KR101859845B1 - Offline Friend Recommendation using Mobile Context and Online Friend Network Information based on Tensor Factorization - Google Patents

Abstract

The present invention provides an offline friend recommendation method using mobile context and tensor factorization based on online friend network information and a system thereof. A social network service method comprises: a step of receiving offline context of a plurality of users using a social network service and evaluation result information for recommended friends to undergo a parsing process to store the offline context and the evaluation result information in a database by a preprocessor; a step of collecting and standardizing friend relationship information of the plurality of users to store the friend relationship information in the database; a step of using information stored in the database to convert the information into a format of input information used in a tensor and a matrix by a data converter; a step of simultaneously factorizing and recoupling a tensor and a matrix by a recommendation engine; and a step of using a recoupled tensor to generate a recommendation friend list for each user by a friend list generation unit.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an offline friend recommendation method based on mobile situation information and an online friend network based on tensor decomposition,

The present invention relates to a method and system for offline friend recommendation through mobile situation information and online friend network information-based tensor decomposition. More specifically, in order to recommend people who have similar life patterns and are likely to connect to each other, it is possible to use the place visit record information collected through the user's terminal and the beacon terminal installed at each place, To a friend recommendation algorithm in which network information is applied to a coupled matrix and a tensor factorization technique.

With the proliferation of online social network services (OSNS) and the popularization of smart mobile devices, users are able to network easily with social networking with people around the world, anywhere, anytime, anywhere. However, paradoxically, these technologies are skipped over the processes of human relationship formation that are formed through trial and error through numerous face-to-face interactions in OSNS and immersed users, It caused difficulties. Research on "recommendation of friends" has been actively researched as a way to provide opportunities for people to form, maintain and expand interpersonal relationships. Friend referral research is based on recommendation methods based on online data (eg, friend relationship information, interests and web visits), and recommendation methods based on offline data (eg, GPS log, Bluetooth log, etc.) , Both of which recommend a person as a friend who has a similar tendency to the user, a higher relationship, or a similar life pattern. The online data-based friend recommendation method allows social relations to be formed at any time with people who are distant from each other without time and space limitations, but the lack of opportunity for face-to-face (F2F) It is not helpful to develop and maintain interpersonal skills in the real world by omitting the process of deeply understanding, understanding and communicating with each other. The offline data - based friend recommendation method focuses on the user 's peripheral activity space based on the context information collected from the user' s mobile device sensor, and thus recommends a friend who can directly increase the F2F interaction opportunity. This enables strong tie formation through skill development and socialization of interpersonal relationships, mutual understanding and positive emotional formation. Therefore, the present invention argues that offline recommendation research based on offline data can be a solution to mitigate the difficulties in forming and maintaining interpersonal relationships in the real world.

An important design consideration factor in recommendation technique is to predict the possibility that recommender A and recommender B can accept each other as friends. Many existing context - based friend recommendation studies analyze the location usage patterns and GPS pattern matching information extracted from the context information and recommend people with similar patterns as friends to increase this possibility. However, because it is based solely on offline status information, it is easy to analyze people with similar life patterns, but does not take into account the user's already established Friend Network, and does not take advantage of the friend network information . For example, if there are four people A, B, C, and D, and their GPS patterns and location records are all the same, everyone is recommended as a friend without discernment. However, if you can take into account the OSNS buddy network information of A, B, C, D, and you have information that A and C can connect through another person Z, then C is recommended rather than A or B This will increase the likelihood of being accepted as a friend. In this way, we need a friend-based recommendation technique based on contextual information that can discover and provide F2F interaction opportunities offline.

Korean Registered Patent No. 10-1714897 (Mar.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and system for recommending friends based on context information that can discover and provide opportunities for offline F2F interaction. To do this, we gather the user 's context information and user' s friend relation information formed in the social network from the smart terminal sensor and recommend friend in offline environment based on tensor decomposition and combination.

In one aspect, a data modeling method for a social network service proposed in the present invention includes a step of deriving a plurality of elements constituting a tensor using information of a first user, second user information and position information, And decomposing the tensor comprised of the plurality of elements based on a matrix representing relationship information between the user and the second user.

The step of deriving the plurality of elements constituting the tensor using the information of the first user, the second user information, and the position information may include: when the first user or the second user enters or exits the place where the beacon terminal is installed, The terminal of the first user or the second user receives the Bluetooth signal from the beacon terminal and transmits the reception time of the Bluetooth signal and the unique code of the beacon terminal to the server.

