The network organisation of consumer complaints

L. E. C. Rocha^1a and P. Holme^1,2

¹ IceLab, Department of Physics, Umeå University - 90187 Umeå, Sweden, EU
² Department of Energy Science, Sungkyunkwan University - Suwon 440-746 Korea

^a luis.rocha@tp.umu.se

Received: 20 April 2010
Accepted: 7 July 2010

Abstract

Interaction between consumers and companies can create conflict. When a consensus is unreachable there are legal authorities to resolve the case. This letter is a study of data from the Brazilian Department of Justice from which we build a bipartite network of categories of complaints linked to the companies receiving those complaints. We find the complaint categories organised in an hierarchical way where companies only get complaints of lower degree if they already got complaints of higher degree. The fraction of resolved complaints for a company appears to be nearly independent of the equity of the company but is positively correlated with the total number of complaints received. We construct feature vectors based on the edge-weight —the weight of an edge represents the times complaints of a category have been filed against that company— and use these vectors to study the similarity between the categories of complaints. From this analysis, we obtain trees mapping the hierarchical organisation of the complaints. We also apply principal component analysis to the set of feature vectors concluding that a reduction of the dimensionality of these from 8827 to 27 gives an optimal hierarchical representation.