CN113095084A - Semantic service matching method and device in Internet of things and storage medium - Google Patents

Semantic service matching method and device in Internet of things and storage medium Download PDF

Info

Publication number
CN113095084A
CN113095084A CN202110279367.1A CN202110279367A CN113095084A CN 113095084 A CN113095084 A CN 113095084A CN 202110279367 A CN202110279367 A CN 202110279367A CN 113095084 A CN113095084 A CN 113095084A
Authority
CN
China
Prior art keywords
user
ijk
intelligent
semantic
model
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.)
Granted
Application number
CN202110279367.1A
Other languages
Chinese (zh)
Other versions
CN113095084B (en
Inventor
黄宏程
田亚楠
胡敏
陶洋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN202110279367.1A priority Critical patent/CN113095084B/en
Publication of CN113095084A publication Critical patent/CN113095084A/en
Application granted granted Critical
Publication of CN113095084B publication Critical patent/CN113095084B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/01—Social networking
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
    • G16Y10/00—Economic sectors
    • G16Y10/75—Information technology; Communication

Abstract

The invention belongs to the field of edge computing and Internet of things, and relates to a semantic service matching method, a semantic service matching device and a semantic service matching storage medium in the Internet of things; the method comprises the steps of obtaining initial interest of a user in the smart object from the use frequency and the average use time of the smart object by the user; establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event; acquiring dynamic relations among the intelligent objects by using a hyperplane time perception knowledge graph embedded model, and estimating social similarity among the intelligent objects; constructing a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects; receiving a service request from a user or an intelligent object, performing similarity matching based on a semantic service matching model, and outputting a service matching result between the user and the intelligent object; the invention ensures the effectiveness and expansibility of service discovery and the accuracy of service matching results.

