Fichas
CBR
Fichas del conjunto (59)
- 1CBR Advantages of using CBR-Knowledge acquisition is improved
-System development time is faster
-Existing data and knowledge are leveraged
-Formalized domain knowledge is not required - 2CBR advantages cont-Experts feel better discussing concrete cases
-Explanation becomes easier
-Acquisition of new cases is easy
-Learning can occur from both successes and failures - 3Issues and challenges of CBR-What makes up a case?
-How can we represent cases in memory?
-Automatic case-adaptation can be very complex!
-How is memory organised (the indexing rules)? - 4CBR-How can we perform efficient searching (i.e., knowledge navigation) of the cases
-How can we organise the cases?
-The quality of the results is heavily dependent on the indexes used - 5Machine Learning Supervised-Explanation-based learning
-Statistical regression
-Inductive learning
-Case-based reasoning
-Neural networks
-Genetic algorithm - 6Machine Learning Unsupervised-Neural networks
-Genetic algorithm
-Clustering - 7Machine-learning concepts and definitions-A major focus of machine learning research is to automatically learn to recognise complex patterns and make intelligent decisions based on data
- 8Machine-learning concepts and definitions-Machine learning
The process by which a computer learns from experience (eg, using programs that can learn from historical cases)
Induce expert knowledge from historical cases - 9Machine-Learning TechniquesHuman learning is a combination of many complicated cognitive processes including:
Induction
Infer by reasoning from particular cases to
general conclusion
Probable based on the evidence given
Eg This ice is cold : - 10Machine-Learning Techniques 2Therefore all ice is cold
3+5=8 and eight is an even number. Therefore, an odd number added to another odd number will result in an even number.
Weak Induction
This snake is not poisonous : All snakes are not poisonous = Overgeneralization
Strong Induction
This blackbird is black: All blackbirds are black : probably true (males anyway!) - 11DeductionInfer from given facts
High degree of Certainty
All Men are mortal; Martyn is a man; Therefore Martyn is mortal.
The premise has to be true
Everyone who eats steak is a rugby player
Martyn eats steak therefore Martyn is a rugby player: false. - 12BPM DefinedA real-time system that alert managers to potential opportunities, impending problems, and threats, and then empowers them to react through models and collaboration
- 13BPM-Budgeting forecasting & planning
-Commercial analytic application
-Custom analytic applications
-Enterprise reporting & scorecards - 14BPMWhat goes in the middle.
-Performance management strategy
-Data warehousing
-ETL & data integration - 15A Closed-Loop Process to Optimize Business Performance1-Strategize: Mission values and goals objectives incentives strategy maps.
2-Plan: Budgets plans, forecasts, models, Initiatives, targets
3-Monitor/analyse: Performance dashboards reports analytical tools - 16A Closed-Loop Process to Optimize Business Performance4-Act/adjust: Interpret, collaborate assess, decide, act, adjust & track
Each with its own process steps… - 17CBR own knowledgeCBR works under the assumption that all the problems have similar characteristics thus be able to apply solutions that have worked in the past on similar problems
- 18Intelligent Agents:-perform a set of operations on behalf of user or IS
-have some degree of independence
-have knowledge of user’s wishes
-are autonomous – empowered
-Senses the environment - 19Why use Intelligent Agent?People becoming overloaded with information.
Amount of data collected by large organisations doubles every year.
Knowledge workers can only analyse about 5% of data.
