摘 要
专家系统是人工智能的一个分支,其发展速度非常快目前专家系统应用领域已经渗透到数学,物理,化学,医学等众多领域,几乎所有的专家系统都至少能将人的效率提高到10倍左右。
本系统是根据人工智能和专家系统的理论知识，利用PowerBuilder9.0开发研制而成的，主要用于实现糖尿病诊断和治疗功能，从而达到利用计算机模拟人类专家诊疗糖尿病的目的。
本文以糖尿病诊断和治疗的知识为基础，论述了专家系统开发的基本原理，结构，方法和过程，详细阐述了知识库、推理机、数据库、人机对话界面、解释器的设计和实现方法，力求将所学到的知识在专家系统开发中得以全面的应用，并使系统在实际的操作中能按照设计的要求安全有效正确的运行。
关键字：专家系统，数据库，PowerBuilder,人工智能
Abstract
Expert system is a branch of artificial intelligence, its development speed is very fast, the application of expert system already permeate ,such numerous fields as physics , chemistry , medical science ,etc.. Almost all the expert systems can bring people''s efficiency up to about 10 times at least.
This system utilizes PowerBuilder 9.0 according to the theory knowledge of artificial intelligence and expert system. The purpose of the system is to realize the diagnosis and treatment of diabetes in order to utilize computer simulation human expert.
The paper introduce the principle and structure of expert system , It explained the design of knowledge base, reasoning machine , database , interactive interface, explaining device and implementation method in detail. During the development the expert system I use all I have learnt to make the system run correctly and eccectively.
Key word: Expert system, the database , PowerBuilder, artificial intelligence
目录
前言 1
第一章人工智能概述 2
1.1 人工智能定义 2
1.2 人工智能应用领域 2
1.3人工智能的研究方法 2
第二章 专家系统理论 4
2.1 专家系统的定义 4
2.2 专家系统的产生和发展 4
2.3 专家系统特点 5
2.4 专家系统的结构和工作原理 6
2.4.1专家系统的结构 6
2.4.2专家系统的工作原理 7
2.5 专家系统的建立 9
2.5.1 专家系统求解的问题 9
2.5.2 专家系统开发步骤 9
第三章 糖尿病知识库设计 10
3.1糖尿病基础知识 10
3.2 糖尿病专家诊断标准（知识库的设计） 12
3.2.1 糖尿病确诊树 12
3.2.2 糖尿病分型树 12
3.2.2 糖尿病分型树 13
3.2.3 糖尿病急性诊断 14
3.2.4 糖尿病慢性诊断 15
3．3二型糖尿病治疗方案 16
第四章 系统分析 17
4.1系统目标 17
4.2 系统需求分析 17
4.3 系统功能结构设计 17
4.4设计思想 18
4.5 开发环境介绍 19
4.5.1总体开发环境 19
4.5.2 开发工具简介 19
4.6 数据库设计 20
4.7数据库连接 20
4.7.1程序设计时与数据库的连接 20
4.7.2程序运行时与数据库的连接 21
第五章 关键模块代码设计 22
5.1 应用程序对象设计 22
5.2 主界面设计 23
5.3 糖尿病初步诊断窗口设计 25
5.4 糖尿病慢性诊断窗口 31
5.5 糖尿病预防模块 31
5.5 糖尿病预防模块 32
5.6 糖尿病治疗模块 32
5.7 血糖检测模块 33
第六章 系统的编译和发布 34
6.1 创建工程对象 34
6.2 编译生成和测试 34
6.3 发布 35
第七章 结 论 36
致 谢 37