A record using the place where the first user or the second user is installed in the beacon terminal is accumulated in the server and the element value of the plurality of elements constituting the tensor represents the similar degree of the life pattern of the first user and the second user.

The connection score among the plurality of elements constituting the matrix is derived based on the number of common friends and the length of the connection relationship path between the first user and the second user, and the proximity between the first user and the second user .

The connection value is varied according to the hop length to reflect the degree of proximity reduction according to the number of hops in the length of the connection relationship path.

In a case where the friend relationship information is a non-public user, in order to grasp the connection relationship path, the friend relationship information is converted into a non-direction graph using the symmetry property, Friend relationship information infer a friend relationship of a private user.

According to another aspect of the present invention, a social network service system proposed in the present invention receives offline status information of a plurality of users using a social network service and evaluation result information on recommended friends, A processor, a database for collecting, shaping, and storing friend relationship information of the plurality of users, a data converter for transforming into a form of input information used in a tensor and a matrix using information stored in a database, a tensor and a matrix, A friend list generating unit for generating a recommended friend list for each user by using the re-combined tensor, and a message generating and notifying unit for generating a notification message for transmitting the generated recommended friend list to the terminals of the users .

The preprocessor parses the received information through a context data parser and an evaluation data parser and stores the parsed information in a database. The preprocessor stores the friend information of the users through a network crawler And stores it in the database in the form of <Source node ID, Destination node ID>.

The database stores the friend relationship information of the plurality of users into a plurality of tables including the entrance and exit time of the user, the friend relationship information, the user profile, the ID of the user terminal, and the satisfaction evaluation result information about the recommended friend do.

The data converter converts the information stored in the database into the form of <User, User, Place, Total visit count> and <User, User, Connection score>.

The recommendation engine creates codes using Matlab and uses a Tensor Toolbox and a CMTF toolbox.

According to another aspect of the present invention, a method for providing a social network service according to the present invention includes receiving offline status information of a plurality of users using a social network service through a preprocessor and evaluation result information on recommended friends, Collecting and formatting friend relationship information of the plurality of users and storing the same in a database, converting the information into a form of input information used in a tensor and a matrix through a data converter using information stored in a database, Disassembling and reassembling the tensor and the matrix simultaneously through a recommendation engine, and generating a recommended friend list for each user using the reassembled tensor through the friend list generator.

The step of receiving the offline situation information of the plurality of users using the social network service and the evaluation result information of the recommended friends through the preprocessor and storing the received information in a database after parsing is performed by a Context Data Parser ) And an Evaluation Data Parser. The collected information is stored in a database. The friend relationship information of users is collected and formulated through a network crawler, and the form of the <Source node ID, Destination node ID> To the database.

The step of collecting and formatting the friend relationship information of the plurality of users and storing the friend relationship information in the database includes storing the friend relationship information of the plurality of users into the user's place entry and exit time, friend relationship information, user profile, And stores the plurality of tables including the satisfaction evaluation result information about the friend.

The step of transforming the information stored in the database into the form of input information used in the tensor and the matrix through the data converter using the information stored in the database includes: score>.

The step of simultaneously decomposing and recombining the tensors and matrices through the recommendation engine creates codes using Matlab and uses a Tensor Toolbox and a CMTF toolbox.

The step of generating a recommended friend list for each user using the re-combined tensor through the friend list generation unit generates a notification message through a message generation and notification unit to transmit the generated recommended friend list to the terminals of the users .

According to another aspect of the present invention, there is provided a file distribution system for distributing a file of an application installed in a user terminal proposed by the present invention, the file transferring unit transferring the file according to a request of the user terminal, When receiving a Bluetooth signal transmitted from a beacon terminal when the beacon terminal is installed in a place where a beacon terminal is installed, a collection module for transmitting information of the beacon terminal to the server in XML format and a notification of a list of recommended friends received from the server A user interface module for confirming a message, a user having a similar life pattern among a plurality of users in the off-line, searching for users having a common friend in a social network, and using multidimensional information using tensors and matrices Model, and concurrently through the CMTF algorithm. The.

Embodiments of the present invention provide a context-based friend recommendation technique that can discover and provide opportunities for offline F2F interaction. To do this, we collect the user 's context information and user friend relationship information formed in the social network from the smart terminal sensor, and recommend friend in offline environment based on tensor decomposition and combination.

