Artificial Intelligence and Expert Systems | Types of expert systems | Expert systems components | Expert systems technologies | Development tools for building the expert system | Web-enabled expert systems
Artificial Intelligence and Expert Systems - Artificial Intelligence (AI) is a branch of computer science, which focuses on the development of computer systems to simulate the processes of problem solving and duplicate human brain functions. Expert systems (ES) are a category of programs based on the theory and methods of artificial intelligence.
The first expert systems appeared in the late sixties. Today, they exist in many forms, from medical diagnosis to investment analysis and from counseling to production control. Due to the advances of the last decade, today's expert systems users can choose from dozens of commercial software packages. At present, we accept as routine such expert systems as weather forecasting, online mapping and driving directions, diagnostic systems for automotive repair shops, and so on.
Many large corporations use expert systems in their business. The list of the companies using expert systems technology is long and varied: NASA, HP, Lockheed, Boing, DaimlerChrysler AG, various power, gas and oil stations, etc.
Expert systems tend to be more effective than other computer based applications, because they:
may combine the knowledge of many experts in a specific field,
can store an unlimited amount of information, and works much faster, than a human,
are available 24 hours a day, and can be used at a distance over a network,
are able to explain their information requests and suggestions,
can process client's uncertain responses and, by combining several pieces of uncertain information, may still be able to make strong recommendations,
can accumulate the knowledge of high level employees for any company, which is especially useful when the company needs to fire them due to worsened market conditions.
Types of expert systems: Expert systems fall into a number of major categories (not exclusive) described below. There are many applications not covered , and some manufacturers and vendors could fit in multiple categories, but due to space limitations, they are only included in one.
Acquisition and interpretation of data. These categories retrieve, analyze, filter and restore missed data, as well as mine data, transform data in another format, and so on.. Examples of this category include:
Netrics Data Quality Software by Netrics Inc. enables cleansing of mission-critical data. Netrics products deliver near-perfect accuracy no matter how "dirty" and error-prone the data is.
Nuggets by Data Mining Technologies, Inc., is an application which builds models to uncover hidden facts and relationships. These models can predict new data, allow generalizing and revealing which indicators (i.e. variables) most impact decisions.
WizWhy by Wizsoft, is a data-mining tool that analyzes the data and issues predictions. By automatically determining the if-then rules in the data it is able to summarize the data, highlight interesting phenomena in the data as well as cases deviating from the rules, reveal all necessary and sufficient conditions and generate predictions for new cases.
Diagnosis and Abnormal Condition Management - These categories recognize the condition of human health or the technical condition of complex industrial systems. Among them:
EasyDiagnosis by Easydiagnosis (a division of MatheMedics) offers automatic online medical diagnosis for consumers.
Intelligent Objects by Gensym Inc. is ES for Abnormal Condition Management (ACM) applications that proactively detect, diagnose, and recover from process problems in real time.
Transformer Oil Analyst (TOA), developed by Delta-X Research. TOA warns the staff of any suspicious symptoms or trends. When there is a suspected equipment or insulating fluid problem , the TOA enables the user to review both historical and current status quickly.
ASTD by Intellectual Systems, Inc.(Russia) is real-time hybrid ES for the pipe line's gas compressor station.
Decision-making or decision support - These categories include the following subcategories: Decision management, Decision optimization, Customer and Account management, Fraud management, Scoring and predictive modeling, Risk management, etc. Examples include:
A variety of expert systems developed by Fair Isaak, Inc, such as: BlazeAdviser, Blaze Decision System, Decision Optimizer, Payment Optimizer, Strategy science, Model Builder for decision Trees, StrategyWare, Capstone Decision Manager, etc;
ReThink by Gensym Corporation, This software helps manage the increasing speed and complexity of business data through simulation, analysis, and automation. Unlike traditional simulation tools, ReThink supports the entire business process lifecycle from as-is analysis, to online automation.
Design and management solutions - After consideration of a large number of variable parameters for solving innovation tasks, it enables optimization of design, and find the best management decisions. Here are Goldfire Innovation and Goldfire Intelligence developed by Invention machine. Some companies use for design goals the TRIZ (see below).
Control and Monitoring - These categories help monitor operation and control certain functions of machinery for heavy industry. Many corporations that developed Supervisory Control and Data Acquisition Systems (SCADA), used the expert systems for different control levels of industrial plants. Here there are companies such as: Emerson Process Management, Matricon, REDUCT & Lobbe Technologies, etc.
