Attribute Relationship Solving Method Based on Nodes 1 and Communities in Opportunistic Social Networks 2

: The penetration of the 5G Internet and big data communication into human society brings about the 6 survival basis of the social opportunistic networks. Using mobile terminal devices for communication makes 7 the communication of nodes in the social opportunistic network intermittent, because nodes may be in motion 8 all the time. In social opportunistic networks, data communication activities can be recorded and analyzed by 9 evaluating communication activities of human beings or determining their interest points. However, the 10 identification of nodes with the same or similar types of attributes among a large number of user nodes, has 11 become a research problem in the field of social opportunistic networks. How to find an effective method to 12 classify nodes according to their social characteristics and similarity degree becomes the key point of social 13 opportunistic network data forwarding process. In this study, we proposed a method of community mining by 14 decomposition of node and community relationship matrix with large social network data attributes. By using 15 the regular type and iterative community features among community-rule-meet nodes, the method is proved to 16 be converged and yield a minimum solution. Experimental results show that the proposed method exhibits 17 strong application value.


21
The explosion of mobile devices in recent years has fueled the development of high-speed communications 22 and highly reliable networks [1][2][3][4]. With the popularization of communication devices, social interaction can be 23 seen everywhere. People can share their interests anytime and anywhere through mobile phones, tablets, smart 24 bracelets and other devices [13][14][15]. As a result, online social platforms like twitter and facebook have become 25 an indispensable part of human life [5][6][7][8]. As human beings interact socially in life, information is stored in 26 portable devices and connected and transmitted intermittently between devices along with human behaviors. 27 Therefore, in the social opportunistic network, data transmission between nodes needs to find "opportunity".

28
The information transfer process needs to look for opportunities meaning that only nodes considered 29 reliable can participate in the communication. The "storage-carry-forward" mechanism is a transmission 30 strategy of social opportunistic network [17][18][19][20][21][22][23]. The node stores the information to be transmitted in its own 31 cache area, carries the information for movement, and does not send the information to the node until it meets 32 the appropriate node. In the urban social scene, people with portable communication equipment represent nodes 33 in the social opportunistic network, so the social attributes of these people will have a great impact on the data 34 transmission strategy [25][26][27][28][29][30][31][32]. Data generated by human behaviors are of great significance for the selection and 35 improvement of information transmission strategies, so they have become the research hotspot of social 36 opportunistic networks.

37
However, a great number of online data that can be retrieved on the basis of human activities are complex 38 and may require extensive calculations. Moreover, many mobile devices that carry the information may 39 overload in big data online social opportunistic networks, they cannot receive or send any messages to others.

40
This characteristic may affect traditional methods in wireless communication networks.
We face big data communication in social opportunistic networks, challenges for nodes are high delay, 42 limited cache space, performing data update and improve delver ratio while appropriate neighbors can be 43 selected by us [33][34][35][36]. How to evaluate transmission states between nodes and neighbors is very important.

44
Data consume significant cache space and energy in devices when people use mobile devices during data 45 transmission and no suitable transmission range target is responding, which eventually causes transmission 46 delay [2]. Especially in big data social opportunistic network environment, where over-flooding and data 47 redundancy are used to create transmission, devices must distribute considerable cache space to save messages 48 [3]. A large number of awaiting information is stored in devices. Some information may be stored for a long 49 time without user acceptance and response status.

50
To avoid over-consumption, the identification of nodes with the same or similar attributes among a large 51 number of user nodes has become a research problem in the field of online social opportunistic networks. The 52 resolve relationship matrix of node and community (RRMNC) method is established in this study. This method 53 is used to conduct attribute decomposition of a large amount of social opportunistic network data by using 54 regular and iterative in the community features of nodes with the minimum solutions after convergence. These 55 nodes comply with community rules.

56
The main contributions of this study include the following:

57
(1) The rules for node iteration are established through the relationship matrix of nodes and communities.

58
(2) After demonstrating the convergence, the node that satisfies the minimum solution is identified. This 59 node exhibits a strong correlation with the community.

60
(3) Numerous experiments show that the proposed method exhibits strong application value.

61
This paper is divided into five chapters.Chapter1 introduces the study. Chapter 2 presents the related works.  In recent years, with the popularity of mobile devices, many researchers have invested in the study of 66 opportunistic social networks. The research on opportunity social networks mainly focuses on routing strategies, 67 making the opportunistic social network suitable for more application scenarios. Next, we will briefly introduce 68 the current status of several methods related to the research of this paper.

69
According to the social attributes and mobile features of nodes, some researchers proposed

80
The method divides communities by social attributes of nodes, and then adopts the strategy of community 81 reduction to remove inefficient nodes in the community, so as to improve the efficiency of data forwarding.     In the problem description, each node in social opportunistic networks can be explained its connected 155 matrix X and characteristic matrix Y . According to evaluate connection and characteristic, nodes can spend 156 little consumption to select appropriate neighbors. Matrix X and Y are non-negative matrices to ensure that 157 the community has many same close node links and attributes, and they can undertake joint decomposition of 158 X and Y, and assume familiarity with the common breakdown factor matrix H . X and Y can be 159 decomposed individually. For example, X can be transformed into three decomposed forms of T X HSH = 160 in community mining, where H is the community affiliation matrix and S is the community connection strength 161 matrix, it can judge relatedness with nodes in community. X is a symmetric matrix, and thus, S is also a 162 symmetric matrix. Therefore,  To prove the minimum value of convergence in a community, we must prove Formulas (7) and (8) 208 (2 ) 4( ) 2( ) 2 ( ) 4( ) ( )

242
To verify the effectiveness of the research method, four typical complex network data sets, including link 243 and attribute information, are selected for the experiment. The specific information of each data set is described 244 as follows：

245
(1) Political blog data set [13]. This data set contains 1490 nodes and 19,190 edges, where each node 246 represents a blog page about American politics and each edge represents the hyperlink relationship between 247 web pages. considering LANMF modeling complex network in the case of an undirected graph so ignore 248 the hyperlinks to sex, the last remaining 16175 side. Each node is associated with an attribute that indicates 249 the political orientation of the web page of the blog, which can either be liberal or conservative

250
(2) Citeseers data set [14]. This data set contains 3312 nodes and 36,141 edges.. Each node represents a piece 251 of literature of science and technology, each edge represents a reference to the relationship between science 252 and technology literature, and each node is associated with a class attribute. The general category attribute 253 value number is 6.

254
(3) CORA data set [15]. This data set is a Citeseer citation network data set with science and technology

266
For the analysis of the ideal community, the result shows that as density increases, entropy decreases, and 267 the higher the degree of the dominant node that will be found in the community.

Conflicts of Interest:
The authors declare no conflict of interest.