Preview (4 of 13 pages)

Preview Extract

Chapter 13 Intelligent Information Systems Learning Objectives Define artificial intelligence, and explain how AI technologies support decision making. Describe an expert system, its applications, and its components. Describe case-based reasoning. Summarize the types of intelligent agents and how they are used. Describe fuzzy logic and its uses. Explain artificial neural networks. Describe how genetic algorithms are used. Explain natural-language processing and its advantages and disadvantages. Summarize the advantages of integrating AI technologies into decision support systems. Explain contextual computing. Detailed Chapter Outline I. What is Artificial Intelligence? Artificial intelligence (AI) consists of related technologies that try to simulate and reproduce human thought behavior, including thinking, speaking, feeling, and reasoning. AI technologies apply computers to areas that require knowledge, perception, reasoning, understanding, and cognitive abilities. To achieve these capabilities, computers must be able to do the following: Understand common sense. Understand facts and manipulate qualitative data. Deal with exceptions and discontinuity. Understand relationships between facts. Interact with humans in their own language. Deal with new situations based on previous learning. Information systems are concerned with capturing, storing, retrieving, and working with data, but AI technologies are concerned with generating and displaying knowledge and facts. In the AI field, knowledge engineers try to discover “rules of thumb” that enable computers to perform tasks usually handled by humans. AI encompasses several related technologies, including robotics, expert systems, fuzzy logic systems, intelligent agents, artificial neural networks, genetic algorithms, and natural-language processing. Although these applications and technologies may not offer true human intelligence, they are certainly more intelligent than traditional information systems. A. AI Technologies Supporting Decision Making Decision makers use information technologies in the following types of decision-making analyses: What-is—this analysis is commonly used in transaction-processing systems and management information systems. What-if—this analysis is used in decision support systems. Decision makers use it to monitor the effect of a change in one or more variables. B. Robotics Robots are one of the most successful applications of AI. They perform well at simple, repetitive tasks and can be used to free workers from tedious or hazardous jobs. Personal robots have attracted a lot of attention recently. These robots have limited mobility, limited vision, and some speech capabilities. Currently, they are used mostly as prototypes to test such services as helping the elderly, bringing breakfast to the table, cooking, opening doors, and carrying trays and drinks. Robots offer the following advantages over humans in the workplace: They do not fall in love with coworkers, get insulted, or call in sick. They are consistent. They can be used in environments that are hazardous to humans, such as working with radioactive materials. They do not spy for competitors, ask for a raise, or lobby for longer breaks. II. Expert Systems Expert systems have been one of the most successful AI-related technologies and have been around since the 1960s. An expert system consists of programs that mimic human thought behavior in a specific area that human experts have solved successfully. For expert systems to be successful, they must be applied to an activity that human experts have already handled, such as tasks in medicine, geology, education, and oil exploration. Decision support systems generate information by using data, models, and well-defined algorithms, but expert systems work with heuristic data. Heuristics consist of common sense, rules of thumb, educated guesses, and instinctive judgments, and using heuristic data encourages applying knowledge based on experience to solve or describe a problem. In other words, heuristic data is not formal knowledge, but it helps in finding a solution to a problem without following a rigorous algorithm. A. Components of an Expert System A typical expert system includes the components described in the following list: Knowledge acquisition facility—a knowledge acquisition facility is a software package with manual or automated methods for acquiring and incorporating new rules and facts so the expert system is capable of growth. Knowledge base—a knowledge base is similar to a database, but in addition to storing facts and figures it keeps track of rules and explanations associated with facts. Factual knowledge—facts related to a specific discipline, subject, or problem. Heuristic knowledge—rules related to a problem or discipline. Meta-knowledge—meta-knowledge is knowledge about knowledge. It enables an expert system to learn from experience and examine and extract relevant facts to determine the path to a solution. It also guides future planning or execution phases of an expert system. Knowledge base management system—a knowledge base management system (KBMS), similar to a DBMS, is used to keep the knowledge base updated, with changes to facts, figures, and rules. User interface—this is the same as the user interface component of a decision support system. Although GUIs have improved this component, one goal of AI technology is to provide a natural language for the user interface. Explanation facility—an explanation facility performs tasks similar to what a human expert does by explaining to end users how recommendations are derived. Inference engine—an inference engine is similar to the model base component of a decision