Evolutionary algorithms for knowledge discovery and model-based decision support

The COMAX expert system contains rules derived from explicit knowledge of t he growth and development process of the cotton plant It has been used for a number of years with the GOSSYM cotton model to provide a decision suppor t tool for agricultural use. Among other things, it is capable of producing advice about irrigation of the cotton crop in order to maximize profitable yield. An alternative approach is proposed, in which an Evolutionary Algor ithm (EA) is used to evolve an irrigation schedule. The EA contains no expl icit representation of botanical knowledge; instead, it utilizes a fitness function which evaluates the effectiveness of each evolved schedule by runn ing the GOSSYM model. The results produced by the model predict the yield a nd profitability of this irrigation schedule, which is used to determine th e relative goodness of the corresponding chromosome. A comparison with the irrigation schedules produced by the COMAX expert system indicates that the EA approach is able to produce better schedules, which increases the profi tability of the cotton crop. This suggests that evolutionary-based techniqu es may find other applications within the field of model-based decision sup port.. Copyright (C) 1998 IFAC.