Aplicação do modelo binomial aos dados do sítio Consumidor.gov.br

Abstract

In this work we used the public database of consumer website federal government (consumi- dor.gov.br), which is a public service for alternative solution of consumer disputes (conflicts between consumers and businesses). This database contains customer information and complaints. Once registered the individual complaint, the company will have ten days to respond to the claim; and subsequent to that the consumer will have a response within twenty days to sort his complaint as "Resolved" or "Unresolved". In practice however, it is observed that there are many complaints which the individual does not evaluate; Difficult to obtain the actual value of the settlement index calculated by the site, which is the proportion of complaints resolved in relation to the total complaints by company. Thus, the government considers complaints not assessed as "resolved" to calculate this index; this work fits a binomial model in conjunction with Criterion for obtainment cutoff points for classification of not evaluated complaints as "resolved" or "not resolved"; and thus could calculate the solution index. Then, there is a comparison between the obtained solution indexes, the index calculated by the government and also other indexes obtained using the classification used in the binomial model. When compared, certain Heterogeneity between the solution indexes was observed.