1 is a flowchart illustrating a data modeling method for a social network service according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining a combination matrix and a tensor decomposition technique according to an embodiment of the present invention. Referring to FIG.
3 is a view for explaining tensor decomposition according to an embodiment of the present invention.
4 is a diagram illustrating a configuration of a social network service system according to an embodiment of the present invention.
5 is a flowchart illustrating a method of providing a social network service according to an embodiment of the present invention.
6 is a diagram illustrating a data transformation process for a join matrix and tensor decomposition according to an embodiment of the present invention.
FIG. 7 illustrates a user interface according to an embodiment of the present invention. Referring to FIG.

With the popularization of smartphones and OSNS (Online Social Network Service), people can interact with people around the world at anytime, anywhere, and as a result, people's OSNS usage continues to increase. However, for those who focus on forming interpersonal relationships through some OSNS, they may experience difficulties in developing and maintaining interpersonal skills by skipping the process of human relationships formed through numerous F2F interactions with a single "recommend" button . The present invention proposes a friend-based recommendation technique based on context information that can detect and provide opportunities for offline F2F interaction. To do this, we gather the user 's context information and user' s friend relation information formed in the social network from the smart terminal sensor and recommend friend in offline environment based on tensor decomposition and combination. In other words, in order to recommend people with similar life patterns and high connection possibility, the place visit information collected through the user's smart terminal and the beacon terminal installed at each place, We propose a friend recommendation algorithm applying network information to join matrix and tensor decomposition method. For the evaluation of the proposed technique, the situation information of 43 persons (12 experimental participants + 31 random users) was collected for one week and 12 people were evaluated for their satisfaction with recommended persons. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a data modeling method for a social network service according to an embodiment of the present invention.

The data modeling method for the proposed social network service fuses the user 's place visit record collected from the Bluetooth - based beacon terminal with the user' s social network friend relationship information through the combination matrix and the tensor decomposition method.

The data modeling method for the social network service includes deriving 110 a plurality of elements constituting the tensor using the information of the first user, the second user information and the location information, And decomposing the tensor comprised of the plurality of elements based on a matrix representing relationship information.

First, in step 110, a plurality of elements constituting a tensor are derived using the information of the first user, the second user information, and the position information. When a first user or a second user enters or leaves a place where the beacon terminal is installed, the terminal of the first user or the second user receives the Bluetooth signal from the beacon terminal, Can be transmitted to the server. A record in which a first user or a second user uses a place where the beacon terminal is installed is accumulated in the server and the element value of the plurality of elements constituting the tensor can represent a similar degree of life patterns of the first user and the second user . For example, a user's place visit record is represented by a three-dimensional tensor X E ^ (U x U x P).

Next, at step 120, a tensor composed of the plurality of elements is decomposed based on a matrix representing relationship information between the first user and the second user. The connection score among the plurality of elements constituting the matrix is derived based on the number of common friends and the length of the connection relationship path between the first user and the second user, and the proximity between the first user and the second user . Where the degree of proximity may be the degree of relevance between the first user and the second user derived based on, for example, the life pattern of first and second users or the number of common friends.

The connection value may be changed according to the hop length to reflect a degree of proximity according to the number of hops of the length of the connection relationship path, and in the case of a user whose friend relationship information is private, The friend relationship information is converted into the non-directional graph using the symmetry property of the friend relationship information, and the friend relationship information can be inferred from the friend relationship of the non-user using the friend relationship information of the other users have.

For example, the friendship information of a social network is represented by a two-dimensional matrix M ∈ R ^ (U × U). Where the matrix M is a joint matrix representing the relationship information between two users, i. E., The first user and the second user, and considered together as additional information when the tensor X is decomposed and decomposed together. Such a coupling matrix and a tensor decomposition technique will be described in more detail with reference to FIGS. 2 and 3. FIG.

FIG. 2 is a diagram for explaining a combination matrix and a tensor decomposition technique according to an embodiment of the present invention. Referring to FIG.

The existing graph and matrix models are very difficult to consider additional information. In the case of the matrix, the two-dimensional limit is limited. In the case of the graph model, a new modeling is required to add new information, and the graph becomes complicated.

Therefore, we use a combination matrix and a tensor decomposition technique to fuse information of different dimensions such as online and offline information. Since the tensor decomposition based model can find hidden elements or missing values in information, it is becoming a way to analyze multidimensional data generated from various sources in recent years.

