Methods and techniques used in the development of an expert system for insect sampling plan design

Abstract

An expert systems prototype for insect sampling plan design in the New Zealand apple orchard environment was developed. The prototype advises on the preliminary design, analysis of the preliminary data and the main sampling plan design for insect sampling on apple trees.

A knowledge acquisition technique to overcome the lack of readily available case histories in this field was developed. The technique was used to specify the possible range of problems within insect sampling plan design, by identifying the factors that contribute to a particular design. The range of each factor was determined and used to simulate test cases for preliminary and main sample plan design. The technique is suitable for other areas of knowledge in which the solutions are diagnostic or classificatory.

The statistician-entomologist relationship was investigated by simulating eight advisory-client interactions. Analysis of these showed that statisticians controlled the relationship. The interaction can be characterised by a three-part model: information-collecting, advice-giving, and closing. In three quarters of the interactions the statistician cycled between the information-collecting and advice-giving episodes. The function that this cycling serves is unclear.