Dynamic forecasts of dairy farming structure; using expert systems

许多地方农业合作社和政府都迅速建立了农业数据和其他相关农业信息的收集和数据服务系统。但由于数据本身的局限性和适合利用这些数据的方法不多，使得这些数据在区域农业政策规划和区域农业结构预测中的利用率不高，而将区域领导的经验和知识引入到专家系统中，可以更有效、更灵活地预测区域经济形势。引入人工智能规划，提高了马尔可夫法对奶牛场规模结构的预测能力。专家系统利用专家的知识构造转移矩阵。系统动力学模型也被视为地方政策决策过程的工具。DEMATEL方法对于明确地方农业区域的政策目标是非常有效的，本研究还对奶牛养殖业的规模经济进行了实证分析。并对制约农户规模扩大的条件，如家庭劳动力的限制、采用的技术和经营能力的差异等进行了分析。通过对北海道和纽约州奶牛场数据的比较表明，采用奶牛养殖新技术、自由畜舍和挤奶厅系统，需要有足够的劳动力来扩大家庭劳动力。这些结果可作为重要信息纳入专家系统，特别是在利用人工语言进行动态预报方面，目前研究还没有完成。本研究正是展示了这种方法和原型的可能性。

The collecting and data service systems of Farm data and other related agricultural information have been established rapidly in many local agricultural cooperatives and governments. But the data is not so much utilized in planning regional agricultural policy and forecasting regional agricultural structure, because there are not so many methods which are appropriate to utilize these data and the limits of data itself.The regional leaders' experience and knowledge are introduced into the expert system to forecast regional economic situation more efficiently and flexible. Forecasting ability of the structure of dairy farm scale using Markof method has been improved introducing artificial intelligence programming. The knowledge of experts is used to make transition matrix in the expert system. The system dynamics model is also examined as a tool of local policy decision making process. The DEMATEL method is very effective to make clear the target of policy in local agricultural region.The positive analysis have also done in this study to test the scale of economy in dairy farm sector. The conditions to restrict enlarging the farm scale, such as restriction of family labor, the difference of adopted technique and managerial ability are also examined. The comparison of dairy farm data between Hokkaido and New York State showed that enough labor extending family labor is necessary to adopt new dairy farming technology, Free stall barn and Milking parlor system. These results might be included in expert systems as a important information.The study has not completed yet especially in dynamic forecast using artificial programing languages. We showed just the possibility of such method and proto-type in this study.