EVALUATION AND CLASSIFICATION OF THE ELECTRICAL HAZARD OF CHEMICAL VAPORS DURING WATER TRANSPORTATION USING PATTERN RECOGNITION TECHNIQUES

Pattern recognition is a statistical technique that allows one to find or predict a property of chemicals that is not directly measurable, but is known to depend upon certain features or properties of the chemicals via some totally unknown relationship. This technique has been applied to a multitude of scientific problems. The same technique was used to classify a chemical according to its relative hazard in bulk water-transportation based on chemical structure and macro-scale properties such as density, vapor pressure, structure-fragments, solubilities, etc. Using the Linear-Learning Machine, the overall prediction of the 47 compounds in training set was 68% correct. The predicted classifications of the 240 compounds in the test set are approximately 68% correct. There are many difficulties associated with properly classifying compounds on the basis of variable derived from structural fragments that must be solved before great reliance can be placed on the results of a Linear-Learning Machine classification.

  • Supplemental Notes:
    • Prepared for Department of Transportation, United States Coast Guard Office of Research and Development, Washington, D.C. 20590.
  • Corporate Authors:

    University of Rhode Island, Kingston

    Department of Chemistry
    Kingston, RI  United States  02881
  • Authors:
    • Fasching, J L
    • Stromberg, E W
    • Weisel, C P
  • Publication Date: 1979-1

Media Info

  • Pagination: 90 p.

Subject/Index Terms

Filing Info

  • Accession Number: 00195756
  • Record Type: Publication
  • Source Agency: United States Coast Guard
  • Report/Paper Numbers: CG-D-16-79 Final Rpt.
  • Contract Numbers: DOT-CG-44160-A
  • Files: TRIS
  • Created Date: Jul 31 1979 12:00AM