Description

Semantic service matching method and device in Internet of things and storage medium
Technical Field
The invention belongs to the field of edge computing and Internet of things, and particularly relates to a semantic service matching method and device in the Internet of things and a storage medium, wherein the semantic service matching method and device is based on user time preference and object social relations.
Background
The internet of things is known as the future of the internet, and the communication form between people at present is expanded to people, people and things, and things. Smart spaces are the names of physical locations where numerous devices interact with each other to provide relevant services to human users under given conditions. The functions that these objects possess may be referred to as services, which provide real-time status information of the physical world. The task of how to provide appropriate services in such an environment faces many new challenges and requirements. Furthermore, the internet of things has dynamic characteristics, the states of physical entities are changing constantly and the heterogeneity between the physical entities causes different interaction modes, and in order to deal with the dynamic characteristics of physical world objects and solve the problem of tight coupling of service-oriented architectures, some researchers try to apply an event-driven service-oriented architecture to service discovery of the internet of things. Meanwhile, the internet of things also enables the definition of the social network to be not limited to the space between people, but to be expanded to the space between people, people and things, and things. In the prior art, the attribute of the social network is added in the internet of things, the relationship between objects of the internet of things can be defined by analogy with the human social network, an abstract model of the social internet of things is designed, and the social network structure established based on the objects of the internet of things is analyzed, so that the model based on the human social network is expanded into the social network of the objects, the object sharing in the internet of things is realized, and the direct interaction between a user and an object is realized.
Some studies have proposed discovering the SIOT object service using semantic Web technologies. The idea behind semantic-based service discovery techniques is to define a metadata model to describe various features of the SIOT object, and then recommend objects that meet the user's needs through semantic similarity matching. Currently, most research in this area is focused on task-oriented schemes. Jin X, Chun S, Jung J, et al.A fast and scalable adaptation for IoT service selection based on a physical service model [ J ]. Information Systems Frontiers,2017, through dynamic clustering of application services and based on the precision quantity to calculate semantic similarity between services and requests, proposes an effective service discovery mechanism, Zhou M, Ma Y.A web service discovery Computing method for IOT system [ C ]// IEEE International Conference on Cloud Computing & Intelligent systems.0. also proposes an ontology-based service matching algorithm, combines semantic similarity and semantic relativity, and is used for service discovery in the Internet of things. Because the types and the number of the service and the object use events of the internet of things are rapidly increased, on one hand, a manual service matching method for event driving becomes more and more inefficient and inflexible, on the other hand, the physical event driving service discovery is simply considered, only the interaction between objects is considered, and the human factors are not considered, but under the SIOT architecture, the social relationship among the objects can be utilized, the difference among the objects is transparent, the social relationship among the users is combined, the sparsity influence of user data is reduced, the distributed service discovery is realized, and the effectiveness and the expansibility of the service discovery are guaranteed. As previously described, the smart store may utilize sensors to collect check-in behavior of customers, such as: check-in time and dwell time, which provides a new way for the present invention to learn about the user's desired service tendencies. However, the service discovery of the smart objects should not only take the selection preference of the user into consideration, but also take the social relationship between the smart objects into consideration, such as the collaborative location relationship, the collaborative work relationship, and so on. In addition, a series of difficulties and challenges are brought to the discovery of the sio service due to the characteristics of spontaneity, context dependence, monopoly and the like of the discovery of the sio service:
1) user preference tendencies for smart objects are often implicit and difficult to measure: unlike social networking or ratings websites, a user's preference for an object can be represented by a rating or comment sentence, while a user's preference for a smart object service is typically hidden in the SIOT, so existing content and collaborative filtering based service discovery and recommendation approaches are not applicable.
2) Smart objects in the SIOT are often heterogeneous, their similarity often being difficult to estimate: in SIOT, objects are heterogeneous in both function and attribute and cannot be represented in a uniform feature space. Furthermore, the service text description of objects is often incomplete, and therefore neither feature-based models nor text-based models can be used to estimate similarity between objects.
3) Data sparsity problem: most of internet of things service discovery systems usually learn user preferences according to user history records, when a new intelligent object is added and the user uses too few intelligent objects, the problem of data sparsity is caused under the condition that interactive history data between the user and the objects are difficult to collect due to privacy and safety problems, and the accuracy of service matching and service discovery is influenced.
Disclosure of Invention
Based on the above analysis, the invention provides a semantic service matching method, a semantic service matching device and a semantic service matching storage medium in the internet of things, so as to solve the following problems:
1) because users tend to use different objects at different time periods of a day and the selection of services is influenced by time factors, the invention provides that the time preference of the users is extracted through the use frequency and the average use time of the objects so as to better utilize the time context and the subjective context information of the users to realize the discovery of the services of the Internet of things.
2) Through a knowledge graph embedding mechanism of hyperplane time perception, dynamic social relations between intelligent objects are captured better, therefore, interactivity between heterogeneous objects is enhanced, and independent service discovery between objects is achieved.
3) Finally, a SIOT semantic service matching model is proposed for receiving and responding to service requests from users and smart objects, and evaluating on a standard data set verifies the validity and accuracy of the model proposed by the invention.
In a first aspect of the present invention, the present invention provides a semantic service matching method in an internet of things, including:
acquiring initial interest of a user on the smart object from the use frequency and the average use time of the smart object by the user;
establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
acquiring dynamic relations among the intelligent objects by using a hyperplane time perception knowledge graph embedded model, and estimating social similarity among the intelligent objects;
