Artificial Neural Networks for the Identification of the Differences Between "Light" and "Heavy" Alcoholics, Starting From Five Nonlinear Biological Variables

• NCJ Number: 170586
• Journal: Substance Use and Misuse Volume: 33 Issue: 3 Dated: (1998) Pages: 693-708
• Author(s): G Maurelli; M Di Giulio
• Date Published: 1998
• Length: 16 pages
• Publication Format: Article
• Publication Type: Survey
• Language: English
• Country: United States of America
• Note: DCC

Annotation

Three types of artificial intelligence systems were compared with respect to their ability to classify different kinds of alcoholics, using a set of data composed of 113 heavy and light alcoholics.

Abstract

The three evaluation systems were (1) a system based on genetic algorithms called BEAGLE, (2) seven different types of Artificial Neural Networks, and (3) a metasystem called MetaNet. The research sought to compare these system's groupings of the alcoholics into light and heavy alcoholics. Light alcoholics were those who drank an equivalent of less than 7 half-pints per day; heavy alcoholics were those who drank an equivalent of more than 7 half-pints per day; Results revealed that the MetaNet system had the best results. The next best results were produced by two Artificial Neural Networks: Squash and Logicon Projection. Results proved that the advanced elaboration data systems used in the social and health areas can be used in prevention programs that aim to reduce the social impact of certain pathologies correlated with different kinds of alcohol dependence. Figures, tables, author biographies and photographs, list of software used, and 14 references (Author abstract modified)