An expert system for making fertilizer recommendations

Project reports the results of research undertaken to develop an expert system for making fertilizer recommendations based on the results of soil analysis.. Developing fertilizer recommendations based on the results of soil analysis is recognized as a standard practice in horticultural crops. The manual process requires expert knowledge and can be tedious and time consuming. This thesis reports the results of research undertaken to develop an expert system for making fertilizer recommendations. Study objectives were to explore and describe methods for the implementation of the interpretation model and the generation of fertilizer recommendations using a user interface that is highly interactive and effective as well as demonstrating the commercial feasibility of such a system. Incremental prototype development occurred both in the DOS and Windows environment, with the Windows version being more commercially acceptable and maintainable than the DOS version. The review of the prototypes by a small team of domain experts proved to be a successful method for iterating through the prototype development system life cycle. At the core of the system is a knowledge base maintained in tabular format and a linear model which resolves an optimum fertilizer recommendation solution that meets certain goals while satisfying a set of constraints. The provision of mixed initiative dialogues that support userexploration of the solution space using the results from the linear model is also reported. Pilot commercialization through a jointly funded project by the co-operating fertilizer company, Incitec and the Horticultural Research and Development Corporation was successfully undertaken and full commercialization is being funded fully by Incitec. Results from this study indicated that the system developed was commercially acceptable as a delivery platform for fertilizer recommendations.