constructing a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
receiving a service request sent by a user or an intelligent object, carrying out similarity matching based on a semantic service matching model, and outputting a service matching result between the user and the intelligent object.
In a second aspect of the present invention, the present invention provides a semantic service matching device in the internet of things, where the device includes:
the interest acquisition unit is used for acquiring the initial interest of the user on the intelligent object from the use frequency and the average use time of the user on the intelligent object;
the preference intensity determining unit is used for establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
the similarity determining unit is used for acquiring the dynamic relation between the intelligent objects by utilizing the hyperplane time perception knowledge map embedded model and estimating the social similarity between the intelligent objects;
the semantic service matching model building unit builds a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
and the matching unit is used for receiving a service request sent by a user or an intelligent object, carrying out similarity matching on the semantic service matching model built by the semantic service matching model building unit and outputting a service matching result between the user and the intelligent object.
In a third aspect of the present invention, the present invention provides a storage medium storing a plurality of instructions, the instructions being suitable for being loaded by a processor and executing the semantic service matching method in the internet of things according to the first aspect of the present invention.
In a fourth aspect of the present invention, the present invention also provides a server comprising a processor and a storage medium, the processor being configured to implement the instructions to:
the instructions are suitable for being loaded by a processor and executing the semantic service matching method in the internet of things according to the first aspect of the invention.
The invention has the beneficial effects that:
the invention provides a semantic service discovery model of a time-aware social internet of things, aiming at the problems that user service selection is influenced by time factors, user preference information is difficult to collect in the environment of the social internet of things, and the social relationship between objects is difficult to estimate due to the characteristics of heterogeneity of the objects and dynamic change of the social relationship between the objects. Firstly, extracting the time preference of a user from the use frequency and the average use time of the objects, constructing the social relationship between the objects by using a hyperplane time perception knowledge graph embedding model, ensuring the effectiveness and the expansibility of equipment service discovery, and finally inputting a service request into a semantic service matching model to realize quick service response and ensure the accuracy of a service matching result.
Drawings
FIG. 1 is a social networking services matching framework diagram based on social relationships in an embodiment of the present invention;
FIG. 2 is a flow chart of a semantic service matching method according to an embodiment of the present invention;
FIG. 3 is a HyTE model diagram employed in the embodiment of the present invention;
FIG. 4 is a diagram of a semantic service matching device according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a social internet of things service matching framework diagram based on social relationships in an embodiment of the present invention, and as shown in fig. 1, the present invention first introduces an overall structure of the service matching framework diagram, and then elaborates each component module of the service matching framework diagram, including: time-based user preference extraction, estimation of social similarity between smart objects using graph embedding, and semantic service matching models. Firstly, extracting the use preference of a user from the use frequency and the average use time of the intelligent object by the user, and aiming at the characteristic that the isomerism and the similarity of the SIOT object are difficult to estimate, the invention uses a graph embedding method to build the social relation of the objects by adopting a hyperplane-based time perception knowledge graph embedding model, and estimates the social similarity between the objects. Finally, the invention provides a semantic service matching model for service matching response, which can accept service requests sent by users or intelligent objects, complete matching based on similarity calculation, maintain a dynamic service list and realize quick response.
Fig. 2 is a flowchart of a semantic service matching method in the embodiment of the present invention, and as shown in fig. 2, the semantic service matching method performs semantic service matching by the following method:
101. acquiring initial interest of a user on the smart object from the use frequency and the average use time of the smart object by the user;
in the process, due to the development of perception technology, a large number of sensors of different types are distributed, so that the invention can easily collect the use events of SIOT objects (intelligent objects), wherein the use events contain abundant context information and can be used for revealing cognitive states of targets, preferences, emotions and the like of users. Due to time, effort, and resource limitations, users tend to select content they like in subject use events. Since the probability of a smart object usage event occurring is positively correlated to the user's preference for that smart object. Therefore, the method and the system can take the use tendency of a user to a certain intelligent object service as a hidden factor of an intelligent object use event and model the hidden factor to extract the preference of the user for service matching.
The user is influenced by the preference of the user, the user often selects the service frequently used by the user, and the preference tendency of the user to the intelligent object is usually implicit, so that the preference of service selection is difficult to measure; secondly, the service selection of the user is influenced by time factors and dynamically changes along with time, the preference strength of the user based on time is used as a hidden factor to model the service selection of the user, the service requirements of the user in two dimensions of time and preference can be fully considered during service matching, and the effectiveness and the accuracy of service matching are improved.
Thus, the modeling calculation process for the initial interest of the smart object is as follows:
user uiIn time slot tkTo intelligent object ojPreferred intensity of(ijk)By e(ijk)Determination of e(ijk)Representing user uiIn time slot tkTo smart object ojOf the initial interest. The invention can estimate the initial interest e from the use frequency and the average use time of the intelligent object(ijk)；e(ijk)The calculation formula of (a) is as follows:
Figure BDA0002978035450000061
wherein, FijkAnd QijkRespectively represent users uiAt time interval tkTo usejThe frequency of use and the average time of use of,
Figure BDA0002978035450000062
representing an object uiIn time slot tkFor the maximum value of the object usage frequency, similarly,
Figure BDA0002978035450000063
the maximum value representing the average time of use of the object. Theta is the frequency of use of the regulation modelAnd weight of time; θ represents a weight for adjusting the smart object usage frequency and the average usage time, and may be set to about 0.5 in the present embodiment.
102. Establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
