building a hyperedge set of a service composition hypergraph: creating a superpoint set, defining a connection relation between the description superpoint of the incidence matrix and the superedge, traversing the sequence combination set and the superpoint set, and constructing the connection relation according to the inclusion relation of the API service corresponding to the sequence combination item and the superpoint; the creation process of the hyperedge set of the service composition hypergraph is as follows:
2.2.1 creating a superside set E and initializing the superside set E as an empty set;
2.2.2 defining an association matrix H for describing the connection relation between the superpoints and the supersides, and initializing the association matrix H into a null matrix;
2.2.3 traversing sequence combination set M, and taking out sequence combination item using symbol M k A representation;
2.2.4 creation of hyperedge e k ；
2.2.5 traversing the super-point set V, and marking the super-point fetched by the ith time as V j ；
2.2.6 find v j Corresponding API service a j ；
2.2.7 if the sequence combination term m k Contains a j Will exceed edge e k And superpoint v j Connecting, otherwise, jumping to the step 2.2.5;
2.2.8 in the association matrix H, assigning the j-th row and k-th column matrix elements to be 1;
2.2.9 when v j When the point is the last super point in V, finishing the traversal;
2.2.10 if e k At least two different superpoints are connected, e k Incorporate E, otherwise delete E k ；
2.2.11 when m k And when the last item in the M sequence combination items is the last item, finishing the creation of the superside set and finishing the traversal.
2.3 constructing a service composition hypergraph and rewriting the hypergraph into a matrix form: the service composition hypergraph is modeling of service call relation, and the structure of the service composition hypergraph can be represented by a symbol G= (V, E), wherein G represents the service composition hypergraph, V represents a superpoint set, E represents a superedge set, and the service composition hypergraph is rewritten into a matrix form for extracting hypergraph signals; the process of rewriting the hypergraph into matrix form is as follows:
2.3.1 definition of the hyperedge matrix D e The method is used for recording the connection relation between any superside and different superpoints;
2.3.2 pair D e Initializing, D e The number of rows and columns are the number of superedges in E, traversing D e The diagonal element of row j and column j is assigned e j The number of all the connected super points, and the other matrix elements are assigned to 0;
2.3.3 definition of adjacency matrix as matrix form of service composition hypergraph, its expression is θ=h T ·D e H, where θ represents an adjacency matrix, H is an association matrix defined in 2.2.2, " T "is the transposed symbol of the matrix,".
5. A cloud computing data recommendation method based on service combination hypergraph convolutional network as described in one of claims 1-3, wherein the process of the third step is as follows:
3.1 approximate convolution on service composition hypergraph: firstly, defining a Laplacian matrix of a hypergraph and a hypergraph convolution process, and then, according to the Chebyshev approximate convolution thought, performing expansion solution on the hypergraph convolution process to obtain a hypergraph signal, wherein the extraction process of the hypergraph signal is as follows:
3.1.1 defining a Laplacian matrix of the service composition hypergraph, denoted by the symbol L;
3.1.2 initializing a laplace matrix, which expression is l=i- θ.
In the formula 3.1.3, I is an identity matrix, and the symbol 'one' represents matrix subtraction;
3.1.4 defining a convolution kernel of the hypergraph convolution network, denoted by the symbol ρ, the hypergraph convolution kernel being essentially a filter in deep learning, the mathematical form of which is a two-dimensional matrix;
3.1.5 defining hypergraph signals on the service composition hypergraph, wherein the hypergraph signals are represented by a symbol x, and the hypergraph signals are vector representations of the service composition relations;
3.1.6 defining a hypergraph convolution operation on the service composition hypergraph, denoted by a symbol x, the whole hypergraph convolution process being definable as x p;
3.1.7 according to the Chebyshev approximate convolution idea, the hypergraph convolution process defined in step 3.1.5 is solved by using the Chebyshev polynomial expansion, and the expression is thatIn the formula, sigma represents summation operation, the upper index N represents approximate maximum order, the lower index N represents order of summation, mu n Is the n-order chebyshev polynomial coefficient,/>Is the scaled n-th order polynomial of the Laplace matrix L, the symbol "≡" representing approximately equal;
3.1.8, a first-order chebyshev convolution operator is selected for solving, and the hypergraph signal extraction process can be further rewritten into x, rho and tau, theta and x, wherein tau is a parameter of a convolution kernel rho;
3.1.9 outputs a hypergraph signal x.