According to the embodiment of the present invention, a tensor is constructed from the location information collected from the user's mobile terminal sensor, and the online friend relationship information of the user is modeled in the form of a matrix. The basic concept of the coupling matrix and the tensor and the decomposition and recombination process are shown in FIG.

For example, the binding matrix and the tensor 210 that are not fully combined may include a first user (User) 212, a second user (User) 213, and a Place (214) And a combining matrix 215 including a tensor 211 and a first user 216 and a second user 217 in an offline state.

The fully coupled coupling matrix and tensor 210 may comprise a plurality of elements (Component R) 220, where R may be rank for decomposition.

The plurality of elements (Component R) 220 may be transformed into a complete coupling matrix and a tensor 230 through a decomposition and recombination process. The complete join matrix and tensor 230 includes a first User 231, a second User 232 place 233, a first User 234 offline ), And a second user (User 235). Much of the data may be missing values, but this can be found through decomposition and combination with additional information.

3 is a view for explaining tensor decomposition according to an embodiment of the present invention.

For example, as shown in FIG. 3, a two-dimensional tensor, that is, a matrix 310, can be modeled by only modeling the relationship between the first user 311 and the second user 312.

In contrast, the three dimensional tensor 320 (or three dimensional or more) may represent and represent the relationship between the first user 321, the second user 322, and the time 323. In other words, the relationship between the first user 321 and the second user 322 according to the time 323 can be shown.

4 is a diagram illustrating a configuration of a social network service system according to an embodiment of the present invention.

The social network service system includes a preprocessor 410, a database 420, a data converter 430, a recommendation engine 440, a Friend Lists Generator ) 450 and a GCM Message Generator & Notification (GCM) 460. In addition, the mobile client application 470 may include a Context Collector User Interface (UI) module.

The preprocessor 410 receives the offline situation information of a plurality of users using the social network service and the evaluation result information on the recommended friends, parses them, and stores them in the database.

The preprocessor 410 parses the received information through a context data parser and an evaluation data parser and stores the parsed information in a database. The preprocessor 410 transmits the parsed information to the user via a network crawler, And stores the information in the database 430 in the form of < Source node ID, Destination node ID >.

The database 420 collects and formats friend relationship information of the plurality of users and stores them. The database 420 stores the friend relationship information of the plurality of users into the user's location entry and exit time, friend relationship information, user profile, user terminal ID (e.g., Google Cloud Message (GCM) ID) And stores it as a plurality of tables including satisfaction evaluation result information about friends. 6 (a) and 6 (b) show examples of such information.

The data converter 430 converts the information stored in the database into the form of input information used in the tensor and matrix. The data converter 430 converts the information stored in the database in the form of <User, User, Place, Total visit count> and <User, User, Connection score> as shown in FIG.

Recommendation Engine 440 simultaneously decomposes and recombines the tensor and matrix. The recommendation engine 440 writes the code using Matlab and uses a Tensor Toolbox and a CMTF toolbox.

The Friend Lists Generator 450 creates a list of recommended friends for each user using the reassembled tensor.

The GCM Message Generator & Notification 460 generates a notification message (for example, a Google Cloud Message (GCM) ID) for transmitting the generated recommended friend list to the user's mobile device Send friend information.

When the user enters the location where the beacon terminal is installed, the mobile client application 470 receives the Bluetooth signal transmitted by the beacon terminal and then transmits the information of the beacon terminal to the server in the XML format And a user interface module (UI) for confirming a notification message informing a collection module (Context Collector) and a list of recommended friends received from the server.

Multidimensional information can be modeled using tensors and matrices, and can be decomposed at the same time through the CMTF algorithm in order to find users with a common friend in the social network with similar lifestyle among multiple users in offline, have.

5 is a flowchart illustrating a method of providing a social network service according to an embodiment of the present invention.

The proposed social network service method includes a step (510) of receiving offline situation information of a plurality of users using a social network service through a preprocessor and evaluation result information on recommended friends, parsing and storing the information in a database (510) A step 530 of collecting, formatting and storing the friend relationship information of the users of the users in the database 520, converting the information stored in the database into the form of input information used in the tensor and matrix through the data converter, A step 540 of decomposing and recombining the tensors and the matrix simultaneously through the engine, and a step 550 of creating a recommended friend list for each user using the re-combined tensor through the friend list generating unit.

In step 510, the offline situation information of a plurality of users using the social network service through the preprocessor and the evaluation result information on the recommended friends are received and stored in a database through a parsing process.