support system. By using different techniques, such as forward and backward chaining, an inference engine manipulates a series of rules. In forward chaining, a series of “if-then-else” condition pairs is performed. The “if” condition is evaluated first, then the corresponding “then-else” action is carried out. In backward chaining, the expert system starts with the goal—the “then” part—and backtracks to find the right solution. The backward chaining technique can be faster in some situations because it does not have to consider irrelevant rules, but the solution the system recommends might not be the best one. Other techniques are used for representing knowledge in the expert system’s knowledge base, such as semantic (associative) networks that represent information as links and nodes, frames that store conditions or objects in hierarchical order, and scripts that describe a sequence of events. B. Uses of Expert Systems Many companies are engaged in research and development of expert systems, and these systems are now used in areas such as the following: Airline industry Forensics lab work Banking and finance Education Food industry Personnel management Security U.S. government Agriculture C. Criteria for Using Expert Systems An expert system should be used if one or more of the following conditions exists: A lot of human expertise is needed but a single expert cannot tackle the problem on his or her own. The knowledge that is needed can be represented as rules or heuristics; a well-defined algorithm is not available. The decision or task has already been handled successfully by human experts, allowing the expert system to mimic human expertise. The decision or task requires consistency and standardization. The subject domain is limited. The decision or task involves many rules (typically between 100 and 10,000) and complex logic. There is a scarcity of experts in the organization, or key experts are retiring. D. Criteria for Not Using Expert Systems The following situations are unsuitable to expert systems: There are very few rules (less than 10) involved. Human experts are more effective at solving these problems. There are too many rules (usually more than 10,000) involved. Processing slows to unacceptable levels. There are well-structured numerical problems (such as payroll processing) involved, which means that standard transaction-processing methods can handle the situation more quickly and economically. A broad range of topics is involved, but there are not many rules. There is a lot of disagreement among experts. The problems require human experts. E. Advantages of Expert Systems An expert system can have the following advantages over humans: It never becomes distracted, forgetful, or tired. Therefore, it is particularly suitable for monotonous tasks that human workers might object to. It duplicates and preserves the expertise of scarce experts and can incorporate the expertise of many experts. It preserves the expertise of employees who are retiring or leaving an organization. It creates consistency in decision making and improves the decision-making skills of nonexperts. III. Case-Based Reasoning Expert systems solve a problem by going through a series of if-then-else rules, but case-based reasoning (CBR) is a problem-solving technique that matches a new case (problem) with a previously solved case and its solution, both stored in a database. In design and implementation of any case-based reasoning application there are four Rs involved: Retrieve: To solve the current case (problem) the system compares it with the cases stored in the database and retrieves the most similar case from the library of the past cases. Reuse: The retrieved case is reused to solve the current problem. Revise: The retrieved case is revised if necessary for further enhancement. Retain: The final solution is retained as a part of the library for future use. IV. Intelligent Agents Intelligent agents, also known as bots (short for robots), are software capable of reasoning and following rule-based processes; they are becoming more popular, especially in e-commerce. A sophisticated intelligent agent has the following characteristics: adaptability, autonomy, collaborative behavior, humanlike interface, mobility, and reactivity. One important application of intelligent agents that is already available is Web marketing. Intelligent agents can collect information about customers, such as items purchased, demographic information, and expressed and implied preferences. Intelligent agents are also used for smart or interactive catalogs, called “virtual catalogs.” A virtual catalog displays product descriptions based on customers’ previous experiences and preferences. Intelligent agents that are currently available fall into these categories: Shopping and information agents Personal agents Data-mining agents Monitoring and surveillance agents A. Shopping and Information Agents Shopping and information agents help users navigate through the vast resources available on the Web and provide better results in finding information. These agents can navigate the Web much faster than humans and gather more consistent, detailed information. They can serve as search engines, site reminders, or personal surfing assistants. Usenet and newsgroup agents have sorting and filtering features for finding information. B. Personal Agents Personal agents perform specific tasks for a user, such as remembering information for filling out Web forms or completing e-mail addresses after the first few characters are typed. C. Data-Mining Agents Data-mining agents work with a data warehouse, detecting trends and discovering new information and relationships among data items that were not readily