Prediction - This category allows to forecast possible outcomes of observable situations. Here are weather forecasting, scoring and predictive modeling, risk management and other financial forecasting:
NeuralWorks Predict, developed by Neuralware Inc, is an integrated tool for rapidly creating and deploying prediction and classification applications. Predict combines neural network technology with genetic algorithms, statistics, and fuzzy logic to automatically find optimal or near-optimal solutions for a wide range of problems.
Developed by NeuroDimention Inc. the following products can solve similar tasks: The neural network modeling software NeuroSolutions 4.2 is a tool for solving data modeling problems. The technical analysis software TradingSolutions 2.1 combines neural network and genetic algorithm technologies with traditional technical analysis to create a highly effective tool for financial modeling.
Predict , developed by Partek, is a toolkit for predictive modeling to deploy models for forecasting unknown cases.
Analysis and Reporting tools: There are a number of applications including the following:
The KBX Data Management Suite by KBACE Technologies, provides an innovative solution to transferring data from an ERP (Enterprise Resource Planning) System to end-user desktops in both a memory and time efficient manner. All reports are designed for use with Microsoft Office.
MicroStrategy 7i developed by MicroStrategy Inc. is the Industrial-Strength Business Intelligence application for integrating, reporting, analysis, and information delivery capabilities into one platform.
PowerAnalyzer by Informatica Corporation is designed for the mass deployment and adoption of business analytics.
Manufacturing- One of most active area of expert systems and AI related activities:
e-SCOR by Gensym Corporation. allows to simulate various configurations, test the robustness of supply chain and identify the required service levels. It can help identify the weak links and areas for improvement in supply chain, saving time and money. e-SCOR is highly flexible and is ideal for performing "what-if" analyses.
Marketing and trade management - A key use of artificial intelligence to determine marketing potential and business opportunities.
Campaign Optimizer, Preview Reporting, Opportunity Mapper, etc. by PreVision Marketing LLC provide full customer marketing database design, development, management and hosting services for clients looking to outsource their multi-channel customer marketing data management.
Ward Systems Group issue the AI Trilogy, a new package of business and scientific software. The AI Trilogy contains the NeuroShell Predictor, NeuroShell Classifier neural net software, and GeneHunter genetic algorithm (GA) software together in one package.
CycAnswers (Cycorp, Inc) is the knowledge-based question-answering product, based on years of common sense reasoning research that supplies answers to client's questions.Human Resource Management Systems (HRMS)- The use of AI technology to find human technical talent.
CCHKnowledgepoint a developer of human resource solutions since 1987, designed a new Performance Impact Workplace for small-to medium businesses (SMB).
ESI (Expert Solutions International) provides Knowledge - based solutions and Workforce Management optimization products.
e-Learning Systems (ELS)- supply individual or group education and training for different disciplines on the basis of special tools. These learning tools and services allow to learn faster and adapt quicker to the challenges of today's business world. ELS can include Learning Management Systems and Course Management systems. Here are some developers and providers of e-Learning systems:
GeoLearning, Inc. is a leading developer of e-learning delivery platforms and Web-based training solutions for organizations around the world, including Fortune 1000 companies, government agencies, etc.
CompanyCollege.com is the e-learning solution from Business Training Library, the provider of affordable training solutions for small companies. CompanyCollege offers high-quality online learning that is easy to use, completely customizable, and inexpensively priced.
Generally, an expert system includes several key components:
Data input/output
Knowledge base or mathematical ( logical models)
Inference engine
Explanation system
Intellectual editor for the knowledge base,
User interface.
The main component of the expert system is knowledge base - an organized collection of facts, heuristics and other information on the subject of system's domain. An ES is built in a process known as knowledge engineering, during which knowledge about the domain is acquired from human experts and other sources. The part of the expert system that applies the knowledge to the problem's solving, is called the inference engine. The explanation system explains the basis for the conclusion, reached by the expert system. The intellectual editor is a tool for correction, learning or self learning of the knowledge base. A friendly user interface enables inexperienced users to specify problems for the system to solve and to understand the system's conclusions.