3.2 pooling the dimension reduction hypergraph signals by utilizing the Hg-Pool method: firstly, sampling an adjacent matrix to obtain an assignment matrix; then, pooling the adjacent matrixes by using the assignment matrix, wherein the pooling result is represented by a symbol theta'; finally, replacing theta in the steps 3.1.2 and 3.1.7 with theta ', and carrying out the step 3.1 again to obtain a hypergraph signal x' after dimension reduction; the pooling dimension reduction process of the hypergraph signal is as follows:
3.2.1 defining a sampler HG, wherein the sampler is a deep learning component and is used for acquiring neighbor node information around a designated node in a graph neural network model;
3.2.2 sampling θ with HG, the sampling result being defined as an assignment matrix, denoted by symbol a,
3.2.3 pooling of adjacency matrices with assignment matrices using the expression θ' =a T θ·a, where θ' is the pooling result;
3.2.4 replacing the theta in the step 3.1.2 with theta', and re-initializing the Laplace matrix;
3.2.5 replace θ in step 3.1.7 with θ', and re-solve the hypergraph signal;
3.2.6 the reduced dimension hypergraph signal is obtained by using step 3.1 and is denoted by the symbol x'.
6. The cloud computing data recommendation method based on the service composition hypergraph convolutional network as recited in claim 5, wherein the process of the fourth step is as follows:
4.1 semantically encoding the API service: inputting the description document of the API service into a pre-training language model to obtain a semantic embedded vector;
4.2, fusing semantic embedded vectors and hypergraph signals to obtain combined embedded vectors: splicing the semantic embedded vector in the step 4.1 with the hypergraph signal pooled in the step 3.2 to obtain a combined embedded vector, wherein the combined embedded vector is represented by a symbol z;
4.3, calculating recommendation probability to obtain a recommendation result: calculating the recommendation score of each API service by using the semantic embedded vector and the combined embedded vector; and then, generating recommendation probability by using the recommendation score, and selecting the API service with the highest probability as a recommendation result after obtaining the recommendation probability.
7. The cloud computing data recommendation method based on the service composition hypergraph convolutional network as recited in claim 6, wherein in 4.1, the encoding process of the semantic embedded vector is as follows:
4.1.1 defining a super parameter h for controlling the quantity of recommended results;
4.1.2 traversing all API services in the cloud computing data, and marking the API service fetched by the ith time as a i ；
4.1.3 will a i Is input into a pre-training language model to obtain a semantic embedded vector, and is marked by a symbol y i Representing that the function description document is a piece of function introduction text information about the API service;
4.1.4 when a i When the last API in the cloud computing data is served, the traversal is ended.