The transmitted information is parsed through Context Data Parser and Evaluation Data Parser and stored in a database. The collected information is collected through a network crawler, It is stored in the database as <Source node ID, Destination node ID>.

In step 520, friend relationship information of the plurality of users is collected, formatted, and stored in a database. The friend relationship information of the plurality of users may be stored in a storage unit such as a user's place entry and exit time, friend relationship information, a user profile, an ID of a user's terminal (e.g., Google Cloud Message (GCM) ID) The result information is stored in a database in a plurality of tables.

In step 530, the information stored in the database is used to transform the data into the form of input information used in the tensor and matrix through the data converter. The information stored in the database is converted as shown in Fig. 6 in the form of < User, User, Place, Total visit count >

At step 540, the tensor and matrix are decomposed and recombined simultaneously through a recommendation engine. For example, you can write code using Matlab and use the Tensor Toolbox and the CMTF Toolbox.

In step 550, a recommended friend list for each user is generated using the re-joined tensor through the friend list generator. Then, a notification message (for example, a Google Cloud Message (GCM) ID) for transmitting the created recommended friend list is generated, and friend information recommended by the user to the mobile device is transmitted through the message creation and notification unit.

6 is a diagram illustrating a data transformation process for a join matrix and tensor decomposition according to an embodiment of the present invention.

The data modeling method for the proposed social network service fuses the user 's place visit record collected from the Bluetooth - based beacon terminal with the user' s social network friend relationship information through the combination matrix and the tensor decomposition method. For example, the user's place visit record can be represented by a 3D tensor XεR ^ (U × U × P), and the social network friend relationship information is represented by a two-dimensional matrix MεR ^ (U × U) . Here, matrix M is a joint matrix representing relationship information between two people, which is considered as additional information when the tensor X is decomposed and decomposed together.

6A is a diagram showing row data of entry / exit time input from a user mobile device and friend relationship route information of a social network. 6 (b) is a diagram showing information transformed into an input form of a tensor and a combining matrix based on FIG. 6 (a). 6 (c) is a diagram showing the modeling of the tensor X and the coupling matrix M, that is, the combination matrix of the user's place-use record and the friend relationship information of the social network and the form of the tensor modeling. The tensor X represents a first user-second user-user-place relationship and each element constituting the tensor X is derived from the frequency of two users using a particular location. When the user enters or exits the place where the beacon terminal is installed, the application installed in the user's mobile device receives the Bluetooth signal transmitted by the beacon terminal and then transmits the time and the unique code of the beacon terminal to the server. However, due to the characteristics of the Bluetooth radio signal, frequent signal breaks due to interference of peripheral signals, so that records entering or leaving are continuously transmitted to the server even if they stay in one place. In order to solve this problem, for example, if the interval between the entry time (EnterTime) and the leaving time (LeaveTime) is less than 5 minutes, the user is classified as staying in one place.

As a result, each user's place use record is accumulated, and each element of the tensor X, χ_ijk, is set to the smallest among the place use records of the users (User) x and y visited at a specific place. The large element value of tensor X means that the places visited by the two users have a great meaning to them and that the life patterns of the two users are similar. The join matrix M represents a first user-second relationship, which means the degree of proximity between the two users. Where the degree of proximity may be the degree of relevance between the first user and the second user derived based on, for example, the life pattern of first and second users or the number of common friends.

The connection score, which is an element of the matrix M, is derived by the number of common friends and the length of the connection relationship path between the two users. Equation (1) shows how the connection value reflecting this meaning is calculated, and Table 1 is a description of the parameters used.

수학식(1)

Equation (1)

<Table 1>

According to an embodiment of the present invention, the Katz metric, which is commonly used in link prediction, is modified to reflect the number of all paths existing between the two. However, the connection value is changed according to the hop length in order to reflect the attenuation of the proximity according to the number of hops of the paths considered. In other words, if the hop between two people is 1, it is judged that there is already a friend relationship in the real world, and the connection value is set to 0. When the hop is 2 or 3, the connection value is calculated using the equation (1) do.

When the hop is 4 or more, it is determined that the possibility of becoming a friend is reduced due to a lot of hops between two persons. In the case of 2 or 3 hops, only the minimum possibility is considered by setting the half of the calculated value. Equation (1) reflects the fact that the greater the number of paths between two people, the less the number of hops on the path, the more likely they are friends.

