Publication: PASS: a simple classifier system for data analysis
Loading...
Identifiers
Publication date
1993-09
Defense date
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Let x be a vector of predictors and y a scalar response associated with it. Consider the regression problem of inferring the relantionship between predictors and response on the basis of a sample of observed pairs (x,y). This is a familiar problem for which a variety of methods are available. This paper describes a new method based on the classifier system approach to problem solving. Classifier systems provide a rich framework for learning and induction, and they have been suc:cessfully applied in the artificial intelligence literature for some time. The present method emiches the simplest classifier system architecture with some new heuristic and explores its potential in a purely inferential context. A prototype called PASS (Predictive Adaptative Sequential System) has been built to test these ideas empirically. Preliminary Monte Carlo experiments indicate that PASS is able to discover the structure imposed on the data in a wide array of cases.
Description
Keywords
Regression analysis, Classifier systems, Machine learning