Title:

Maximizing the Spread of Influence in Temporal Social Networks

Title in english:

Maksymalizacja wpływu społecznego w temporalnych sieciach społecznych

Creator:

Michalski, Radosław

Contributor:

Kazienko, Przemysław. Promotor

Subject and Keywords:

wpływ społeczny ; maksymalizacja wpływu społecznego ; strategie wyboru węzłów początkowych ; sieci temporalne ; sieci społeczne ; analiza sieci społecznych ; social influence ; social influence maximization ; seed selection strategies ; temporal networks ; social networks ; social network analysis

Abstrakt:

People do not live in isolation. They are constantly exchanging information, they form groups, establish families. Nowadays they are also overwhelmed by the constant media stream. All of the above influence ones opinions, visions and attitudes. By looking at human relationships from a social network perspective, these opinions spread through the network convincing some people to share them and pass to their neighbours, and others to discard or disagree with them. This process, named "the spread of influence", is happening all the time and following it is one of the challenges. The other challenge, definitely more difficult, is to predict whether a particular opinion will spread across the network. The third aspect discussed in this dissertation, is how to choose an initial set of nodes - the seed - in order to maximise the overall spread of influence in the social network. In contrast to the state-of-the-art in this area, the dissertation concerns a problem of the dynamics in social networks. As we amplify or reduce the intensity of contacts with others over time or sometimes even completely change our environment, the assumption that the social network mapping these relationships is static is too strong simplification of reality. That is why the temporal aspects of networks are influencing the whole process strongly and discarding that information may lead to wrong conclusions about whom to influence in the beginning. Moreover, as contacts between people are ordered in time, it is impossible to look at this process as at a reversible one - information cannot travel against time. The static approach assumes this as well by saying that relationships established some time ago, which ceased over time, may be used for passing information to others. This assumption is simply wrong. In this work it has been studied how the spread of influence behaves in temporal social networks and whether it is possible to introduce a new approach outperforming typical strategies in this area by making use of the temporal nature of networks. The model of social influence studied in this dissertation is Linear Threshold, one of two most commonly used in the research on social influence. In order to face common problems in the area, a new method providing better results is proposed and evaluated in this dissertation - the tInf method. It proves that instead of simplifying the reality it is better to make use of it in order to achieve more satisfying results. However, it should be stated, that this approach is not a win-win solution, since it also introduces a new set of challenges that should be solved to take full advantage of it.

Publisher:

Politechnika Wrocławska

Place of publication:

Wrocław

Date:

2014

Resource Type:

rozprawa doktorska

Source:

<sygn. PWr 379258>

Language:

eng

Relation:

Politechnika Wrocławska. Wydział Informatyki i Zarządzania

Rights:

Wszystkie prawa zastrzeżone (Copyright)

Access Rights:

Dla wszystkich w zakresie dozwolonego użytku

Rights Owner:

Własność autora

Location:

Politechnika Wrocławska