参考文献 38
附 录 39
前言
一般认为，临床诊断是运用已有的医学知识对疾病的表现进行辩证分析，得出符合逻辑的结论的过程。就是把问诊、体格检查、试验室及其他各项辅助检查所得到的资料，运用医学知识进行辩证分析、综合、推理与判断，从而得出符合患者病情的诊断结论。其基本环节如下：收集资料以综合分析、推理，作出诊断。
在传统的疾病诊断过程中，其综合分析、推理除了从过去的历史资料中得来以外，还主要依据医生的个人经验。因此，其诊断存在一定的片面性和主观性。计量诊断与传统的疾病诊断基本环节一致，但其分析、推理不是凭经验，而是用一种定量的推理模式代替，再根据—定的法则作出合理的临床判断。因此，计量诊断也称定量诊断或数值诊断。随着计算机应用的发展，计量诊断过程可由计算机部分或完全实现，所以也称之为计算机辅助诊断。
从计算机科学角度讲，计算机辅助诊断就是利用机器模仿医生的智能。利用机器模仿人类的智能即人工智能。目前，基于神经网络模型人工智能的研究是世界智能技术研究的热点。人工智能技术在医学上的典型应用是专家系统。专家系统的实质就是让计算机系统代替专家为患者诊断，换句话说就是利用机器模仿人类专家的智能。
本文所介绍的糖尿病专家系统正是基于人工智能和专家系统的理论， 利用POWERBUILDER9.0开发与实现的，系统包括知识库、推理机、数据库、人机对话界面、解释器等。系统可以根据用户输入的症状，推理出他患糖尿病的概率，对糖尿病进行分型，以及糖尿病急性、慢性并发症诊断，并且给出糖尿病及并发症治疗方案，从而实现计算机模拟专家医生诊断、治疗糖尿病的功能。
第一章 人工智能概述
1.1 人工智能定义
智能可以理解为：知识如何获取、表达和存储；智能行为如何产生和学习；动机、情感和优先权如何发展和运用；传感器信号如何转换成各种符号；怎样利用各种符号执行逻辑运算、对过去进行推理及对未来进行规划；智能机制如何产生幻觉、信念、希望、畏惧、梦幻甚至善良和爱情等现象。广义地讲，人工智能是关于人造物的智能行为，而智能行为包括知觉、推理、学习、交流和在复杂环境中的行为。人工智能的一个长期目标是发明出可以像人类一样或能更好地完成以上行为的机器；另一个目标是理解这种智能行为是否存在于机器、人类或其他动物中。因此，人工智能包含了科学和工程的双重目标。
1.2 人工智能应用领域
人工智能的研究领域非常广泛,而涉及的学科也很多。目前,人工智能研究领域为以下几个方面
l 专家系统 （Expert System）
l 自然语言处理 (Natural Language Processing)
l 机器学习 （Machine Learning）
l 自动定理证明 （Automatic Theorem Proving）
l 分布式人工智能 (Distributed Artificial Intelligence)
l 机器人 (Robots)
l 模式识别 (Pattern Recognition)
l 博弈 (Game Plying)
l 计算机视觉 (Computer Vision)
l 人工神经网络 (Artificial Neural Network)
1.3人工智能的研究方法
尽管人工智能已经创造了一些实用系统，但人们不得不承认这些远未达到人类的智能水平。正因为如此，就选择人工智能研究的最佳方法—既为人工智能的最终研究目标打好基础，又能创造出短期效益—存在大量的讨论和争辩。这样，在过去的四十年里涌现出大量方法，每一种方法均有其拥护者，有些甚至有趣得令人爱不释手。也许所有这些方法应该综合起来运用。总之，所有这些拥护者都认为自己的研究方法具有突破性进展，值得特别关注。其中的一些方法可分为两大类。第一类包括符号处理的方法。它们基于N e w e l l和S i m o n的物理符号系统的假说。尽管不是所有人都赞同这一假说，但几乎大多数被称为“经典的人工智能”（即哲学家John Haugeland所谓的“出色的老式人工智能”或G O FA I）均在其指导之下。
第二类包括所谓的“子符号”方法。它们通常采用自下而上的方式，从最低阶段向上进行。在最低层阶段，符号的概念就不如信号这一概念确切了。 由子符号派制造的著名样品机器包括所谓的“神经网络（ Neural network）”。受到生物学方法的启发，这些系统主要因其学习的能力而十分有趣。根据模拟生物进化方面的进程，一些有趣的机器应运而生。
介于自上而下和自下而上之间的方法是一种动机“环境自动机（ situated automata ）”[Kaelbling & Rosenschein 1990 ，Rosenschein & Kaelbling 1995]的方法。Kaelbling 和Rosenschein 建议编写一种程序设计语言来说明a g e n t在高水平上所要求的行为，并编写一编译程序，以从这种语言编写的程序中产生引发行为的线路。
第七章 结 论
l 总结
