A fact data base (FDB) was introduced to construct a new expert system to support tomato cultivation by using an AI-computer. In this expert system, inference for cultivation management of tomato plants can be optimized by interactions between forward inference and backward inference.
When a user of this expert system asked an arbitrary question concerning tomato cultivation, an inference engine of the computer began to acquire data related to the question from FDB. First, the computer conducted the forward inference, and the obtained result could be stored temporarily in the meta-working area of the computer memory. Furthermore, results obtained from the forward inference were used for backward inference, and inference could be optimized for the question.
In order to test this new expert system, FDB's for physiological data for tomato plants and climatic conditions for various cultivation locations were constructed in the computer memory and 836 rules for inference were set in the inference engine. When optimal times for seeding and transplanting for tomato plants were estimated for 41 different locations in Japan by using this expert system, time inferred by the computer was similar to that practiced actually in the locations, indicating that this expert system can be used for optimization of cultivation management.
Although a few deviations were observed, such deviation could be corrected readily, because FDB could be expanded easily. Thus, a new expert system developed here by using FDB may be used for other crops, if sufficient FDB and rules for inference are set appropriately.
抄録全体を表示