28
Artificial Intelligence for Automated Decision Support Valerii Klymchuk

Artificial Intelligence for Automated Decision Support Project

Embed Size (px)

Citation preview

Artificial Intelligence for Automated Decision

SupportValerii Klymchuk

AI is an area of computer scienceArtificial Intelligence is a collection of concepts and ideas for development of intelligent systems in different areas and domains.

Artificial Intelligence is concerned with 2 basic ideas:• What is intelligence? (study of human thought process).• Representation and duplication of the thought processes in machines.

Alan Turing designed a test to determine whether a computer exhibits intelligent behavior - the Turing Test: “A computer can be considered smart only when a human interviewer cannot identify the computer while conversing with both: an unseen human being and an unseen computer.”

Abilities considered to be signs of intelligence:• Learning or understanding from experience.• Making sense out of ambiguous or contradictory messages.• Responding quickly and successfully to a new situation (flexibility).• Using reasoning in solving problems and directing conduct effectively. • Understanding and inferring in a rational way.• Applying knowledge to manipulate the environment.• Thinking and reasoning, Dealing with perplexing situations.• Recognizing and judging the relative importance of different elements

in a situation.

AI: Field Applications

AI: Disciplines

Knowledge engineering is a collection of activities:

• for the acquisition of knowledge from human experts and other sources, and conversion of this knowledge into a repository.

• to help experts articulate how they do what they do and to document this knowledge in usable form.

It often applies principles and tools of Artificial Intelligence.

Knowledge Engineering in Rule-based SystemTwo types of rules are common in artificial intelligence:• Knowledge rules, or declarative rules, state all the facts and relationships

about the problem.• Inference rules, or procedural rules, offer advice on how to solve a problem,

given that certain facts are known. Knowledge rules go to the knowledge base, whereas inference rules become part of the inference engine. The inference engine uses the rules and facts to draw conclusions. It directs the search through the collection of rules in the knowledge base (pattern matching). When all the IF parts of a rule are satisfied, the rule said to be fired. The new knowledge generated by the rule is inserted into the memory as a new fact. Engine keeps doing so until the goal is achieved.

Knowledge rules might look like:• Rule 1: IF an international conflict begins, THEN the price of gold goes up.• Rule 2: IF the inflation rate declines, THEN the price of gold goes down.Inference (procedural) rules:• Rule 1: IF the data needed are not in the system, THEN request them from

the user. • Rule 2: IF more than one rule applies, THEN deactivate any rules that add

no new data.A set of activities for the acquisition of knowledge from human experts and other sources, and its conversion into a repository is called Knowledge Engineering. It applies principles and tools of Artificial Intelligence.

The Process of Knowledge Engineering

Inference Mechanisms (or reasoning) in Rule-based Systems:• Backward chaining is a goal-driven approach in which you start from

an expectation of what is going to happen (i.e., hypothesis) and then seek evidence that supports (or contradicts your expectation. Often, this entails formulating and testing intermediate hypotheses (or subhypotheses).• Forward chaining is a data-driven approach. We start from available

information as it becomes available or from a basic idea, and then we try to draw conclusions. The ES analyses the problem by looking for the facts that match the IF part of its IF-THEN rules. As each rule is tested, the program works its way toward one or more conclusions.

• Feedforward (multilayer perceptron with backpropagation).• Recurrent Networks - a pictoral representation • Associative Memory• Kohonen’s Self-Organizing feature Maps• Hopfield Networks

Artificial Neural Network (ANN) Architectures:

Ultimately, the architecture of a neural network model is driven by the task it is intended to carry out.

Hopfield network

Elements of ANN:• A neural network is composed of

processing elements, organized in different ways to form the networks structure. • The basic processing elements (PE)

of an ANN are artificial neurons. Each neuron receives inputs, processes them, and delivers a single output, as shown. Processing information in an Artificial Neuron.

General ANN Learning Process• In supervised learning, the learning

process is inductive; that is, connection weights are derived from existing cases. The usual process of learning involves three tasks:

The information flows unidirectionaly from input layerto hidden layers to output layer.

Feedforward (multilayer perceptron with backpropagation) Network

Architecture, where the connections between the layers are not unidirectional; rather, there are many connections in every direction between the layers and neurons.

Recurrent Networks - a pictorial representation

A Recurrent Neural Network Architecture

• Provide a way to represent multidimensional data in usually in one or two dimensions.• Learn to classify data

without supervision (i.e., there is no output vector).• Commonly used for

clustering tasks due to self-organizing capability.

Kohonen’s Self-Organizing Feature Maps (SOM)

4 x 4 nodes connected to the input layer (with three inputs) representing a two-dimensional vector

Applications. NN models have been used as:• Classifiers. Typically are multilayer models in which information is

passed from one layer to the next, with a goal of mapping the input to the network to a specific category, as identified by an output of the network.• Forecasting tools.• Customer segmentation mechanisms.• General optimizers. In contrast, can be a single layer of neurons,

highly interconnected, and can compute neuron values iteratively until the model converges to a stable state. This stable state represents an optimal solution to the problem under analysis.