专家系统是人工智能的一个分支,具有极强的应用性。由于对人工只能这门较为复杂学科的挑战，和对医学领域的兴趣，这次毕业设计课题，我选择的是糖尿病专家系统。通过一个学期的学习和实践,我阅读了大量有关人工只能和专家系统的书籍，基本掌握了专家系统的基本理论、基本设计方法，通过担任知识专家的指导老师的合作, 从大量的糖尿病诊断知识,预防知识,中提取精华,建立知识库,采取软件工程的方法和专家系统的设计原则进行系统分析和设计,实现推理机，完成人机交互界面。
在设计过程中，遇到许多的难题，比如：如何进行正确糖尿病诊断，如何进行BAYES概率推理，在各种医学数值下，如何取的让普通用户都能看懂的表达形式。还有如何使专家系统功能更加完善，让患者使用起来更为方便上都下了很大的工夫去改进，去克服。
l 不足之处
由于时间的限制，本系统仍有许多地方值得改进。
1. 知识库嵌套在程序中
专家系统的设计原理是程序与知识库分开，通过程序调用知识库，并能修改知识库的内容，由于本人第一次尝试专家系统设计，对如何分开知识库和程序并不十分了解，而且知识库的更新也仅限于一小部分程序中，由于我以后的研究方向可能是人工智能方向，所以一定能在未来内学到更深入的方法。
2. 实现过程原始化
在编写代码的过程中，发现大部分代码都是IFELSE语句不免让人觉得过于简单，现代专家系统绝大部分采用了神经网络和BAYES模糊推理的方法，在这个方面可以加以改进。
3. 知识提取不全面
在提取糖尿病知识中，对医学知识的缺乏，无法完整的提取出推理策略，比如慢性病诊断中，只是简单的推理有某些症状就能得出某个结论。
致 谢
即将走完大学的历程，在此，我首先要向我的指导老师表示最真挚的谢意。无论在毕业设计阶段还是在参加实习阶段，我都得到了何老师学习上的指点和生活上的关心。在我论文的前期准备、实验和写作过程中，何老师坚持每个星期抽出固定时间为我们解决阶段性的问题，并且花费了很多的精力来指导和点拨，提出了许多极有价值的意见和建议，使我在毕业设计的完成过程中少走了很多弯路，并如期的完成了我的论文。他严谨的治学态度，真诚的待人处事方式，我始终印象深刻，也是我今后治学、为人的榜样。对于他的鼓励和教诲，我将终生铭记。
衷心感谢我的同学，在这期间能及时地交换各种意见，提出了不少有益的见解。
参考文献
郭宝利编著.PowerBuilder 8.0 完全解析 .北京：电子工业出版社 .2002
樊金生等编著.PowerBuilder 8.0 实用教程 .北京：科学出版社 .2003
何旭洪，余建英编著.PowerBuilder数据库系统开发实例导航 .北京：人民邮电出版社 2003.6
黄厚宽, 田胜风著 .人工智能. 电子工业出版社. 2004
武波,马玉祥 .专家系统. 北京理工大学出版社. 2000
超楚生. 内分泌学 .北京:人民卫生出版社 .2001
王士同.人工只能教程.北京: 电子工业出版社. 2002
林尧瑞，张钹，石纯一. 专家系统原理与实践 .清华大学出版社.2001
Ｐ.Ｈ温斯顿蓍 傀光南,周少柏译. 人工智能 .科学出版社.2003
Anderson, P. F. Expert Systems, Expertise, and the Library and Information Professions. Library and Information Science Research 10, no. 4 (Oct-Dec 1988)
Abstract
An expert system is a branch of artificial intelligence, the development of very fast current expert system applications have penetrated into the mathematics, physics, chemistry, medicine and other fields, almost all of the expert system are at least able to raise the efficiency of people about 10 times .
This system is based on the theory of artificial intelligence and expert systems of knowledge, the use made of PowerBuilder9.0 developed, mainly for diagnosis and treatment of diabetes to achieve functionality, so as to achieve the use of computer simulation of human experts, the purpose of treatment of diabetes.
