Grafos de mundos pequenos e difusão de inovações

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

This is a research in the area of Information Technology about how are the innovations disseminated in society. Taking the work of Everett Rogers (1962) about "diffusion of innovations" as reference, the study takes p1ace in the interface of this area with the fields of Sociology and Marketing, allowing evaluating how do the changes or new concepts are received or rejected in groups of different natures. The representation of how do the innovations penetrate in these populations has been done by simulations based on graphs, where each vertex represents one member of the population. In the studies of Rogers and other researchers, prevails a random shape of interconnection of these members, here known as "random world". But recent1y, Watts (1999) discovered that, under certain conditions, members of groups interconnect in such a way where all of them have on1y a few intermediates between each other, in contexts where there are neighborhoods with many members already connected. Still today, it s studied the causes for this phenomenon, known as "small world". The objective of this work is to verify how is the behavior of the curve of "diffusion of innovations" in contexts with these characteristics and identify the differences regarding the model where the interconnections were random. The first results showed that, in the majority of the cases here analyzed, the velocity of acceptance of innovations (performance) is lower in contexts of "small world", if compared to those from "random world". But, after redesigning models of "small world" based on the same definitions of Watts and extracting the randomness, we found a superior performance, even though it s more sensible than versions of Rogers and Watts regarding the characteristics of the population. We concluded that the adoption of some models of "small world" (if with good parameters), can be more productive and allow the preservation of characteristics of real world, that means, to obtain a more "natural" effect We tested different parameters, finding advantages in the following: the degree of independence of new connections regarding the previous connections (parameter a), the random probability of connection between any two members (parameter p), the medium quantity of connections per member (parameter km), the selection of innovators members as local leaders (they have connection with other groups and high index of connections with their own group). The model used here for simulation, with little adjust, was the "multi-agents" idealized by Maienhofer and Finholt (2000)

ASSUNTO(S)

algoritmos redes neurais teoria dos grafos arquitetura de redes

ACESSO AO ARTIGO

Documentos Relacionados