the invention makes the use preference I of the user(ijk)The invention may further be F, modeling the user object usage events as hidden factorsijkAnd QijkIntroducing auxiliary variables
Figure BDA0002978035450000071
And
Figure BDA0002978035450000072
to increase the accuracy of the model. The invention models the initial interest e(ijk)Preference for user I(ijk)Influence of and I(ijk)The effect on the smart object usage time, modeled to represent the relationship between these variables, is represented by a joint distribution function:
P(I(ijk),Fijk,Qijk|e(ijk))＝P(I(ijk)|e(ijk))P(Fijk|I(ijk))P(Qijk|I(ijk))
wherein, P (I)(ijk),Fijk,Qijk|e(ijk)) Representation based on initial interest e(ijk)Below I(ijk)、Fijk、QijkA joint distribution probability of (a); p (I)(ijk)|e(ijk)) Representing a first conditional probability; p (F)ijk|I(ijk)) Representing a second conditional probability; p (Q)ijk|I(ijk)) Representing a third conditional probability; e.g. of the type(ijk)Representing user uiIn time slot tkTo smart object ojThe initial interest of (1); i is(ijk)Representing user uiIn time slot tkTo intelligent object ojThe preferred strength of (d); fijkRepresentative user uiIn time slot tkUsing smart object ojThe frequency of use of (c); qijkRepresentative user uiAt time interval tkUsing smart object ojAverage usage time of (2).
After the joint distribution model is established, how to specifically extract the preference intensity of the user needs further processing, according to the statistical analysis of sample data, the influence of the initial interest on the user presents Gaussian distribution, the probability that the service is selected again is known as a function of the service use frequency and the average time of the user, and a power law distribution trend is presented(ijk)|e(ijk)) Using a gaussian distribution model to represent:
P(I(ijk)|e(ijk))＝(ηe(ijk),σ2)
wherein η is a parameter estimated by EM algorithm, and is updated in each iteration, and its value determines the accuracy of the service matching model, in this embodiment, σ2Set to 0.5 in the EM algorithm.
The second conditional probability P (F)ijk|I(ijk)) And a third conditional probability P (Q)ijk|I(ijk)) Each is represented by a power law distribution model:
Figure BDA0002978035450000073
Figure BDA0002978035450000074
wherein alpha is1，β1，α2，β2Parameters estimated by the EM algorithm are adopted and updated in each iteration, the convergence of the maximum likelihood function is determined by the values of the parameters, and when the maximum value is obtained, the precision of the service matching model can be increased by accurate parameter values, so that the satisfaction degree of a user on the service matching result is increased.
In order to avoid over-learning, the present invention needs to perform regularization on these parameters, which is expressed as:
Figure BDA0002978035450000081
Figure BDA0002978035450000082
Figure BDA0002978035450000083
l＝1,2
wherein λ isη，
Figure BDA0002978035450000084
Is according to L2Parameters introduced by regularization are used for preventing under-fitting and over-fitting conditions of the model in the training process, the effect of optimizing a service matching model is achieved, and the model precision is increased.
After the parameters are processed, training and learning are required to be carried out on the parameters, and the sample data is expressed as phi ═ UxO × T, (i, j) ∈ phi, namely, a user-smart object-time triple, and is expressed as D { (i)1,j1,k1),...,(iM.jN,kS) I.e. i denotes a user, j denotes a smart object, k denotes a time (time slot), and the subscripts thereof denote the respective serial numbers or numbers. During model training, variable e(ijk)、FijkAnd QijkTherefore, the invention brings the regularized model into a joint distribution model, and can obtain:
Figure BDA0002978035450000085
the above formula represents the maximum likelihood function of the joint distribution model, and when the maximum likelihood function is obtained, the model parameters approach the accurate solution, so that convergence is achieved, and the model training is finished. The invention regularizes a parameter lambdaη，
Figure BDA0002978035450000086
And
Figure BDA0002978035450000087
all the parameters are 0.01, and it can be understood that the regularization parameter represents a parameter required to be used in the regularization process, the regularized parameter in the invention is a parameter required to be solved by the invention, and the parameter is used for optimizing a joint distribution model, so that more accurate preference intensity of the user is extracted; the invention can use the maximum likelihood function to estimate unknown model parameter sigma eta, alphal,βl1,2, taking logarithm of two sides of the maximum likelihood function, converting the continuous multiplication into summation convenient operation, and obtaining:
Figure BDA0002978035450000088
to speed up the training process of the model, at the parameters η, αl，βlAnd variable I(ij)The random gradient descent algorithm is adopted in the optimization, and in each iteration, the loss function on a certain piece of training data is optimized randomly, so that the updating speed of each parameter is greatly accelerated. The gradient vector is the derivative of the function to each variable, the direction of the vector is the direction of the gradient, the magnitude of the vector is the magnitude of the gradient, and the partial derivative is calculated for each variable after the normalization processing, so that the following can be obtained:
Figure BDA0002978035450000091
Figure BDA0002978035450000092
Figure BDA0002978035450000093
in each iteration, the parameter eta, alpha is measured by Newton methodl，βlAnd variable I(ij)Updating is performed so that they can quickly approach the exact solution, convergence is achieved, the model training process is ended, and the parameter update formula can be written as:
Figure BDA0002978035450000094
Figure BDA0002978035450000095
Figure BDA0002978035450000096
Figure BDA0002978035450000097
the model parameter optimization process is shown in Table 1, and model parameters output by the algorithm are brought into the model after power law distribution modeling, namely
Figure BDA0002978035450000098
The magnitude of the user preference strength can be described using the object use frequency and the average use time. Probability of use P (I)(ijk),Fijk,Qijk|e(ijk)) To describe the dependency relationship between the variables, which represents how much the object usage frequency and the average usage time have influence on the user preference strength, for example, when the object usage frequency is 4 times per hour, and the average usage time in all time periods is 12 minutes, the user preference strength can be 0.7 by the above method.
TABLE 1 model parameter optimization Algorithm
Figure BDA0002978035450000099
Figure BDA0002978035450000101
In the embodiment of the invention, as shown in table 1, in each iteration of the newton method processing, sample data of each triple is firstly traversed, and preference intensity of a user to an intelligent object is updated; traversing regularization parameters after power law distribution modeling processing; finally, the regularization parameters after the Gaussian distribution modeling processing are updated
103. Acquiring dynamic relations among the intelligent objects by using a hyperplane time perception knowledge graph embedded model, and estimating social similarity among the intelligent objects;
in the embodiment, the social similarity between the smart objects is estimated by using graph embedding, and the traditional SIOT service discovery method is to disclose the service selection preference of a user according to the historical comment or score of the user, which ignores the semantic information of the SIOT object. Moreover, with the development of semantic technologies, such as a semantic sensing network, an object semantic network and SIOT, a large amount of open semantic data is accumulated, the invention can utilize the data to construct a time-aware SIOT knowledge graph to capture various dynamic relations between heterogeneous SIOT objects, and then the entity and various social relations are mapped into a low-dimensional space to estimate the semantic similarity between the objects.