apparent. D. Monitoring and Surveillance Agents Monitoring and surveillance agents usually track and report on computer equipment and network systems to predict when a system crash or failure might occur. V. Fuzzy Logic Fuzzy logic allows a smooth, gradual transition between human and computer vocabularies and deals with variations in linguistic terms by using a degree of membership. A degree of membership shows how relevant an item or object is to a set. A higher number indicates it is more relevant, and a lower number shows it is less relevant. Fuzzy logic is designed to help computers simulate vagueness and uncertainty in common situations. Fuzzy logic allows computers to reason in a fashion similar to humans and makes it possible to use approximations and vague data yet produce clear and definable answers. Fuzzy logic works based on the degree of membership in a set (a collection of objects). In a conventional set (sometimes called a “crisp” set), membership is defined in a black-or-white fashion; there’s no room for gray. A. Uses of Fuzzy Logic Fuzzy logic has been used in search engines, chip design, database management systems, software development, and other areas. VI. Artificial Neural Networks Artificial neural networks (ANNs) are networks that learn and are capable of performing tasks that are difficult with conventional computers, such as playing chess, recognizing patterns in faces and objects, and filtering spam e-mail. Like expert systems, ANNs are used for poorly structured problems—when data is fuzzy and uncertainty is involved. Unlike an expert system, an ANN cannot supply an explanation for the solution it finds because an ANN uses patterns instead of the if-then-else rules that expert systems use. An ANN creates a model based on input and output. An ANN has an output layer, an input layer, and a middle (hidden) layer where learning takes place. Every ANN has to be trained, and when organizational policies change, the network needs to be retrained so it can mimic the new policies. ANNs are used for many tasks, including the following: Bankruptcy prediction Credit rating Investment analysis Oil and gas exploration Target marketing VII. Genetic Algorithms Although they are not as widely accepted, genetic algorithms (GAs) have become more recognized as a form of artificial intelligence. They are used mostly to find solutions to optimization and search problems. Genetic algorithms are used for optimization problems that deal with many input variables, such as jet engine design, portfolio development, and network design. They find the combination of inputs that generates the most desirable outputs, such as the stock portfolio with the highest return or the network configuration with the lowest cost. Genetic algorithms can examine complex problems without any assumptions of what the correct solution should be. In a GA system, the following techniques are used: Selection or survival of the fittest—gives preference or a higher weight to better outcomes. Crossover—combines good portions of different outcomes to achieve a better outcome. Mutation—tries combinations of different inputs randomly and evaluates the results. Chromosome—a set of parameters that defines a proposed solution to the problem the GA is trying to solve. Genetic algorithms are already used with neural networks and fuzzy logic systems to solve scheduling, engineering design, and marketing problems, among others. Some hybrid products that combine AI technologies use GAs. VIII. Natural-Language Processing Natural-language processing (NLP) was developed so users could communicate with computers in human language. An NLP system provides a question-and-answer setting that is more natural and easier for people to use. It is particularly useful with databases. The size and complexity of the human language has made developing NLP systems difficult. NLP systems are generally divided into the following categories: Interface to databases Machine translation, such as translating from French to English Text scanning and intelligent indexing programs for summarizing large amounts of text Generating text for automated production of standard documents Speech systems for voice interaction with computers NLP systems usually perform two types of activities. The first is interfacing: accepting human language as input, carrying out the corresponding command, and generating the necessary output. The second is knowledge acquisition: using the computer to read large amounts of text and understand the information well enough to summarize important points and store information so the system can respond to inquiries about the content. IX. Integrating AI Technologies into Decision Support Systems AI-related technologies, such as expert systems, natural-language processing, and artificial neural networks, can improve the quality of decision support systems (DSSs). These systems are sometimes called integrated (or intelligent) DSSs (IDSSs), and the result is a more efficient, powerful DSS. AI technologies, particularly expert systems and natural-language processing, can be integrated into the database, model base, and user interface components of a DSS. The benefits of integrating an expert system into the database component of a DSS are: Adding deductive reasoning to traditional DBMS functions Improving access speed Improving the creation and maintenance of databases Adding the capability to handle uncertainty and fuzzy data Simplifying query operations with heuristic search algorithms Similarly, one can add AI technologies to a DSS’s model base component. In addition, integrating expert system capabilities into the user interface component can improve the quality and user friendliness of a