An Algorithm for Clustering Non-Deterministic Data

Abstract

This paper presents an algorithm for clustering non-deterministic discrete data for use in classification applications. The clustering algorithm analyzes input data and ranks the relative predictive strength of each value of each attribute. There are two levels of clustering: intra-attribute and inter-attribute. The values of each attribute are clustered to a level that provides the relatively deterministic dataset. The clustering technique is guided by the number of binary input units for the Artificial Neural Network (ANN) and training time requirements. Experimental results, on predicting the lapse intentions of Automobile Club of Missouri members, show that the networks trained with clustered data classify as well or better than networks trained with data clustered by other methods. The clustering algorithm improved the accuracy of the ANN algorithm over prediction accuracy achieved on raw data.

Department(s)

Computer Science

Document Type

Article - Journal

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2024 The Authors, All rights reserved.

Publication Date

01 Dec 1996

This document is currently not available here.

Share

 
COinS