In order to calculate the connection value, it is necessary to consider all the paths between the two persons, but the user can not grasp the route when the friend information of the social network is made private. Therefore, since the friendship information of a social network is represented by a directed graph, the friendship information of the social network uses a symmetric characteristic in order to convert it into an undirected graph. In other words, in the case of a user who does not disclose the friend relationship information, the hidden friend relationship of the closed user is inferred by using the friend relationship information of the other users. We compute the friendship information represented by the non-directional graph into the depth-first search (DFS) algorithm and calculate all the paths between two hops and three hops.

In other words, multidimensional information is modeled with tensors and matrices to find potential friends who have a similar life pattern in offline and mutual friends and have common friends in OSNS, and are decomposed simultaneously using CMTF (Coupled Matrix and Tensor Factorization) algorithm. The CMTF algorithm uses a gradient-descent-based optimization method. The stop condition is set to 10 ^-8 and the maximum number of iterations is set to 10 ⁵ . Also, since it is an NP-hard problem to set a rank value for disassembly, it is necessary to gradually increase the rank value from 2 to 50, and to decompose and recombine to obtain the rank value at the lowest error value before and after decomposition Respectively. Algorithm 1 shows the procedure for this.

FIG. 7 illustrates a user interface according to an embodiment of the present invention. Referring to FIG.

When the user touches the notification message, the message is displayed through the UI as shown in FIG. The UI is made up of three tabs 710, 720 and 730, the first tab 710 being the social network profile of the recommended friend, the second tab 720 being shared between the user and the recommended friend And the last tab 730 shows a question item about a recommended friend.

For example, a beacon terminal is installed in 27 places (a laboratory, a rest room, a cafeteria, a restaurant, a gymnasium, a meeting place, etc.) on campus and 43 students (12 students (graduate students) And 31 respondents). The collected information was collected for one week for the users, and the questionnaire was evaluated. Satisfaction was assessed with the questionnaire items as shown in Table 2 for the recommended friends. Each item consists of three evaluation factors: 1) Yes, 2), 2) I do not know, 1), and 3) No (0).

<Table 2>

In order to increase the likelihood that a user accepts a friend recommended by a friend as a friend, the proposed model includes a combination matrix for analyzing a user's place use record occurring offline and friendship information of a friend connected to the social network, The tensor decomposition method was used and the user satisfaction scores collected through actual user experiments showed the validity of the proposed method. Although we used only place information and social network friend relationship information, the combination matrix and the tensor method can be considered together with more dimension information.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be embodyed temporarily. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI &gt; or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (19)