Considering the influence of time on SIOT service discovery and the timeliness of the relationship between any entity, the invention adopts a hyperplane-based time-aware knowledge graph embedding model (HyTE) to construct the relationship between objects. HyTE divides the whole KG into a plurality of static sub-graphs according to time, each sub-graph corresponds to a time stamp, then an entity and the sub-graphs are projected onto a specific plane, and therefore a hyperplane normal vector containing time information and a distribution representation of KG elements are obtained.
First, unlike conventional (h, r, T), the HyTE model proposes a time-stamped triplet (h, r, T, [ T ] Ts,Te]) As shown in FIG. 3, the entities are connected toThe relation between the entities is mapped to each sub-plane through time segments, each sub-plane represents the relation between the entities at the current time, TsAnd TeRepresenting the start time and the end time of the validity period of the triplet, respectively. Thus, KG can be expressed as:
Figure BDA0002978035450000111
t is a discrete point in time. Time is divided into T time steps, then
Figure BDA0002978035450000112
The standard normal vector of the hyperplane divided by the T time steps is shown, and then the triples are projected into the hyperplane of different time stamps according to different effective times, and the mapping formula is as follows:
Figure BDA0002978035450000113
Figure BDA0002978035450000114
Figure BDA0002978035450000115
wherein entity ehRepresenting a head entity, etRepresenting tail entities, erRepresenting the relationship between the two entities, wγSatisfy the constraint | | wγ||2Time hyperplane is represented on all time slices as 1. Constraint eh、etAnd erAnd a score function measuring the distance between them, in the following formula:
Pγ(eh)+Pγ(er)≈Pγ(et)
fΥ(h,r,t)＝||PΥ(eh)+PΥ(er)-PΥ(et)||l1/l2
wherein l1And l2Expressing a distance measurement mode, respectively expressing Manhattan distance and Euclidean distance, and optimizing an objective by using an interval-based pairwise minimum ranking loss function for differentiating correct triples from incorrect triples, wherein the equation is as follows:
Figure BDA0002978035450000121
D+indicating the correct triplet, D-Representing the wrong triplet, the invention adopts time-dependent negative sampling to obtain negative samples, and the generated negative samples only exist in the SIOT knowledge graph and do not exist in a subgraph of a specific time slice. And (3) by regularizing the entity vector and performing time embedding normalization, randomly accessing triples in the SIOT knowledge graph for multiple times, and obtaining the minimum value of the loss function by adopting a random gradient descent algorithm. K for the inventionjAnd kvRespectively represent an object ojAnd ovBecause the similarity calculation is sensitive to the value range of the variable and the embedded vector is high-latitude, the method adopts the Pearson correlation coefficient to estimate the object ojAnd ovSocial similarity between sems (o)j,ov) The calculation formula is as follows:
Figure BDA0002978035450000122
since the social relationships between objects in the SIOT are typically dynamic rather than static, for example: co-located object relationships and co-working object relationships, the present invention therefore represents the knowledge graph as G ═ E, R, T, Λ, E represents the set of objects,
Figure BDA0002978035450000123
representing a set of typed relationships between objects, T representing the time at which the object relationship begins and ends, representing the ontology of a set of relationship types. For example,G:(h,r,t,[γs,γe]) Is shown in the time gap gammasTo gammaeValid triplets (h, r, t) within. The invention first embeds the SIOT knowledge graph into graph G and then learns to each group
Figure BDA0002978035450000124
And
Figure BDA0002978035450000125
represents the embedded vectors and finally estimates the social similarity of each pair of SIOT objects.
104. Constructing a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
the smart objects in the SIOT are usually heterogeneous, the similarity of the smart objects is usually difficult to estimate, and the social similarity between the smart objects needs to be calculated according to the social internet of things relationships such as ownership relationship (owership), parent social relationship (parent social relationship), co-location relationship (co-location), collaboration relationship (co-work) and social relationship (social object relationship) established among the smart objects of the internet of things in a low-dimensional vector space. In service matching, not only the time preference of the user but also the social relationship among the objects are considered, so that a plurality of candidate services which are interested in the user can be matched for the user based on the social similarity among the objects, and the expansibility and the usability of the model are increased.
The essential purpose of service matching is to establish a correlation between service requests and service responses. Existing research is done mainly by manual and a priori knowledge, in other words, services and events are matching pairs specified by subjective experience, not automatic matching. Furthermore, unlike conventional service discovery, the trigger of event-driven service discovery is a physical event, which is extracted from the awareness information, and the aspect of user-initiated request is ignored, making it unsuitable for use in a human-centric service discovery method. Aiming at the problems, the invention provides a semantic service matching model of the invention, which is mainly divided into three parts: service request identification, semantic service matching, and maintaining a list of available services. The main work of service request identification is used for identifying a service request initiated by a user or an intelligent object, and then semantic similarity matching is carried out on the service request and a service to be selected.
Services are described by OWL-SE, which mainly contains four aspects of information in the service profile, namely input, output, preconditions and results. The inputs and outputs generally represent the basic functional features of service matching, and the present invention represents the capabilities of service request identification (RS) and semantic service Matching (MS) as a service based on description logic. In the specification of the present invention, the present invention defines a service request initiated by a user or an object as an output identified by the service request and an input matched by a semantic service, which means that the final matching result can be expressed as an output matched by the semantic service.
Figure BDA0002978035450000131
Figure BDA0002978035450000132
Figure BDA0002978035450000133
According to the above formula, the service matching model can be expressed as:
Figure BDA0002978035450000134
where τ is a given threshold value, ErAnd EmIndividual watchInput indicating that the output of the service request identification matches the semantic service, Sim (E)r,Em) Is the similarity between the two, when Sim (E)r,Em) Above τ, are considered similar and match successfully.
Sim (E) is calculated by the following formular,Em) Wherein sem (o)j,ov) Is an object ojAnd object ovSocial similarity between them, sim (o)j,ov) Is an object ojAnd ovSimilarity based on correlation between them, by simultaneous pair ojAnd ovA set of user estimates for rating.
Sim(Er,Em)＝λ*sem(oj,ov)
+(1-λ)*sim(oj,ov)
