Automatic Message Classification

Published

October 11, 2006

Author(s)

Arthur Griesser

Abstract

Rising automation means fewer humans, maintaining more equipment, which is exchanging more information. This information torrent is expected to increase even further, with the implementation of advanced process control and new information transfer standards such as Interface A and Interface C. The information deluge will make it difficult to pinpoint the information necessary for resolve problems. Sooner or later automated support for problem diagnosis will become valuable. Data classifiers(such as neural networks and Bayesian filters) typically require features extracted from the classified data. They are also typically trained: the classifier is provided with example data, and their categories. The selection of categories and samples for training is something of an art: under-training and over-training can easily result in poor classification. This presentation considers classification without training or domain knowledge.

Proceedings Title

AEC/APC Symposium XVIII

Conference Dates

September 30-October 5, 2006

Conference Location

Westminster, CO

Pub Type

Conferences

Keywords

Classification, Clustering, Pattern Recognition, Machine Learning

Citation

Griesser, A. (2006), Automatic Message Classification, AEC/APC Symposium XVIII, Westminster, CO (Accessed September 13, 2024)