Optimization of link prediction in social networks with the help of fuzzy logic

Continuous Master's Course in Computer Engineering

Abstract

Today, the popularity of social networking sites among people is undeniable, sites that provide users with many possibilities for communication between people. One of the basic problems in the analysis of this type of networks is the prediction of new connections between network people. The fuzzy method, as one of the popular methods in artificial intelligence, provides a simple way to make clear, ambiguous, noisy and missing results. As a result, fuzzy logic has become a tool for modeling the complexities of the real world. These models are usually much more accurate than their similar ones and provide us with more accurate results. For this reason, fuzzy logic will have the necessary potential to provide a more accurate proposed link, and the framework that we will present in this research will be developed based on this logic.

According to the above approaches, in this research, we tried to improve the results by presenting a proposed framework to provide an intelligent algorithm based on the combination of fuzzy logic with CN, Jaccard, and PA algorithms, which are algorithms for predicting links in the social graph. Examining the results showed that the proposed algorithm is more accurate in link prediction, but due to the presence of fuzzy and defuzzification steps, it has a lower speed.

Key words:

link prediction in social networks - fuzzy logic - link prediction algorithms based on similarity

Chapter 1

Introduction and problem statement

Introduction:

Social networks are a new generation of websites that have been in the focus of the users of the global Internet these days. Such sites operate based on the formation of online communities, and each one gathers a group of Internet users with a specific characteristic. Social networks are considered as a type of social media that have made it possible to achieve a new way of communicating and sharing content on the Internet. Hundreds of millions of Internet users are members of hundreds of different social networks, and part of their daily online activity takes place on these sites.

Predicting the occurrence of links is a basic and fundamental issue in networks. In the topic of link prediction, a view of a network is given and we want to know what transactions are likely to happen between the current members of the network in the near future or which of the existing transactions we will lose. Although this issue has been widely studied and investigated; However, the problem of how to optimally and effectively combine the information obtained from the network structure with the abundant descriptive data related to nodes and edges remains to a large extent. L. Backstrom, , 2011))

To model social networks, they use a graph in which people form nodes and relationships between people are represented by edges. Meanwhile, a large social graph is created.

In this work, we will try to predict possible connections by analyzing social networks. Communication prediction is a sub-branch of social network analysis, in which a series of connections that are not directly visible or do not exist must be inferred or guessed according to the observations and existing connections.

In this chapter, the reason for addressing the project and its problem is examined. For this purpose, first, introductions related to social networks, social graph, fuzzy logic, link prediction in social networks are briefly presented, then the problem that the thesis tries to solve is presented, and finally the approach used in the research and the structure of the thesis are presented.

The main goal of the research is to examine the existing methods in the field of predicting new communication links in social networks and to provide a new solution with the help of fuzzy logic in the field of link prediction in the social graph.

1-2- Social networks (Parhizkar, 2012)

In recent years, the Internet space has become more important in people's daily lives. People use the Internet to communicate with others, buy and sell products electronically, search for information and do many other things, and in this way, the Internet has become a vast social network.

A social network is a social structure composed of nodes (generally individual or organizational) connected by one or more specific types of dependencies, for example: prices, inspiration, ideas and financial exchanges, friends, kinship, trade, web links, disease transmission (epidemiology) or airline routes. The resulting structures are often very complex. Social networks are groups of people or organizations with common tastes or interests that come together to achieve certain goals. Social network analysis looks at social relationships with the terms vertex and edge. Each member is called an actor. The characteristic of social networks is the existence of complex relationships [2] and interactions [3] between actors. Vertices are individual actors in networks and edges are relationships between these actors. Many types of edges can exist between vertices. The results of various researches indicate that it is possible to use the capacity of social networks at many individual and social levels in order to identify problems and determine their solutions, establish social relations, manage organizational affairs, make policies and guide people in the path of achieving goals. For example, the results of studies in the field of tourism policy show that social networks are effective in attracting foreign tourists to various destinations by influencing behavioral variables, and these networks can be used in order to build trust and reduce the risk of users' decisions in choosing a specific tourist destination. The social network can also be used to identify the social status of each actor. These concepts are often shown in a social network graph, where points are vertices and lines are edges.

One of the main reasons for the formation of social networks are personal relationships, work relationships, scientific relationships, shared tastes and interests, and socio-political motives. (Parhizkar, 2012)

1-3-Analysis of social networks [4] (Parhizkar, 2012)

Analysis of social networks means the study of the characteristics of social networks and relationships between people and parts of a network with a network or graph theory approach.

Analysis of social networks is a type of interdisciplinary study in various fields, including: sociology, information science, Communication sciences, management and organization, anthropology, geography, social psychology, linguistics, epidemiology, economics and business.

Social network analysis is a research approach that deals with the patterns of relationships between individuals, groups and organizations and was first formed in sociology, social psychology and communication sciences. Web analytics scientists have used this approach in web studies and have analyzed social networks in the continuous environment. They believe that computer networks can be studied based on social network analysis because they can connect people and organizations together. Many other types of social networks have been formed in the virtual environment and are expanding every day. In addition to dealing with individuals (individuals, organizations, states) as discrete units of analysis, the analysis of social networks also focuses on the structure of disciplines that affect individuals and the relationships between them. Partially strong networks are less useful to their members than networks that have many weak ties to people outside the core network. More open networks with social connections and loose threads provide more opportunities to access new ideas and achievements than closed networks with long threads. In other words, a group of friends who only have connections with each other, share the same information and achievements. But a group of people who have connections with other social sectors have more chances to access a wider range of information. For people to achieve success, it is better to connect with various networks than to have many connections within one network. Similarly, individuals can practice influencing and acting as mediators in communication between two networks that are not connected to each other (this is called filling structural holes).