60% of time spent trying to discover patterns in data
30% of time spent analysing patterns
10% of time spent acting on patterns
Information overload reduces decision-making capability - 20Simple Reflex agentIntelligence
Reactivity – Perceive their environment and respond in a timely fashion to changes that occur
Proactiveness – Take the initiative
Social Ability - Interact with other agents
Autonomy – cont - 21Reasons for Success of AgentsDecision Support
Support Knowledge workers
Frontline decision support
Support frontline workers
Repetitive Office activities
Automate repetitive tasks
Mundane Personal Activity
Automate mundane tasks
Search & Retrieval
Domain Experts - 22Intelligent agentIn the information age, there is information overload occurring intelligent agents help to go through the information to retain the information that is relevant to the problem domain
- 23Intelligent agentsHelps managers or decision makers to identify the information they should take action on
- 24Level of intelligenceLevel 0
Agents retrieve info under strict orders
eg Web browsers where the user specifies URL - 25Level of intelligenceLevel 1
User-initiated searching facility
Referred to as Search Engines (eg Google) - 26Level of intelligenceLevel 2
Maintain user profiles
Monitor Internet info
Alert user when info found/new info
Referred to as Semi-Intelligent Agents
Confused.com - 27Level of intelligenceLevel 3
Include Learning and deductive component
Assist user who can not formalise a query
Referred to as Truly Intelligent Agents
Or Reconnaissance Agents
Eg Letizia - 28Components of an agentOwner
Username, parent process name
Author
The development owner
Account
Anchor to owner’s account for billing purposes
Goal
The objectives
Subject description
Boundaries, context
Creation and duration
Dates of request creation and dates info required by
Background
Any Supporting information
Intelligent subsystem
Underpinning ANN or ExS - 29CharacteristicsGoal orientated
Accepts high-level requests
Decides how to satisfy those requests
Collaborative
Can modify requests
ask clarification
work with other Intelligent Agents
Flexible
Dynamically which actions to choose
responds to the environment
Self-starting
Sense changes in environment and decides when to act - 30BI Organisational impactsLarge numbers of people participating in decisions
Decrease in face-to-face meetings
Less time spent in meetings
Fewer organisational levels involved in authorising actions
More rapid identification of problems and opportunities
More efficient use of organisational databases
Higher quality decisions - 31Individual ImpactsJob satisfaction
Job content
Job flexibility
Demand for experts
More time planning, less time firefighting - 32Social ImpactsEmployment Levels
Unemployment higher in non-industrialised countries
Work expands to fill the time available
Automation is an ongoing process
Some jobs may not be automated
Automation reduces costs thereby increasing demand - 33Other Social ImpactsWork in hazardous environments
Opportunities for disabled
Opportunities for women
Working from Home
Improvements in health
Aids for the Consumer
Increased Leisure time
Law enforcement - 34Potential negative effectsComputer crime and internet fraud
Too much power
The dangers of the web (the dark side of the “Force”)
Blaming the computer phenomenon
Social responsibility - 35The Future of BIBI will play a larger role in strategic management due to BPM systems
Increased use of the Web for BI development and applications
BI is being combined with other Web-based enterprise systems (CRM, ERP, and KM)
Intelligent systems have become major contributors in the fight against terrorism and fraud, as well as contributors to productivity and quality of life improvements - 36The Future of BIWeb-based advisory services will mushroom on the Internet
More complex BI applications are expected—so more complex problems will be solved.
The trend toward making BI more intelligent will continue
Advancements in mobile, wireless, and pervasive computing require the use of a greater and more diversified number of intelligent devices - 37The Future of BIBI will be available for dissemination via “The Cloud”
Natural language-based search engines will populate
the Internet, facilitating BI construction by reducing costs
The semantic Web will increase the use of BI
The use of voice technologies and natural language processing will further facilitate the usage of BI - 38Cloud ComputingA style of computing in which dynamically scalable and often virtualized resources are provided over the Internet
Users need not have knowledge of, experience in, or control over the technology infrastructures in the cloud that supports them
Related terms: utility computing, application service provider grid computing, on-demand computing, software as a service (SaaS)
Cloud = The Internet (as it is represented in diagrams) - 39Cloud Computing BIFragments of cloud computing
Infrastructure as a service (IaaS)
Platforms as a service (PaaS)
Software as a service (SaaS)
Data as a service (DaaS)
Example: Web-based e-mail and Google Docs
Cloud-computing service providers
Salesforce.com, IBM, Sun Microsystems, Microsoft (Azure), Google, and Yahoo!
Different service compensation models exist - 40Cloud Computing BICloud computing, like many other IT trends, has resulted in new offerings in business intelligence
Cloud-based data warehousing (by 1010data, LogiXML, Lucid Era)
Cloud-based dashboard and data management tools (by Elastra, Rightscale)
Advantage: rapid diffusion, cutting-edge technology, less investment,…
Concerns: loss of control and privacy, legal liabilities, cross-border political issues - 41BPM1-Strategize: Mission values and goals objectives incentives strategy maps.