In this paper, diagnosis and treatment of diabetes, knowledge-based expert system development are discussed the basic principles, structure, methods and processes, detailed knowledge base, inference engine, database, man-machine conversation interface, interpreter design and implementation methods, sought to have learned knowledge in the expert system development to be a comprehensive application, and to the system in actual operation, the design requirements in accordance with the proper operation of safe and effective.
Keywords: expert systems, databases, PowerBuilder, artificial intelligence
Abstract
Expert system is a branch of artificial intelligence, its development speed is very fast, the application of expert system already permeate, such numerous fields as physics, chemistry, medical science, etc .. Almost all the expert systems can bring people''s efficiency up to about 10 times at least.
This system utilizes PowerBuilder 9.0 according to the theory knowledge of artificial intelligence and expert system. The purpose of the system is to realize the diagnosis and treatment of diabetes in order to utilize computer simulation human expert.
The paper introduce the principle and structure of expert system, It explained the design of knowledge base, reasoning machine, database, interactive interface, explaining device and implementation method in detail. During the development the expert system I use all I have learnt to make the system run correctly and eccectively.
Key word: Expert system, the database, PowerBuilder, artificial intelligence
Directory
Preface 1
Chapter 2 Overview of Artificial Intelligence
1.1 Definition of Artificial Intelligence 2
1.2 Artificial Intelligence Applications 2
1.3 Artificial Intelligence Research Methods 2
Chapter II Expert System Theory 4
2.1 Expert System Definition 4
2.2 The emergence and development of expert systems 4
2.3 Expert System Features 5
2.4 The expert system structure and working principle of 6
2.4.1 Structure of Expert System 6
2.4.2 Expert System Works 7
2.5 Expert System for the establishment of nine
2.5.1 Expert System to solve the problem 9
2.5.2 Expert System Development Step 9
Chapter 10 Diabetes Knowledge Base Design
3.1 Diabetes Basics 10
3.2 Diabetes expert diagnosis criteria (knowledge base design) 12
3.2.1 Diabetes diagnosed Tree 12
3.2.2 Diabetes classification tree 12
3.2.2 Diabetes classification tree 13
3.2.3 Acute diagnosis of diabetes mellitus 14
3.2.4 Diagnosis of diabetes, chronic 15
3.3 Type II diabetes treatment plan 16
Chapter IV Analysis 17
4.1 system target 17
4.2 System Requirements Analysis 17
4.3 The structural design of system functions 17
4.4 The design idea 18
4.5 Development Environment Introduction 19
4.5.1 Overall Development Environment 19
Development Tools 4.5.2 Introduction 19
4.6 Database Design 20
4.7 Database Connectivity 20
4.7.1 Programming with a database connection 20
4.7.2 program is running with the database connection 21
Chapter 22 code design key modules
5.1 Application Object Design 22
5.2 Main Interface Design 23
5.3 The initial diagnosis of diabetes, the design of window 25
5.4 Diagnosis of the window 31 with chronic diabetes
5.5 Diabetes Prevention Module 31
5.5 Diabetes Prevention Module 32
5.6 diabetes treatment module 32
5.7 Glucose Detection Module 33
Chapter VI system to compile and publish 34
6.1 Create Project Object 34
6.2 Compile generate and test 34
6.3 Release 35
Chapter VII Conclusion 36
Thanks 37
References 38
Appendix 39
Preface
Is generally believed that clinical diagnosis is to use the medical knowledge of disease has been the performance of dialectical analysis, the logical conclusion drawn process. Is to Wen Zhen, physical examination, laboratory and other auxiliary examination of the data obtained from the use of medical knowledge in a dialectical analysis, synthesis, reasoning and judgments to arrive at conclusions consistent with the diagnosis of patient''s condition. The basic aspects are as follows: to collect data for comprehensive analysis, reasoning, make a diagnosis.