Visual Interactive Models (VIM)• Systems developed for the military and the video-game industry have

thinking characters who can behave with a relatively high level of intelligence in their interactions.

The VIM approach can be used in conjunction with Artificial Intelligence. Integration of the two techniques adds several capabilities that range from the ability to build systems graphically - to learning about the dynamics of the system.

Case: Predictive integrated modeling and decision support for Power Generators by StatSoft:• Optimizes operation ParametersProblem: A coal-burning 300MW multicyclone unit required optimization for consistent high flame temperatures to avoid forming slag and burning excess fuel oil.Results: After optimizing the control parameters, flame temperatures showed strong responses, resulting in cleaner combustion for higher and more stable flame temperatures.• Predicts problems before they happenResults: Optimized settings resulted in consistently lower NOx emissions with less variability and no excursions.• Reduces Emissions (NOx, CO)Results: After optimization, NOx emissions under low-load operations were comparable to NOx emissions under higher loads.

Case: Watson• Facilitates evidence based support at MSKCC.• It learned the process of diagnosis and treatment through natural language

processing. Was trained to gain knowledge by comparing an individual patient’s medical information against a variety of treatment guidelines, published research, and other insights. • Provides individualized, confidence-scored recommendations to the physicians.

Provides a platform to look at the case from different angles. • Its voice capabilities allow physician to speak to Watson.• Watson also Assists the insurance providers in detecting fraudulent claims• Provides approval for medical treatments based on clinical and patient data at

WellPoint insurance provider.

Knowledge-based Management System• In the medical field amount of medical information doubles every 5

years. This massive growth limits physician’s decision-making ability in diagnosis and treatment. Patients histories and electronic medical records can be analyzed in combination with existing medical knowledge.• One of the most widely publicized knowledge-based DSS is IBM’s Watson

system. It employs such techniques as: natural language processing; hypothesis generation and evaluation, evidence-based learning.• Watson successfully played Jeopargy television show and beat the other

human competitors. Later it evolved into a question-answering computing platform that is being used commercially in the medical field.

Knowledge-driven DSS(KBDSS) or an Intelligent Decision Support System (IDSS) is a system that integrates knowledge from experts.• Involve application of knowledge technologies to address specific

decision support needs. • Are utilized in the creation of automated decision-making process.• Rules are used to automate the decision-making process. These rules

are either an expert system (ES) or structured like one. Rule-based expert systems – a technique developed in the area of artificial intelligence are the foundation for building KBDSS.• All intelligence-based DSS fall into this category including ANN and ES.

Expert Systems Features, Symbolic ReasoningSuch computer-based information systems use well-stored, organized and quickly retrievable expert knowledge to improve productivity and quality of performance. ES must have following features:• Symbolic Reasoning is the basic rationale of Artificial Intelligence. Knowledge must be represented

symbolically, and primary reasoning mechanism must be symbolic, rather then mathematical calculation. Reasoning mechanisms include backward chaining and forward chaining.

• Expertise. Associated with a high degree of intelligence and vast quantity of knowledge. • Deep knowledge: knowledge base should not be trivial for non-experts. • Self knowledge. ES must be able to examine its own reasoning and provide explanations as to why a

particular conclusion was reached. It should be able to learn from its past success and mistakes.First generation ES use if-then rules to represent and store their knowledge. The second-generation ES are more flexible in adopting multiple knowledge representations and reasoning methods. They may integrate fuzzy logic, neural networks, or genetic algorithms with rule-based inference to achieve a higher level performance. Expertise is the extensive, task-specific knowledge that an experts possesses.

Automated Decision System (ADS or DAS )• New approach to supporting decision making.• Attempts to automate highly repetitive decisions based on business rules and are mostly

suitable for frontline employees who must make quick decisions. • Is a rule-based system that provides a solution in one area to a specific repetitive problem

usually in one industry (e.g., to approve or disapprove a request for a loan; to determine the price of a store-item, etc.) ADS provide a rule-based solution.

The following are examples of business rules: • “If only 70 percent of the seats on a flight are sold, offer X discount to nonbusiness travelers,”• “If an applicant owns a house and makes over $ 100,000 a year, offer a $10,000 credit line.” Such rules are based on experience or derived through data mining, can be instantly applied to problems (e.g., “Based on the information provided and subject to verification, you will be admitted to our university”), or rules can be provided to a human, who makes the final decision.

Automated DSS facts:• Make the decisions in real time or near-real time.• Almost all airlines use automated DS to assign dynamic prices based

on actual demand. ADS initially appeared in the airline industry where they were called revenue (or yield) management systems.• Giant Food Stores worked with DemandTec to deploy a system for its

pricing decisions. The system handles massive amount of point-of-sale and competitive data to model and forecast demand, handles a large amount of price changes without increasing staff. Giant’s productivity has doubled as a result.

General Architecture of Automated Decision Systems

ReferencesBusiness Intelligence and Analytics. Systems for Decision Support, 10th,

Edition, 2015, Sharda, Delen, Turban, Pearson, ISBN: 978-1-292-00920-9