And finally, evaluating the quality of the semantic service matching result of the invention through a collaborative filtering model based on the article. Setting U e U to represent user in user set U, O e O to represent an object in object set O, according to different service request initiator, the invention divides scoring function into two categories: r isu(ui,oj,tk) And ro(oi,oj,tk)，ru(ui,oj,tk) Indicating that the user initiated the service request, obtained the rating of the service response, ro(oi,oj,tk) It means that the object initiates a request to obtain the score of the service response. The calculation formula is as follows:
Figure BDA0002978035450000141
Figure BDA0002978035450000142
105. receiving a service request sent by a user or an intelligent object, carrying out similarity matching based on a semantic service matching model, and outputting a service matching result between the user and the intelligent object.
In this embodiment, after a user or an intelligent object sends a service request, no matter the user or the intelligent object sends, matching may be performed according to the similarity of the semantic service matching model, and a corresponding user or a corresponding intelligent object is determined according to the matching.
In the embodiment of the invention, the service is described by OWL-SE language, the language information mainly comprises four aspects of information in service configuration files, namely input, output, preconditions and results, after receiving a service request sent by a user or an intelligent object, the service request described by OWL-SE language is firstly converted into a word vector represented semantically through a word embedding technology, and the converted word vector is input into a semantic service matching model for similarity matching, so that the service matching result between the user and the intelligent object is output.
The scheme of the embodiment of the invention can solve the problem of dynamic interaction between the 'things' and the 'things' in the Internet of things, realize the autonomous service discovery between the 'things' and the 'things', improve the interoperability and provide personalized service discovery results for users.
The method of the embodiment can be applied to user-intelligent object interaction systems such as a search engine, a question-answering system and the like in an Internet of things platform.
An embodiment of the present invention further provides a semantic service matching device in the internet of things, as shown in fig. 4, the semantic service matching device includes:
the interest acquisition unit is used for acquiring the initial interest of the user on the intelligent object from the use frequency and the average use time of the user on the intelligent object;
the interest acquisition unit may model an initial interest of the smart object by a formula
Figure BDA0002978035450000151
The initial interest of each user in the smart object at a time slot is solved.
The preference intensity determining unit is used for establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
the preference intensity unit is used for indicating the use preference I of the user(ijk)The invention may further be F, modeling the user object usage events as hidden factorsijkAnd QijkIntroducing auxiliary variables
Figure BDA0002978035450000152
And
Figure BDA0002978035450000153
to increase the accuracy of the model; the preference intensity unit models the initial interest e(ijk)Preference for user I(ijk)Influence of and I(ijk)The method comprises the steps of modeling to express the relation among variables by the influence of the service time of an intelligent object, mining the relation among the variables again according to a Gaussian distribution model and a power law distribution model, finally obtaining learned parameters by learning the gradient of the parameters in the relation among the variables, and obtaining the preference intensity of a user by bringing the learned parameters into the model again.
The similarity determining unit is used for acquiring the dynamic relation between the intelligent objects by utilizing the hyperplane time perception knowledge map embedded model and estimating the social similarity between the intelligent objects;
the similarity determining unit estimates social similarity between the intelligent objects by utilizing graph embedding, constructs a SIOT knowledge graph based on time perception by utilizing semantic information of the SIOT objects to capture various dynamic relations between heterogeneous SIOT objects, and then maps the entities and various social relations into a low-dimensional space to estimate the semantic similarity between the intelligent objects.
The semantic service matching model building unit builds a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
the semantic service matching model building unit builds a semantic service matching model through service request identification, semantic service matching and maintenance for a service list, the main work of the service request identification is used for identifying a service request initiated by a user or an intelligent object, and then semantic similarity matching is carried out on the service request and a service to be selected.
And the matching unit is used for receiving a service request sent by a user or an intelligent object, carrying out similarity matching on the semantic service matching model built by the semantic service matching model building unit and outputting a service matching result between the user and the intelligent object.
The matching unit evaluates the quality of the semantic service matching result of the invention through a collaborative filtering model based on the article. Setting U e U to represent the user in the user set U, O e O to represent an object in the object set O, and according to the difference of service request initiators, dividing the scoring function into two types in the embodiment of the invention: r isu(ui,oj,tk) And ro(oi,oj,tk)，ru(ui,oj,tk) Indicating that the user initiated the service request, obtained the rating of the service response, ro(oi,oj,tk) It means that the smart object initiates a request to obtain the score of the service response, and a result with a higher score may be selected as a matching result.
The embodiment of the invention also provides a storage medium, wherein the storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by the processor and executing the semantic service matching method in the internet of things.
An embodiment of the present invention further provides a server, where the server includes a processor and a storage medium, and the processor is configured to implement each instruction:
the instructions are suitable for being loaded by a processor and executing the semantic service matching method in the internet of things as described in the embodiment.
In the description of the present invention, it is to be understood that the terms "coaxial", "bottom", "one end", "top", "middle", "other end", "upper", "one side", "top", "inner", "outer", "front", "center", "both ends", and the like, indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplicity of description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, are not to be construed as limiting the present invention.
In the present invention, unless otherwise expressly stated or limited, the terms "mounted," "disposed," "connected," "fixed," "rotated," and the like are to be construed broadly, e.g., as meaning fixedly connected, detachably connected, or integrally formed; can be mechanically or electrically connected; the terms may be directly connected or indirectly connected through an intermediate, and may be communication between two elements or interaction relationship between two elements, unless otherwise specifically limited, and the specific meaning of the terms in the present invention will be understood by those skilled in the art according to specific situations.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (8)

