Predicting Identity Theft: A Comparative Assessment of Logistic Regression, Classification and Regression Trees, and Neural Networks Using Multi-Wave NCVS Data

• In Event: Advances in Comparative Methodology

Abstract

This study evaluates the comparative performance of three statistical techniques - logistic regression (LR), classification and regression trees (CART), and a multi-layer perceptron neural network (MLPNN) - in predicting identity theft victimization. Drawing on five waves of data from the Identity Theft Supplement (ITS) of the National Crime Victimization Survey (NCVS), we employ a multi-validation approach to assess model accuracy. Multiple evaluation indicators, including the sensitivity, specificity, overall accuracy, and the area under the ROC curve, are used to measure the predictive effectiveness of each technique. Our findings show that some models excel at identifying likely victims (high sensitivity) while others are better suited for correctly classifying non-victims (high specificity).

Authors

• Fawn Ngo, University of South Florida

• Anurag Agarwal, Florida Gulf Coast University

• Ramakrishna Govindu, University of South Florida