Systems Dramatically improved business processes which reduce costs Better coordination and support for individuals and groups in the firm The ability to capture and retain knowledge that exists in the organization Measures of the Contribution of Decision and Intelligent Systems Improvements in efficiency Better customer service Cost savings per project Decision Support Systems (DSS) A computer-based system that helps the decision maker utilize data and models to solve unstructured problems E.g., a spreadsheet package DSS can be classified as data-oriented provide tools for the manipulation and analysis of data model-based generally have some kind of mathematical model of the decision being supported DSS Examples American Airlines Yield Management maximizes the revenue or yield from each flight through overbooking, discount seats, and traffic management resulted in total quantifiable benefits of more than $1.4 billion for AA Pfizer distribution system supports decisions about the US distribution network for distributing finished goods, including warehousing, transportation and ultimate delivery to the customer returns are hard to estimate Other DSS Examples General Motors OnStar a two-way vehicle satellite communication system offering a number of services for safety, security and entertainment GM has an 80% share of the telematics market, and OnStar has a market value of several billion dollars Merrill Lynchs Integrated Choice Account Structure helped design appropriate account structures and pricing for the company Integrated Choice program analysis considered the total revenue at risk, estimated what accounts customers would choose, and the impact of their choice on revenues Helped the company increase assets and customers Executive Information Systems (EIS) EIS bring to senior management information that needs its attention summarize data and makes them available for downloading to a personal computer in the executive's office have an appealing interface and an easy to use system drill down feature Group Decision Support Systems (GDSS) GDSS consist of special software and physical facilities such as a conference room containing PCs for each person in the room The software helps identify issues and evaluate alternative decisions and actions Enable workgroup decision making Technology-assisted meetings Intelligent Systems Turings test for Artificial Intelligence (AI) place a computer and a human in two separate rooms an interviewer in a third room, who cannot see the human or the computer user, asks questions that are passed to the computer and to the human if the interviewer cannot tell the difference between the answers from the computer and the human, the machine is said to exhibit intelligent behavior AI Versus Traditional Programs AI programs manipulate symbols rather than numbers AI programs are non-algorithmic often employing heuristics or rules of thumb Many AI programs are concerned with pattern recognition Applied AI : Expert Systems Advisory programs that attempt to imitate the reasoning process of human experts Reasons to build Expert Systems to make the expertise of an individual available to others in the field to capture knowledge from an expert who is likely to be unavailable in the future to provide consistency in decision making Components of Expert Systems The user interface a good interface make the system much more pleasant to use and helps promote its acceptance The knowledge base For example, the use of rules to represent the experts knowledge a rule in a knowledge base contains some of the logic of an application The inference engine the reasoning part of the expert system
The Inference Engine Forward chaining involves going through the rules one at a time to infer the best recommendation In backward chaining, the system begins with a goal and at each stage, the inference engine establishes sub-goals Expert System Development Prototyping Learning through test cases The person designing the system is sometime referred to as the knowledge engineer Knowledge engineering is difficult because experts are not conscious of decision-making steps and have difficulty explaining their logic in some cases, experts are reluctant to reveal their expertise to systems developers Examples of Expert Systems AESOP: A System for Stock Options Pricing Examples of Expert Systems The Port of Singapore Authority Expert Systems planning and managing all operations of the port E.g., allocating berths to ships, planning the stowage of containers, the allocation of resources in general, and reading container numbers and operating trucking gates managing shipping traffic and the activities of the port E.g., assigning ships to anchorages, scheduling the movement of vessels through channels to terminals, deploying pilots to tugs and launches, routing launches, and deploying tugboats Knowledge Discovery Combines AI with large databases Knowledge discovery programs look for patterns in the data and report the results to the user Neural Networks The first neural networks were loosely based on how the brain functions The most popular type of such programs is used to classify input into different categories A neural network has to be first trained by presenting it with past cases After training the network can be used for classification Case-Based Reasoning Captures lessons from past experience and uses them to find solutions to a new problem It is both a problem-solving approach and a model of how some experts think individuals learn, remember, and think about problems A CBR needs cases, a similarity index, a case retrieval mechanism, and an explanation module Genetic Algorithms Involves generating a population of possible problem solutions and rating them based on some fitness function applying a selection function to the un-rated population to select ``parents'' for the next generation of solutions using a reproduction function to generate copies of the parents The inversion operation reverses the order of randomly selected, contiguous portions of the vector A point mutation alters a single feature, replacing it with a randomly chosen value The crossover operator randomly selects a sequence of features and swaps them between two parents Intelligent Agents An agent is a piece of software that performs a task for its owner involves AI combined with networks applications for intelligent agents have been for consumer tasks like shopping and providing recommendations based on profile matches Summary IT can do more than process transactions and help people communicate Some of the decision and Intelligent Systems applications carry high risks but can also lead to substantial rewards