An Expert System for Raising Pigs

(File Last Modified Wed, May 29, 2002.)

• Review: An Expert System for Raising Pigs
• Problem Addressed
• Approach Proposed
• Validation
• Contribution
• Future Directions
• Comments

Review: An Expert System for Raising Pigs

Problem Addressed

AUSPIG is a successful MS-DOS tool (a mathematical modelling package) for studying the performance of pigs growing in pens in a piggery. A problem was that it required an expert to fully utilise the output of the mathematical models. What is needed is an itelligent post-processor to analyse the output. Besides its primary function, the post-processor should also satisfy the following:

• The expert shell can be integrated without inflate the whole package cost.
• It is possible to continue the expert system after the feasibility study.
• The developers wanted a high-level symbolic-procesing language that could be easily modified to handle a variety of inferencing techniques.

Approach Proposed

• PROLOG was selected as the high-level symbolic processing language.
• First, a rapid prototyping methodology was adopted to create a DSL which allowed the expert to write rules.
• A forward chaining rule Interpreter was built for that DSL.

Validation

• pigE is assessed according to the speed at which it reaches a conclusion and how much its recommendations can improve the profit per pig.
• The expert system out performed the human expert by 6.5 percent and increased the profit per pig by 227 percent.
• When extended the use to cover the profit of the entire piggery, the difference between the expert system and human experts are further increased.
• In all cases, the expert system out performed the human expert.

Contribution

• Heuristic optimisation provides a superior method of optimising complex systems like AUSPIG.
• The use of the application language approach (DSL) simplified knowledge acqusition, thus breakthrough the bottleneck in building expert systems.
• The division of the rules into knowledge bases and rule groups simplifies the maintenance problem.

Future Directions

• Human users could be assessed on a continuum of expertise and the expert system could adapt appropriately.
• Remove data entry if the expert system could provide default parameter values for user to comfirm or override.

Comments

Dr. Menzies regards this paper as his best paper in all times. I can not agree with him more. This paper, although very short, provide as much information as possible. The content is concise and to the point, in depth but not abstract, comprehensive but not confusing. It is well orgnized and well written. Reading papers would no longer be a pain in back if all papers are like that.

• pigE is a successful paradigm of commercial application of DSL.
• PROLOG's excellent adaptablity and extendability make it a powerful tool for expert system implementation.
• I think the most successful point of pigE, besides its prominent performance, is its ability to liberate the maintenance, which is a very serious issue after a software has been delivered. Can this DSL approach extend to domains other than expert system? I am not sure. It seems to me that DSL is suitable to be a restricted customized development enviroment between domain experts/developers and the actual end-user application. It is not a solution for commercial application but a solution for a intermediate layer or a post-processor.