8. The cloud computing data recommendation method based on the service composition hypergraph convolutional network as recited in claim 6, wherein the process of computing the recommendation probability and making the recommendation in 4.3 is as follows:
4.3.1 traversing the API service in the cloud computing data, and marking the API service acquired at the ith time as a i ；
4.3.2 definition a i Is signed by the symbol se i A representation;
4.3.3 defining an exponentially formed recommendation score, symbolized byA representation;
4.3.4 pair a is based on step 4.1 i Semantic coding is carried out to obtain a semantic embedded vector y i ；
4.3.5 obtaining a combined embedding vector z based on step 4.2;
4.3.6 calculating the transpose of z and y i And normalized by the symbol y i ' indicating that normalization is a preprocessing method in deep learning, which can map the original data value to [0,1 ]]Between them;
4.3.7 y is i ' assign to se i ；
4.3.8 performing exponential operation on the natural constant e, wherein the exponent is y i ' assign the calculation result to
4.3.9 when a i When servicing the last API in the cloud computing data, finishing the traversal;
4.3.10 accumulate all exponentially related recommendation scores, the result being signed Σe se A representation;
4.3.11 traversing API service in cloud computing data and marking the API service fetched by the jth time as a j ；
4.3.12 definition a j Recommended probability rm of (1) j ；
4.3.13 use a j Recommendation score of (2)Divided by Σe se Assigning the calculation result to rm j ；
4.3.14 when a j When servicing the last API in the cloud computing data, finishing the traversal;
4.3.15 the API services are ordered according to the size of the recommendation probability;
4.3.16 the first h services with higher recommendation probability are selected as recommendation results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311271535.8A CN117370650A (en) | 2023-09-28 | 2023-09-28 | Cloud computing data recommendation method based on service combination hypergraph convolutional network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311271535.8A CN117370650A (en) | 2023-09-28 | 2023-09-28 | Cloud computing data recommendation method based on service combination hypergraph convolutional network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117370650A true CN117370650A (en) | 2024-01-09 |
Family
ID=89403263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311271535.8A Pending CN117370650A (en) | 2023-09-28 | 2023-09-28 | Cloud computing data recommendation method based on service combination hypergraph convolutional network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117370650A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013130A (en) * | 2024-04-08 | 2024-05-10 | 烟台大学 | Service recommendation method, system, equipment and storage medium based on super service network |
-
2023
- 2023-09-28 CN CN202311271535.8A patent/CN117370650A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118013130A (en) * | 2024-04-08 | 2024-05-10 | 烟台大学 | Service recommendation method, system, equipment and storage medium based on super service network |
| CN118013130B (en) * | 2024-04-08 | 2024-06-04 | 烟台大学 | Service recommendation method, system, equipment and storage medium based on super service network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111291266B (en) | Artificial intelligence based recommendation method and device, electronic equipment and storage medium | |
| CN112633419B (en) | Small sample learning method and device, electronic equipment and storage medium | |
| CN110175628A (en) | A kind of compression algorithm based on automatic search with the neural networks pruning of knowledge distillation | |
| CN110019616B (en) | POI (Point of interest) situation acquisition method and equipment, storage medium and server thereof | |
| CN109816438B (en) | Information pushing method and device | |
| US11841839B1 (en) | Preprocessing and imputing method for structural data | |
| CN110795944A (en) | Recommended content processing method and device, and emotion attribute determining method and device | |
| CN113190670A (en) | Information display method and system based on big data platform | |
| CN111815432A (en) | Financial service risk prediction method and device | |
| CN111325030A (en) | Text label construction method and device, computer equipment and storage medium | |
| CN117370650A (en) | Cloud computing data recommendation method based on service combination hypergraph convolutional network | |
| CN112836509A (en) | Expert system knowledge base construction method and system | |
| CN114780746A (en) | Knowledge graph-based document retrieval method and related equipment thereof | |
| CN115438169A (en) | Text and video mutual inspection method, device, equipment and storage medium | |
| CN108427756A (en) | Personalized query word completion recommendation method and device based on same-class user model | |
| CN113609337A (en) | Pre-training method, device, equipment and medium of graph neural network | |
| CN115131698A (en) | Video attribute determination method, device, equipment and storage medium | |
| CN108647295B (en) | Image labeling method based on depth collaborative hash | |
| CN112632948B (en) | Case document ordering method and related equipment | |
| CN113806579A (en) | Text image retrieval method and device | |
| CN111259975B (en) | Method and device for generating classifier and method and device for classifying text | |
| CN117194742A (en) | Industrial software component recommendation method and system | |
| CN111597336A (en) | Processing method and device of training text, electronic equipment and readable storage medium | |
| CN115186738B (en) | Model training method, device and storage medium | |
| CN115587261A (en) | Government affair resource catalog recommendation method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |