Assessment of Logistics Regression for Classification of Drug Data

Full Text PDF PDF
Author(s) S.S. Abdulkadir
Pages 505-508
Volume 1
Issue 10
Date September, 2012
Keywords Logistic regression, classification, odds ratio, discriminant Analysis.
Abstract

The classification of individuals who involve in illicit drugs into drug peddlers and non-peddlers on the basis of oral evidence, type and the quantity of exhibit found with them usually pose problem for the purpose of prosecution. This paper uses logistic regression to classify offenders into the two dichotomous dependent variable (peddlers or non-peddlers) in order to ease the work of the agency, National Drug Law Enforcement Agency (NDLEA), responsible for illicit drugs. The data used in this study include age of offenders, type and weight of exhibit collected from the agency. The discriminant analysis was first used by Abdulkadir and Emmanuel (2010) to classify the illicit drug offenders into peddlers and non-peddlers based on the data. In literature the author discovered that the discriminant analysis cannot handle mixed data (discrete and continuous data) effectively instead logistic regression better fit the data. The model correctly classified 95.6% original grouped cases with positive and negative predictive values 97.35% and 91.55% respectively, which are higher than values obtained under the discriminant analysis. It was discovered that large quantity of exhibit has more effect than other variables, yet their inclusion significantly improves the outcome of the findings.

< Back to Oct Issue