Fermentation database mining by pattern recognition

摘要: A large volume of data is routinely collected during the course typical fermentation and other processes. Such provide required basis for process documentation occasionally are also used analysis improvement. The information density these often low, automatic condensing, analysis, interpretation ("database mining") highly desirable. In this article we present a methodology whereby variables processed to create database derivative quantities representative global patterns, intermediate trends, local characteristics process. powerful search algorithm subsequently attempts extract specific their particular attributes that uniquely characterize class outcomes such as high- or low-yield fermentations.The basic components our pattern recognition described along with applications two sets from industrial fermentations. Results indicate truly discriminating do exist in they can be useful identifying causes symptoms different outcomes. has been implemented user-friendly software, named db-miner, which facilitates application efficient speedy data. (c) 1997 John Wiley & Sons, Inc. Biotechnol Bioeng 53: 443-452, 1997.