Predicción del alcance de propagación de información e influencia en redes sociales online
- Autores: Raul Marcelo Ortiz Gaona
- Directores de la Tesis: José Luis Melús Moreno (dir. tes.), Marcos Postigo Boix (codir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2020
- Idioma: español
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Online social networks (OSNs) are increasinglyused by different sectors of societywith the purpose of informing and influencing people. These sectors are: govemments, political parties, non-govemmental organizations, suppliers ofgoods and services, etc. Theywish they could predict the extent ofthe spread of information and the influence theytransmit through these networks. Currently there are many m athem atical models that allow us to predict the scope of these propagations, many ofwhich are based on Linear Threshold Nlodel (L TM) or lndependent Cascad e Model (ICM).
In this document we present Lucy Nlodel, a new predictive mathematical model that incorporates several elements that have not been considered or taken into account in the base models orín their derived models, but which are present in real life and affect the extent of propagation of messages on online social networks. These new elements are the following: We distinguish three classes of messages: pure.ly informative messages; messages that are intended to influence appealing to people's emotions and feelings; and messages that are intended to influence by appealing to personal interests and conveniences. We argue that the intensity of the relationship between individua Is (tie-strength) depends on the relevan ce and affinitybetween them . Toe probabilitythat the message is received byan individual is directlyproportional to this relationship intensity. Also, we consider that a message with greater im portance is more likely to spread among people than a message with less importance. Finally, we maintain that individuals are informad or influenced (activated) after exceeding a threshold, which depends on the type and importance ofthe message and the relevance or affinity, as the case may be, between individuals.
On the other hand, we determine a way to quantify the affinity between nodes and the strength of the relations hip between nodes, and we link this to the probabilitythat a node receives a message.
To validate the results of LM, we performed laboratorytests of LM, L TM, ICM and WCM with a Facebook network tracked of an anonymous user. These results have been comparad with empirical tests ofinformation propagation and influence on the Facebook platform in the network of the same anonymous user. LM represents the message propagation process in a finer way and its results are notab(y closer to the empirical results in comparison with the results that were obtained with the models L TM, ICM and WCM. In order to study the behavior of LM, we carry out additional laboratorytests, modifying its parameters and using a synthetic (artificial) Power-Law network and two real networks (Facebook and YouTube).
LM is suitable to be used in different spheres of human activity, for example: 1- lt will allow to know more preciselythe different sectors of society about the impact produced bythe messages that they publish in the OSNs. 2- lt can be incorporated into the maximization models ofinfluence propagation in social networks. 3- lt can be used in models that predict the loss of customers in mobile telephone services. 4. lt can also be used in systems that simulate the spread of informatics virus es in computer networks, cell phone networks and online social networks.