DSS. Integrating NLP can improve the effectiveness of an interface, too, by making it easier to use, particularly for decision makers who are not computer savvy. X. Contextual Computing: Making Mobile Devices Smarter People have been using GPSs on their smartphones for years. This is a great service for somebody who is not familiar with a city. These kinds of applications could have significant commercial value. Contextual computing is expected to carry this idea much further still. Humans make decisions based on what they know and how they feel about something, drawing on experiences they have accumulated throughout their lives. Sometimes referred to as the sixth, seventh, and eighth senses, contextual computing refers to a computing environment that is always present, can feel people’s surroundings, and—based on who they are, where they are, and whom they are with—offer recommendations. The principle behind contextual computing is that computers can both sense and react to their environments similar to how human brains understand and interpret stimuli. In essence, contextual computing allows for tailoring a course of action to a user in a particular situation and environment based on what it knows about the user. Key Terms Artificial intelligence (AI) consists of related technologies that try to simulate and reproduce human thought behavior, including thinking, speaking, feeling, and reasoning. AI technologies apply computers to areas that require knowledge, perception, reasoning, understanding, and cognitive abilities. (P. 275) Robots are one of the most successful applications of AI. They perform well at simple, repetitive tasks and can be used to free workers from tedious or hazardous jobs. (P. 277) Expert systems mimic human expertise in a particular field to solve a problem in a well-defined area. (P. 279) A knowledge acquisition facility is a software package with manual or automated methods for acquiring and incorporating new rules and facts so the expert system is capable of growth. (P. 279) A knowledge base is similar to a database, but in addition to storing facts and figures it keeps track of rules and explanations associated with facts. (P. 279) A knowledge base management system (KBMS), similar to a DBMS, is used to keep the knowledge base updated, with changes to facts, figures, and rules. (P. 279) An explanation facility performs tasks similar to what a human expert does by explaining to end users how recommendations are derived. (P. 280) An inference engine is similar to the model base component of a decision support system. By using different techniques, such as forward and backward chaining, it manipulates a series of rules. (P. 280) In forward chaining, a series of “if-then-else” condition pairs is performed. (P. 280) In backward chaining, the expert system starts with the goal—the “then” part—and backtracks to find the right solution. (P. 281) Case-based reasoning (CBR) is a problem-solving technique that matches a new case (problem) with a previously solved case and its solution, both stored in a database. After searching for a match, the CBR system offers a solution; if no match is found, even after supplying more information, the human expert must solve the problem. (P. 283) Intelligent agents are software capable of reasoning and following rule-based processes; they are becoming more popular, especially in e-commerce. (P. 283) Shopping and information agents help users navigate through the vast resources available on the Web and provide better results in finding information. These agents can navigate the Web much faster than humans and gather more consistent, detailed information. They can serve as search engines, site reminders, or personal surfing assistants. (P. 284) Personal agents perform specific tasks for a user, such as remembering information for filling out Web forms or completing e-mail addresses after the first few characters are typed. (P. 284) Data-mining agents work with a data warehouse, detecting trends and discovering new information and relationships among data items that were not readily apparent. (P. 284) Monitoring and surveillance agents usually track and report on computer equipment and network systems to predict when a system crash or failure might occur. (P. 285) Fuzzy logic allows a smooth, gradual transition between human and computer vocabularies and deals with variations in linguistic terms by using a degree of membership. (P. 285) Artificial neural networks (ANNs) are networks that learn and are capable of performing tasks that are difficult with conventional computers, such as playing chess, recognizing patterns in faces and objects, and filtering spam e-mail. (P. 287) Genetic algorithms (GAs) are search algorithms that mimic the process of natural evolution. They are used to generate solutions to optimization and search problems using such techniques as mutation, selection, crossover, and chromosome. (P. 288) Natural-language processing (NLP) was developed so users could communicate with computers in human language. (P. 289) Contextual computing refers to a computing environment that is always present, can feel our surroundings, and—based on who we are, where we are, and whom we are with—offer recommendations. (P. 291) Instructor Manual for MIS Hossein Bidgoli 9781305632004, 9781337625999, 9781337625982, 9781337406925

Document Details

Related Documents

Close

Send listing report

highlight_off

You already reported this listing

The report is private and won't be shared with the owner

rotate_right
Close
rotate_right
Close

Send Message

image
Close

My favorites

image
Close

Application Form

image
Notifications visibility rotate_right Clear all Close close
image
image
arrow_left
arrow_right