A hybrid expert system and neural network approach for environmental applications

An integrated knowledge-based system, RAISON, was developed for environmental applications, using expert system and neural network technologies. Practical considerations dictate that access and linkage with databases, spreadsheet, maps, documents, graphics and other modules are essential. It was found necessary to implement the expert system and neural network modules in the same software package, sharing common databases and a macro programming language to facilitate cross-linkage in hybrid systems. Some experiences with both techniques for solving acid rain problems, as well as the conversion from microcomputer to the workstation platform, are discussed. Comparison of results with independent data sets is found to be about 20 to 25% in error which is well within the measurement uncertainties.<<ETX>>