Some measures for the evaluation of the predicitive capacity of a classification model

Josimara MAZUCHELI [1]

Francisco LOUZADA-NETO 1

Lorene GUIRADO 1

Edson Zangiacomi MARTINEZ[2]

§    ABSTRACT: In this article we present some of the usual measures considered in the evaluation of the predicitve capacity of a classification model, among the which, the sensibility, the specificity, the negative and positive predictive values and to acuracy. Also we describe the ROC curve, that consists of a graphic representation of the performance of a model, according to the assembly of its possible answers. The procedures are illustrated in an real data set considering a logistics regression model.

§    KEYWORDS: Classification model; predictive capacity measure; sensibility; specificity; negative and positive predictive values; ROC curve.



[1] Departamento de Estatística, Universidade Federal de São Carlos -- UFSCar, CEP: 13565-905, São Carlos, SP, Brazil. E-mail: dfln@power.ufscar.br

[2] Departamento de Medicina Social, Faculdade de Medicina de Ribeirão Preto -- FMRP, Universidade de São Paulo -- USP, CEP: 14049-900, Ribeirão Preto, SP, Brazil. E-mail: edson@fmrp.usp.br