1. A semantic service matching method in the Internet of things is characterized by comprising the following steps:
acquiring initial interest of a user on the smart object from the use frequency and the average use time of the smart object by the user;
establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
acquiring dynamic relations among the intelligent objects by using a hyperplane time perception knowledge graph embedded model, and estimating social similarity among the intelligent objects;
constructing a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
receiving a service request sent by a user or an intelligent object, converting the service request into a semantically expressed word vector through a word embedding technology, inputting the word vector into a semantic service matching model for similarity matching, and outputting a service matching result between the user and the intelligent object.
2. The matching method for semantic services in the internet of things according to claim 1, wherein the initial interest of the user in the smart object is calculated as:
Figure FDA0002978035440000011
wherein e is(ijk)Representing user uiIn time slot tkTo smart object ojThe initial interest of (1); theta is the weight for adjusting the use frequency and the use time of the model; fijkRepresentative user uiIn time slot tkUsing smart object ojThe frequency of use of (c);
Figure FDA0002978035440000012
representing user uiIn time slot tkFrequency of use of the smart object; qijkRepresentative user uiIn time slot tkUsing smart object ojAverage usage time of (a);
Figure FDA0002978035440000013
representing user uiIn time slot tkThe average usage time of the smart object is used.
3. The matching method for semantic services in the internet of things according to claim 1, wherein the joint distribution model is expressed as:
P(I(ijk),Fijk,Qijk|e(ijk))＝P(I(ijk)|e(ijk))P(Fijk|I(ijk))P(Qijk|I(ijk))
wherein, P (I)(ijk),Fijk,Qijk|e(ijk)) Representation based on initial interest e(ijk)Below I(ijk)、Fijk、QijkA joint distribution probability of (a); p (I)(ijk)|e(ijk)) Representing a first conditional probability; p (F)ijk|I(ijk)) Representing a second conditional probability; p (Q)ijk|I(ijk)) Representing a third conditional probability; e.g. of the type(ijk)Representing user uiIn time slot tkTo smart object ojThe initial interest of (1); i is(ijk)Representing user uiIn time slot tkTo intelligent object ojThe preferred strength of (d); fijkRepresentative user uiIn time slot tkUsing smart object ojThe frequency of use of (c); qijkRepresentative user uiIn time slot tkUsing smart object ojAverage usage time of (2).
4. The matching method for semantic services in the internet of things according to claim 3, wherein the first conditional probability is modeled by adopting Gaussian distribution, the second conditional probability and the third conditional probability are modeled by adopting power law distribution, and the modeled parameters are normalized; inputting sample data of a user-intelligent object-time triple, training the parameters after regular processing, and optimizing the joint distribution model after training.
5. The matching method for semantic services in the internet of things according to claim 4, wherein the parameters after regularization are trained, parameter estimation is performed by adopting an EM (effective electromagnetic) algorithm, the first conditional probability is modeled by using Gaussian distribution, the second conditional probability and the third conditional probability are modeled by using power law distribution, a maximum likelihood function is obtained according to the joint distribution probability, and in each iteration of processing by adopting a Newton method, sample data of each triplet is traversed first, and the preference intensity of a user on an intelligent object is updated; traversing regularization parameters after power law distribution modeling processing; and finally updating the regularization parameters after the Gaussian distribution modeling processing.
6. A semantic service matching device in the Internet of things is characterized in that the device comprises:
the interest acquisition unit is used for acquiring the initial interest of the user on the intelligent object from the use frequency and the average use time of the user on the intelligent object;
the preference intensity determining unit is used for establishing a joint distribution model to extract the preference intensity of the user by utilizing the influence of the initial interest on the preference intensity and the influence of the preference intensity on the intelligent object use event;
the similarity determining unit is used for acquiring the dynamic relation between the intelligent objects by utilizing the hyperplane time perception knowledge map embedded model and estimating the social similarity between the intelligent objects;
the semantic service matching model building unit builds a semantic service matching model according to the preference degree of the user and the social similarity between the intelligent objects;
and the matching unit is used for receiving a service request sent by a user or an intelligent object, carrying out similarity matching on the semantic service matching model built by the semantic service matching model building unit and outputting a service matching result between the user and the intelligent object.
7. A storage medium storing instructions adapted to be loaded by a processor and to perform a method for semantic service matching in the internet of things according to any one of claims 1 to 5.
8. A server, comprising a processor and a storage medium, the processor configured to implement instructions to:
the instructions are suitable for being loaded by a processor and executing the semantic service matching method in the Internet of things according to any one of claims 1-5.
CN202110279367.1A 2021-03-16 2021-03-16 Semantic service matching method and device in Internet of things and storage medium Active CN113095084B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110279367.1A CN113095084B (en) 2021-03-16 2021-03-16 Semantic service matching method and device in Internet of things and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110279367.1A CN113095084B (en) 2021-03-16 2021-03-16 Semantic service matching method and device in Internet of things and storage medium

