权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Machine Learning Approaches to the Analysis of Organizational Behaviors

组织行为分析的机器学习方法

基本信息

批准号：
05680287
负责人：
TERANO Takao
金额：
$ 1.28万
依托单位：
University of Tsukuba
依托单位国家：
日本
项目类别：
Grant-in-Aid for General Scientific Research (C)
财政年份：
1993
资助国家：
日本
起止时间：
1993 至 1994
项目状态：
已结题

项目摘要

The high productivity of Japanese production systems are well-known. Various analyzes have been carried out to explain the principles. However, conventional organization and management theory has not succeeded in the explanation. The theory should formally explain the mechanisms of such typical activities in Japanese companies as Kaizen, Nemawashi, and so on. The important but difficult features of these activities are that they heavily rely on informal information processing among members and cannot be quanititatively measured.Recent advances in Artificial Intelligence have made it possible to re-examine Simon's approaches with physical symbol systems hypothesis. To develop a rigorous theory on organizational learing, therefore, AI symbolic approaches are promising because of their descriptive powers and capabilities of computer simulation.In this project, first we discuss the requirements of AI models applicable to organizational theory.Second, in order to facilitate the analyzes, we propose a computational model : LPC.The model consists of a set of agents with (a) a knowledge base for learned concepts, (b) a knowledge base for the problem solving, (c) a prolog-based inference mechanisms, and (d) a set of beliefs on the reliability of the other agents. Each agent can improve its own problem solving capabilities by inductive and/or deductive learning on the given problems and by reinforcement learing on the reliability of communications among the other agents.Several experimental systems of the model have been implemented in CESP and Prolog languages. Experiments were carried out to examine the feasibility of the machine learning mechanisms of agent for problem solving and communication capabilities. The experimental results suggest that the proposed model is executable for analyzing the learning mechanisms applicable to distributed knowledge systems.

日本生产系统的高生产率是众所周知的。已经进行了各种分析来解释这些原理。然而，传统的组织管理理论并没有成功地解释这一现象。这一理论应该从形式上解释日本企业中典型活动的机制，如改善（Kaizen）、根和（Nemawashi）等。这些活动的重要而困难的特点是，它们严重依赖于成员之间的非正式信息处理，而且无法定量地测量。人工智能的最新进展使得用物理符号系统假说重新审视Simon的方法成为可能。因此，为了建立一个严格的组织学习理论，人工智能符号方法是很有前途的，因为它们的描述能力和计算机模拟的能力。在这个项目中，我们首先讨论了适用于组织理论的人工智能模型的要求。其次，为了便于分析，我们提出了一个计算模型：该模型由一组代理组成，具有（a）用于学习概念的知识库，（B）用于解决问题的知识库，（c）基于序言的推理机制，以及（d）关于其他代理的可靠性的一组信念。每个Agent通过对给定问题的归纳和/或演绎学习以及对其他Agent之间通信可靠性的强化学习来提高自身的问题解决能力，该模型的几个实验系统已经用CESP和Prolog语言实现。通过实验验证了智能体机器学习机制在解决问题和沟通能力方面的可行性。实验结果表明，所提出的模型是可执行的分析适用于分布式知识系统的学习机制。