Expert systems technologies are typically based on the types of knowledge bases or mathematical (logical) models used for problem solving with the help of inference engines. Here are some examples of different types of technologies:
Model Based Reasoning - Using a modeling environment to explore data , create algorithms, and create custom rules that offer early insights and competitive advantages
In this field MATLAB®, Simulink®, Stateflow®, etc. (The Mathworks, Inc.) have become the fundamental tools for engineering and scientific work MATLAB integrates mathematical computing, visualization, and a powerful language to provide a flexible environment to explore data, create algorithms, and create custom tools that provide early insights and competitive advantages. Simulink is an interactive tool for modeling, simulating, and analyzing dynamic, multi-domain systems. Stateflow is an interactive design tool for modeling and simulating event-driven systems.
Opal-RT Technologies, Inc. specializes in new low-cost PC based tools for large-scale simulation, rapid control prototyping, and software or systems testing with hardware-in-the-loop.
Case-Based Reasoning (CBR): A CBR system compares the problem with previous cases, and then suggests the most successful course of action. Typically, in addition to manually built CBR systems, data mining methods also used.. You can find the list of the CBR's software vendors on the AI-CBR site
Evolutionary model reasoning - An evolutionary algorithm is a class of algorithms that finds approximate solutions to difficult-to-solve problems. They are inspired by mechanisms from biological evolution such as natural selection, mutation and recombination to find the best configuration for a specific system within specific constraints.
These algorithms are developed by the company Frontline Systems, Inc. in the Solver's products
Stochastic model reasoning -Bayesian Belief Network (BBN) use a tree-based structure rather than a logical "if-then", Bayesian systems iteratively use probabilities for finding any solutions.
The Netica APIs (Application Program Interfaces) are a family of powerful Bayesian Network toolkits from the Norsys software corporation.. They enable the building of Bayesian belief networks and Influence Diagrams, do probabilistic inferencing, learning nets from data, modifying nets, and save and restore nets.
Fuzzy Logic reasoning - Fuzzy logic is the emulation of human reasoning on computers. The key concepts in fuzzy logic are the linguistic variable and membership function. A linguistic variable is a variable whose value is a word. (for example, the linguistic variable "temperature" might have the values: "normal", "high," "cold," "huge", "danger" ,"freezing," and so on). A membership function describes these linguistic values in terms of numerals. For example, to describe the word "cold", we can say, "If temperature is below 40 degrees, we can say it's definitely cold." The linguistic variables and their membership functions allow fuzzy logic to perform the imprecise, non-numerical reasoning performed in the same manner as humans.
Aptronix® is developing The Ultimate Fuzzy Design Tool - FIDE (Fuzzy Inference Development Environment).
Brain simulation reasoning -Artificial Neural Networks (ANN) are named after the cells in the human brain, that include billions of neuron cells. ANN are formed from hundreds or thousands of simulated neurons connected together in much the same way as the brain's neurons. Neural networks are fast, tolerant of imperfect data, and do not need formulas or rules. ANN are good at pattern recognition, generalization, and trend prediction. Neural networks learn from experience, because they are trained by repeatedly reviewing examples. Each example includes both inputs (information you would use to make a decision) and outputs (the resulting decision, prediction, or response).
NeuralWorks Professional II/PLUS by NeuralWare Inc. is a comprehensive neural network technology platform that implements over 28 standard paradigms in an advanced model development workbench that is both integrated and extensible.
TRIZ systems (TRIZ is a Russian acronym for Theory of Inventive Problem Solving), that provides systematic methods for solving design and technological problems. This practical problem-solving paradigm was invented and developed by Genrich Altshuller, who discovered that technological systems evolve according to specific laws. TRIZ is rapidly gaining popularity with worldwide innovation communities by virtue of its success in guiding creative thinking and innovative product development.
The ALTSCHULLER INSTITUTE for TRIZ studies.has been founded to complete more related research. Companies and groups appeared, such as Technical Innovation Center, Ideation International , Got Innovation etc., that are evolving ideas and principles, innovated by G. Altshuller. Their products include software, training, publications, e-learning, and consulting services.
Hybrid technologies - Many industrial expert systems combine different technologies, taking advantage of the strengths of each technology.