Publications (2)

Publication Number Publication Date
CN113095084A true CN113095084A (en) 2021-07-09
CN113095084B CN113095084B (en) 2022-09-23

Family

ID=76668121

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110279367.1A Active CN113095084B (en) 2021-03-16 2021-03-16 Semantic service matching method and device in Internet of things and storage medium

Country Status (1)

Country Link
CN (1) CN113095084B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113807977A (en) * 2021-09-02 2021-12-17 北京建筑大学 Method, system, device and medium for detecting Touchi attack based on dynamic knowledge graph

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103188755A (en) * 2013-01-06 2013-07-03 西安交通大学 Mobile perception service node selection method facing to internet of things
KR20150120555A (en) * 2014-04-17 2015-10-28 전자부품연구원 Global IoT Resource Discovery Service Method and Server using the same
CN107205016A (en) * 2017-04-18 2017-09-26 中国科学院计算技术研究所 The search method of internet of things equipment
CN108460489A (en) * 2018-03-15 2018-08-28 重庆邮电大学 A kind of user behavior analysis based on big data technology and service recommendation frame
CN109194746A (en) * 2018-09-06 2019-01-11 广州知弘科技有限公司 Heterogeneous Information processing method based on Internet of Things
CN109271497A (en) * 2018-08-31 2019-01-25 华南理工大学 A kind of event-driven service matching method based on term vector
WO2020119699A1 (en) * 2018-12-11 2020-06-18 Oppo广东移动通信有限公司 Resource publishing method and apparatus in internet of things, device, and storage medium
CN111901144A (en) * 2020-06-19 2020-11-06 深圳奇迹智慧网络有限公司 Interaction method and device for Internet of things equipment, computer equipment and storage medium

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103188755A (en) * 2013-01-06 2013-07-03 西安交通大学 Mobile perception service node selection method facing to internet of things
KR20150120555A (en) * 2014-04-17 2015-10-28 전자부품연구원 Global IoT Resource Discovery Service Method and Server using the same
CN107205016A (en) * 2017-04-18 2017-09-26 中国科学院计算技术研究所 The search method of internet of things equipment
CN108460489A (en) * 2018-03-15 2018-08-28 重庆邮电大学 A kind of user behavior analysis based on big data technology and service recommendation frame
CN109271497A (en) * 2018-08-31 2019-01-25 华南理工大学 A kind of event-driven service matching method based on term vector
CN109194746A (en) * 2018-09-06 2019-01-11 广州知弘科技有限公司 Heterogeneous Information processing method based on Internet of Things
WO2020119699A1 (en) * 2018-12-11 2020-06-18 Oppo广东移动通信有限公司 Resource publishing method and apparatus in internet of things, device, and storage medium
CN111901144A (en) * 2020-06-19 2020-11-06 深圳奇迹智慧网络有限公司 Interaction method and device for Internet of things equipment, computer equipment and storage medium

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
MUTASIM ELSADIG ADAM: "Usages of Semantic Web Services Technologies in IoT Ecosystems and its Impact in Services Delivery: A survey", 《INTERNATIONAL JOURNAL OF COMPUTER》 *
SAHRAOUI DHELIM 等: "Cyberentity and its consistency in the cyber-physical-social-thinking hyperspace", 《COMPUTERS AND ELECTRICAL ENGINEERING》 *
邓达成: "语义物联网中事件驱动的服务发现关键问题研究", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113807977A (en) * 2021-09-02 2021-12-17 北京建筑大学 Method, system, device and medium for detecting Touchi attack based on dynamic knowledge graph

Also Published As

Publication number Publication date
CN113095084B (en) 2022-09-23

Similar Documents

Publication Publication Date Title
CN108363804B (en) Local model weighted fusion Top-N movie recommendation method based on user clustering
CN108717408B (en) Sensitive word real-time monitoring method, electronic equipment, storage medium and system
CN111177538B (en) User interest label construction method based on unsupervised weight calculation
CN109255033A (en) A kind of recommended method of the knowledge mapping based on location-based service field
CN111061946A (en) Scenario content recommendation method and device, electronic equipment and storage medium
CN112541122A (en) Recommendation model training method and device, electronic equipment and storage medium
Li et al. Image sentiment prediction based on textual descriptions with adjective noun pairs
CN109783614B (en) Differential privacy disclosure detection method and system for to-be-published text of social network
CN115917535A (en) Recommendation model training method, recommendation device and computer readable medium
Li et al. Product innovation concept generation based on deep learning and Kansei engineering
CN111143704B (en) Online community friend recommendation method and system integrating user influence relationship
CN111222847B (en) Open source community developer recommendation method based on deep learning and unsupervised clustering
CN116244513B (en) Random group POI recommendation method, system, equipment and storage medium
CN113239264A (en) Personalized recommendation method and system based on meta-path network representation learning
CN111429161B (en) Feature extraction method, feature extraction device, storage medium and electronic equipment
CN107506419B (en) Recommendation method based on heterogeneous context sensing
Kakad et al. Ontology construction from cross domain customer reviews using expectation maximization and semantic similarity
CN113095084B (en) Semantic service matching method and device in Internet of things and storage medium
CN117271818B (en) Visual question-answering method, system, electronic equipment and storage medium
CN108389113B (en) Collaborative filtering recommendation method and system
CN114358807A (en) User portrayal method and system based on predictable user characteristic attributes
CN117235264A (en) Text processing method, apparatus, device and computer readable storage medium
CN116089708A (en) Agricultural knowledge recommendation method and device
Gao et al. Construction of digital marketing recommendation model based on random forest algorithm
CN114637920A (en) Object recommendation method and device

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