- 42BPM2-Plan: Budgets plans, forecasts, models, Initiatives, targets
- 43BPM3-Monitor/analyse: Performance dashboards reports analytical tools
- 44BPM4-Act/adjust: Interpret, collaborate assess, decide, act, adjust & track
- 45DATA MINING TYPES OF PATTERNSAssociation
Identifies relationships that occur at one time (eg contents of a shopping basket)
High probability of repetition
Develop an associative algorithm
eg 80% of records that contain A, B, C also contain D,E.
Market Basket Analysis
Analyses thousands of basket purchases to derive associations
Many intuitive but some not: (eg Nappies and Beer!) - 46DATA MINING TYPES OF PATTERNSPrediction
Prediction
Predict nature of future occurrences based on events in the past. Eg Who will win Grand National based on “the Going”
Example
Using predictive data mining on their mountains of data, Caterpillar's quality managers uncovered the manageable process parameters most important to the quality outcomes of finished products. Tribal knowledge, which contributed to process variation, was now consolidated and focused by providing a line of sight between the manufacturing processes and the finished product. By distilling the relevant relationships from a large candidate pool of data, the manufacturing process was fine-tuned to consistently produce a finished product that met engineering and customer requirements without added costs. - 47DATA MINING TYPES OF PATTERNSCluster (segmentation)
Identifies groups of items that share a particular characteristic
All members of a set share some attribute
Objects grouped together because of similarity
eg credit card purchases on a weekday and credit card purchases on a weekend
People who live in a particular area may behave in a particular way (GIS) - 48DATA MINING TYPES OF PATTERNSSequential (or time series) relationships
Identifies relationships that exist over a period of time
eg People who buy a new TV may then buy a new video player 3 weeks later.
eg People who purchase a holiday may later buy holiday insurance
Analyse credit card transactions
Similar sequence events
eg when oil stock prices follow a trend pharmaceutical stocks follow the same trend. - 49DATA MININGPulling data from the DW to find patterns
To find things that are not immediately obvious
It's like mining for gold it's not obvious at the face surface but you have to dig down deeper - 50Modeling BenefitsCost of modelling is lower than experimentation with real system
Models enable the compression of time
Manipulation of variables is easy
Cost of mistakes is low
Managers can calculate risks involved in making specific actions
Models reinforce learning
Models act as a communication vehicle - 51Modelling ProcessIntelligence
Finding/Discovering the problem
The difference between actual and desired output.
Does the problem really exist
Where is it located
How significant is it?
Classification
Structured, semi-structured, unstructured
Statement
What the problem is
To whom does it belong - 52Modelling ProcessDesign – Developing alternatives
Decision-making under certainty
Complete information available
longer the time span more uncertain
Decision-making under risk
Risk (probability) attached to certain courses of action
Decision-making under uncertainty
Unable to estimate probability
very little information available
difficult to structure problem
Outcome depends on experience and attitude to risk - 53Modelling Process – DesignInventing, Developing & Analysing alternative courses of action
Components
Result variables
Decision variables
Uncontrollable variables
Structure
The principle of choice
Optimisation
Highest level of goal attainment for a given set of resources
Lowest cost
Alternative with highest ratio of goal attainment per unit invested
Study impact of each alternative on whole organisation
Sub-optimisation
Study a department in isolation
only partially evaluate the problem
Satisficing
A satisfactory solution
Not absolute best - 54Executive ISText & time series information add value
Provides drill down for quick access to detailed information
Ease of Use is critical for reducing initial tracking and for reducing reorientation time when the IS is used on an occasional basis
Extensively uses data on the organisations environment
Multiple search engines
Multidimensional data mining
Electronic messaging, calendar & schedule - 55EISDsicussion
- 56EISPlanning
- 57EISExecute and control
- 58EISTracking
- 59EISPerformance monitoring
Información
- Descripción
- Fichas sobre CBR, creado por Tongai Blessing Matongorere el 05/05/2016.
- Creado por
- Tongai Blessing Matongorere
- Visitas
- 1
- Valoración