Empirical analysis of case-based reasoning and other prediction methods in a social science domain: repeat criminal victimization

  • Authors:
  • Michael A. Redmond;Cynthia Blackburn Line

  • Affiliations:
  • Mathematics & Computer Science, La Salle University, Philadelphia, PA;Department of Law & Justice Studies, Rowan University, Glassboro, NJ

  • Venue:
  • ICCBR'03 Proceedings of the 5th international conference on Case-based reasoning: Research and Development
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Case-Based Reasoning (CBR) has been used successfully in many practical applications. In this paper, we present the value of Case-Based Reasoning for researchers in a novel task domain, criminology. In particular, some criminologists are interested in studying crime victims who are victims of multiple crime incidents. However, research progress has been slow, in part due to limitations in the statistical methods generally used in the field. We show that CBR provides a useful alternative, allowing better prediction than via other methods, and generating hypotheses as to what features are important predictors of repeat victimization. This paper details a systematic sequence of experiments with variations on CBR and comparisons to other related, competing methods. The research uses data from the United States' National Crime Victimization Survey. CBR, with advance filtering of variables, was the best predictor in comparison to other machine learning methods. This approach may provide a fruitful new direction of research, particularly for criminology, but also for other academic research areas.