In the traditional process of disease diagnosis, its comprehensive analysis, reasoning, apart from the historical data from the past come in, is also largely based on doctor''s personal experience. Therefore, there is a certain one-sidedness of their diagnosis and subjectivity. Measurement of disease diagnosis and conventional diagnosis of the same basic links, but its analysis, reasoning is not a rule of thumb, but use a quantitative reasoning mode instead, and then according to - the rule set to make a reasonable clinical judgments. Therefore, the measure also known as diagnostic value of quantitative diagnosis or diagnosis. With the development of computer applications, measurement diagnostic process can be partially or fully implemented a computer, it is also known as computer-aided diagnosis.
From the perspective of computer science, computer-aided diagnosis is to use the machine to imitate intelligent doctor. The use of machines to imitate human intelligence or artificial intelligence. At present, the neural network model based on artificial intelligence research is the world''s smart technology research hot spots. Artificial Intelligence in Medicine typical application is the expert system. The essence of expert systems enable computer systems to replace expert diagnosis for patients, in other words is to use the machine to imitate intelligent human experts.
Described in this article diabetes expert system is based on the theory of artificial intelligence and expert systems, using POWERBUILDER9.0 development and implementation, the system includes the knowledge base, inference engine, database, man-machine conversation interface, interpreter and so on. System can be based on user input symptoms, reasoning out the probability that he is suffering from diabetes, for diabetes type, as well as diabetes, acute and chronic complications, diagnosis, and complications of diabetes and treatment programs are given in order to achieve a computer simulation of expert medical diagnosis, treatment, Diabetes functions.
Chapter I Overview of Artificial Intelligence
1.1 Definition of Artificial Intelligence
Intelligence can be understood as: knowledge on how to obtain, expression, and storage; intelligent behavior how to generate and learning; motivation, emotion and priority to the development and use; sensor signal how to convert into a variety of symbols; how to use a variety of symbolic implementation of the logical operation of reasoning in the past and the future planning; intelligent mechanism for how the illusion, beliefs, hopes, fears, dreams and even goodness and love and so forth. Broadly speaking, artificial intelligence is about man-intelligent behavior and intelligent behavior, including perception, reasoning, learning, communication and behavior in complex environments. AI a long-term goal is the invention that can, like humans, or be able to complete the above acts of a better machine; Another objective is to understand the existence of such intelligent behavior in machines, humans or other animals. Therefore, the artificial intelligence includes the dual goals of science and engineering.
1.2 Artificial Intelligence Applications
Artificial intelligence field of study is very broad, and there are many disciplines involved. At present, the artificial intelligence research for the following aspects
l expert system (Expert System)
l Natural Language Processing (Natural Language Processing)
l Machine Learning (Machine Learning)
l Automatic Theorem Proving (Automatic Theorem Proving)
l Distributed Artificial Intelligence (Distributed Artificial Intelligence)
l Robots (Robots)
l Pattern Recognition (Pattern Recognition)
l game (Game Plying)
l Computer Vision (Computer Vision)
l Artificial Neural Networks (Artificial Neural Network)
1.3 Artificial Intelligence Research Methods
Although artificial intelligence has created a number of practical systems, but people have to admit that they fall far short of the level of human intelligence. Because of this, on the choice of the best way to artificial intelligence research - both for the study of artificial intelligence, the ultimate goal of laying a solid foundation, but also to create a short-term benefits - there is a lot of discussion and debate. Thus, in the past four decades emerged a large number of methods, each method has its supporters, and some interesting and attractive too. Perhaps all of these methods should be used together. In short, all of these advocates believe that their research method is a breakthrough worthy of special attention. Some of these methods can be divided into two broad categories. The first category includes symbol-processing methods. They are based on N ewell, and S imon the physical symbol system hypothesis. Although not everyone agreed with this hypothesis, but almost most of them were referred to as "the classic artificial intelligence" (ie, the philosopher John Haugeland so-called "good old-fashioned artificial intelligence" or GO FA I) are under their guidance.
The second category includes the so-called "sub-symbolic" approach. They are usually used bottom-up approach, from the lowest stage up to. In the lowest level stage, the symbol of the concept of the concept of precise than the signal. Signs were made by the sub-samples of the machine, including the famous so-called "neural network (Neural network)". Inspired by biological methods, these systems mainly because of their ability to learn and very interesting. According to the simulation aspects of the process of biological evolution, a number of interesting machines came into being.