The expert system NuEngineer of the Numan Intelligence, Inc., generalizes many existing Intelligent Search Algorithms such as Heuristic Search, Genetic Algorithms, Optimization, Gradient, and Neural Networks;
The expert system FlexTool (CynapSys, Inc,) uses Genetic Algorithm, Neural Network, and Fuzzy Expert System Matlab Toolbox for scientific and business fields.
Tools for Expert System Development
There are a number of different programming approaches, programming languages and development environment. Here are a few:
Programming languages for expert systems - Almost every high level programming language ever created, and many lower languages, have been used to develop AI and ES applications. The most common seems to be: LISP, Prolog, C/C++ followed by Java. Here are some information on two programming languages Prolog and LISP, that were key to programming languages for Artificial Intelligence for the last two decades. Prolog is a full, industry-standard programming language, ideally suited to writing rules. Common Lisp is well suited to large programming projects, dynamic modifications and explorative programming.
In the 1980s several versions of Prolog came out, including Cogent Prolog, Turbo Prolog (Borland), Amzi! Prolog, and LPA (Logic Programming and Associates). AMZI and LPA are medium-priced Prolog implementations with excellent integration with Delphi, Java, and many other programming environments and platforms. Another successful and attractively priced implementation is also available from Visual Prolog - from the Prolog Development Center (PDC), the successor of Turbo Prolog .
The two major dialects of LISP in use today are Common Lisp and Scheme.
Franz Inc. is currently the primary supplier of LISP development environment , created Allegro CL, a dynamic object-oriented development environment for Common Lisp/CLOS.
Shells - A shell is an expert system without a knowledge base. A shell furnishes the ES developer with the inference engine, user interface, and the explanation and knowledge acquisition facilities. All of this makes it unnecessary to rebuild the components for each new expert system.
Acquire, product of Acquired Intelligence, Inc. is both a knowledge base authoring tool and expert system shell.
CLIPS is a productive development and delivery expert system tool which provides a complete environment for the construction of rule and/or object based expert systems.
Using Jess, it is possible to develop Java software that has the capacity to "reason" using knowledge you supply in the form of declarative rules.
SDK and environments - Using shells to write expert systems generally greatly reduces the cost and time of development (compared with writing the expert system from scratch). Software Development Kits (SDK) expand the capabilities of shells in various directions. Here are come examples of SDK and environments, all though most AI products provide an some form of environment.
G2 by Gensym Corporation is a graphical, object oriented software platform for rapidly building real-time expert system application.
GURU, product of Micro Data Base Systems, is a total system for developing applications that combine expert system and database technology.
The DX Solution Series product line, developed by Knowledge industries, consists of three components: DXpress, WIN-DX and API-DX.
Machine learning systems (MLS)- are becoming an important part of the expert system environment. In contrast to performance systems that acquire knowledge from human experts, machine learning systems acquire knowledge automatically from examples, i.e., from source data. MLS are frequently used for learning neural network, knowledge bases, using inductive rule based algorithms, and evolution-based genetic algorithms, etc. MLS usually use data-driven, hypothesis-driven, and multi-strategy constructive induction.
Here are some popular products and services:
EXSYS, Inc. is the first company providing the ability to run knowledge automation expert systems on the Web. Over the years, the EXSYS tools have been enhanced in many ways, but always with a focus on solving real-world problems through a practical easy-to-use development paradigm. Exsys CORVID is flagship "Knowledge Automation" paradigm resulting from a close examination of what is needed to build and field decision-making systems in today's Internet-oriented world. CORVID is designed to allow knowledgeable" experts" to develop and deploy powerful, interactive online advisory applications quickly and easily.
Amzi!Prolog has a server based Prolog
implementations ideal for integration with the web.
Webcom, Inc.
produces WebSource Configurator, an
application that
Vanguard Software Corporation is developing the following products: DecisionPro -Integrated software for management decision analysis, risk management, forecasting, optimization, and modeling. Vanguard Vista-Web-based survey tool that allows user to create, post, and track survey results online.
Conclusion: Although technical progress in the last decade has been satisfactory, as shown in examples above (far from complete), the number of widespread expert systems in use today is less than was expected. Now embedded expert systems tend to be more successful than advisory systems. In my opinion, the primary reason for the lower usage of expert system in advisory type roles is the very expensive procedures of retrieving knowledge in a short time. Developments in the machine learning systems could accelerate the number of new expert systems we see in the future.
Copyright © 2002-2007, Clever Ace. All Rights Reserved.