A data modeling method for a social network service of a social network service system including a data converter and a recommendation engine,
Deriving a plurality of elements constituting the tensor using the first user's information, the second user information, and the location information via the data converter; And
Decomposing a tensor comprised of the plurality of elements based on a matrix representing relationship information between a first user and a second user through a recommendation engine
Lt; / RTI &gt;
The data modeling method for the social network service includes:
Second information of users, second user information, location information of a first user and information of a second user by using a matrix and a tensor decomposition method that represent relationship information between first and second users to fuse information of different dimensions such as online and offline information Dimensional information including the relation between the time and the time is modeled using the tensor and the matrix, and simultaneously decomposed through a Coupled Matrix and Tensor Factorization (CMTF) algorithm
Data modeling method for social network services. The method according to claim 1,
Wherein the step of deriving the plurality of elements constituting the tensor using the information of the first user, the second user information and the position information through the data converter comprises:
When a first user or a second user enters or leaves a place where the beacon terminal is installed, the terminal of the first user or the second user receives the Bluetooth signal from the beacon terminal, To send the unique code of
Data modeling method for social network services. 3. The method of claim 2,
A record using a place where the first user or the second user is installed in the beacon terminal is accumulated in the server and the element values of the plurality of elements constituting the tensor are accumulated in the server to indicate the degree of similarity of the life patterns of the first user and the second user
Data modeling method for social network services. The method according to claim 1,
The connection score among the plurality of elements constituting the matrix is derived based on the number of common friends and the length of the connection relationship path between the first user and the second user, and the proximity between the first user and the second user Indicative of
Data modeling method for social network services. 5. The method of claim 4,
The connection value is varied according to the hop length to reflect the degree of proximity to the number of hops in the length of the connection relationship path
Data modeling method for social network services. 5. The method of claim 4,
In a case where the friend relationship information is a private user, in order to grasp the connection relationship path, the friend relationship information is converted into the non-direction graph using the symmetry property, and the friend relationship information is converted into the non- Friend relationship information is inferred from friendship of non-public users
Data modeling method for social network services. A preprocessor for receiving offline situation information of a plurality of users using the social network service and evaluation result information about recommended friends, parsing and storing the information in a database;
A database for collecting, shaping and storing friend relationship information of the plurality of users;
A data converter for converting the information stored in the database into a form of input information used in a tensor and a matrix;
A recommendation engine for decomposing and recombining the tensor and the matrix simultaneously; And
A friend list generating unit for generating a recommended friend list for each user by using the re-
Lt; / RTI &gt;
The recommendation engine includes:
Second information of users, second user information, location information of a first user and information of a second user by using a matrix and a tensor decomposition method that represent relationship information between first and second users to fuse information of different dimensions such as online and offline information Dimensional information including the relation between the time and the time is modeled using the tensor and the matrix, and simultaneously decomposed through a Coupled Matrix and Tensor Factorization (CMTF) algorithm
Social network service system. 8. The method of claim 7,
The pre-
The transmitted information is parsed through Context Data Parser and Evaluation Data Parser and stored in a database. The collected information is collected through a network crawler, Saved to database in the form of <Source node ID, Destination node ID>
Social network service system. 8. The method of claim 7,
The database includes:
Storing the friend relationship information of the plurality of users as a plurality of tables including the entrance and exit time of the user, friend relationship information, user profile, ID of the user terminal, and satisfaction evaluation result information about the recommended friend
Social network service system. 8. The method of claim 7,
The data converter includes:
To convert the information stored in the database into <User, User, Place, Total visit count> and <User, User, Connection score>
Social network service system. 8. The method of claim 7,
The recommendation engine includes:
Write code using Matlab and use the Tensor Toolbox and CMTF toolbox
Social network service system. 8. The method of claim 7,
Generates a notification message for transmitting the created recommended friend list, and transmits the generated notification message to the terminals of the users,
Wherein the social network service system further comprises: Receiving offline status information of a plurality of users who use the social network service through the preprocessor and evaluation result information on the recommended friends, and parsing and storing the received information in a database;
Collecting and formatting friend relationship information of the plurality of users and storing the same in a database;
Converting the information stored in the database into a form of input information used in a tensor and a matrix through a data converter;
Simultaneously decomposing and recombining the tensor and the matrix through a recommendation engine; And
A step of generating a recommended friend list for each user by using the re-combined tensor through the friend list generating unit
Lt; / RTI &gt;
The step of simultaneously decomposing and recombining the tensor and the matrix through the recommendation engine includes:
Second information of users, second user information, location information of a first user and information of a second user by using a matrix and a tensor decomposition method that represent relationship information between first and second users to fuse information of different dimensions such as online and offline information Dimensional information including the relation between the time and the time is modeled using the tensor and the matrix, and simultaneously decomposed through a Coupled Matrix and Tensor Factorization (CMTF) algorithm
How to service a social network. 14. The method of claim 13,
Receiving offline status information of a plurality of users using the social network service through the preprocessor and evaluation result information on recommended friends, and parsing and storing the received information in a database,
The transmitted information is parsed through Context Data Parser and Evaluation Data Parser and stored in a database. The collected information is collected through a network crawler, Saved to database in the form of <Source node ID, Destination node ID>
How to service a social network. 14. The method of claim 13,
Collecting and formatting the friend relationship information of the plurality of users and storing the same in a database,
Storing the friend relationship information of the plurality of users as a plurality of tables including the entrance and exit time of the user, friend relationship information, user profile, ID of the user terminal, and satisfaction evaluation result information about the recommended friend
How to service a social network. 14. The method of claim 13,
The step of converting the information stored in the database into a form of input information used in a tensor and a matrix through a data converter,
To convert the information stored in the database into <User, User, Place, Total visit count> and <User, User, Connection score>
How to service a social network. 14. The method of claim 13,
The step of simultaneously decomposing and recombining the tensor and the matrix through the recommendation engine includes:
Write code using Matlab and use the Tensor Toolbox and CMTF toolbox
How to service a social network. 14. The method of claim 13,
The step of generating a recommended friend list for each user using the re-joined tensor through the friend list generating unit may include:
In order to transmit the created recommended friend list, a notification message is generated through the message creation and notification unit and transmitted to the terminals of the users
How to service a social network. delete

Images