Between between top-down and bottom-up approach is a motive "environment automata (situated automata)" [Kaelbling & Rosenschein 1990, Rosenschein & Kaelbling 1995] method. Kaelbling and Rosenschein recommended the preparation of a programming language to illustrate the agent at a high level of conduct required and the preparation of a compiler to a program written in that language from the generated trigger behavior of the line.
Chapter VII Conclusion
l Summary
An expert system is a branch of artificial intelligence, with strong applicability. This door can only be due to the more complex artificial discipline challenges, and interest in the medical field, this graduation design topics, I chose the diabetes expert system. Through a semester of study and practice, I read a lot about artificial and expert systems can only books, the basic grasp of the expert system''s basic theory, basic design methodology, as knowledge, expert guidance through the cooperation of teachers, from a large number of diabetes diagnosis knowledge, prevention knowledge, to extract the essence of building a knowledge base approach to software engineering and expert system design principles of systems analysis and design, implement inference engine to complete the man-machine interface.
In the design process, encountered many problems, such as: how to correctly diagnosis of diabetes, how to BAYES Probabilistic Reasoning in a variety of medical value, how do we get to let ordinary users can understand the expression. Expert system functions and how to make more perfect, so that it more convenient for patients to use both made a great effort to improve, to overcome.
l shortcomings
Due to time constraints, this system there are still many areas that warrant improvement.
1. Knowledge nested in the program
Expert system design principle is to separate the procedures and knowledge base, through the process called knowledge base, and can modify the contents of the knowledge base, because my first attempt to expert system design, and procedures on how to separate the knowledge base is not very understanding, and knowledge base The update also limited to a small part of the program, as I am the future direction of research could be artificial intelligence, so it must be able to learn more in-depth within the next approach.
2. The original of the implementation process
In the process of writing code, found that most of the code is IF ELSE statements can not help but think it is too simple, most of the modern expert system using neural networks and fuzzy inference BAYES method, in this regard can be be improved.
3. Knowledge Extraction incomplete
In the extraction of diabetes knowledge, the lack of medical knowledge can not extract the full reasoning strategies, such as the diagnosis of chronic diseases, but the simple reasoning has some of the symptoms will be able to come to a conclusion.
Thanks
About to finish the university course, in this, I would first like to express my most sincere guidance of teacher appreciation. In terms of graduation to participate in the design stage or in the internship stage, I have been learning what the teacher''s pointing and life concerns. In my paper preparation, experimentation and the writing process, where teachers insist on a fixed time every week taking the stage for us to solve the problem, and spent a lot of energy to guide and Enlightenment from and offered many valuable comments and proposal, so I graduated from the completion of the design process a lot less to go roundabout, and on schedule and completed my thesis. His rigorous scholarship, attitude, honest dealing with others ways of doing things, I have always been impressed, but also my scholarly pursuits in the future, man''s example. For his encouragement and teachings, I will bear in mind for life.
Heartfelt thanks to my classmates, during which the timely exchange of views and ideas, put forward some useful insights.
References
Guo Baoli compiled. PowerBuilder 8.0 completely resolved. Beijing: Electronic Industry Publishing House .2002
Fan Jinsheng eds. PowerBuilder 8.0 and practical tutorial. Beijing: Science Press, .2003
Xu-Hong He, Yu Jian-Ying eds. PowerBuilder database instance of navigation system development. Beijing: People''s Posts & Telecom Press 2003.6
HUANG Hou-kuan, TIAN Sheng wind with. AI. Electronic Industry Press. 2004
Bo Wu, Yu-Xiang Ma. Expert systems. Beijing University of Technology Press. 2000
super-Chusheng. Endocrinology. Beijing: People''s Health Publishing House .2001
WANG Shi-tong. Man can only tutorial. Beijing: Electronic Industry Press. 2002
Lin Yao-Rui, Zhang Bo, SHI Chun-yi. Expert System Principles and Practice. Tsinghua University Press .2001
P.H Winston Achillea puppet light south, Zhou Shao-bai translation. AI. Science Press .2003
Anderson, PF Expert Systems, Expertise, and the Library and Information Professions. Library and Information Science Research 10, no. 4 (Oct-Dec 1988)
