Está en la página 1de 23

Unit 5:

Decision Making & DSS

•Understanding Business Decisions


•Making Rational Decisions
•Decision-Making Process: Herbert Simon Model
•Closed & Open Decision-Making Systems
•Decision Types & Nature
•Decision Alternatives Evaluation & Exercises
•DSS- Basic Concept & Understanding
•Generic Model & Elements of DSS
•DSS Classification/ Types; Benefits from DSS
•Data Mining Basics
•Expert Systems Overview

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 1
Business Decisions
1. Any Decision (Personal or Business-related) is associated with an Objective/ Goal. It involves
application of specific techniques/ algorithms and/ or Knowledge & Judgment/ Wisdom by the
decision maker, to available Information, for settlement/ conclusion/ resolution of some problem/
situation/ issue/ task.

2. Business Decisions have to be taken keeping in mind best interests of the Firm. However, the decision-
making is not done by the Firm (which has an artificial existence) but by PERSON(s), representing
it & acting on its behalf. Its important to understand this, since it brings in a PERSONAL TRAIT
factor in business decision-making.

3. Characteristics of Business Decisions:


Characteristic What it means
a) Sequential in nature Not isolated events, but linked to other earlier situations/ decisions
b) Can be quite complex, with risks/ trade-offs At higher management levels, decisions can involve significant
forecasting element, with many variables and alternative scenario
evaluations.
c) Is influenced by Personal traits/ values Specific personal traits (high risk taker, aggressive, conservative/
safe, etc.) can change a decision based on logic & principles.
d) Can be based on specific, temporal situational/ Generally associated with “fire-fighting” or “quick action” situations.
environmental factors May not involve complete/ comprehensive analysis.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 2
Making Rational Decisions
1. Decisions made based on related & available Information & which can be explained with logic & reasoning, are
Rational. On the other hand, Decisions based on “gut feel, instinct, etc.” are Irrational.

2. In reality, the 2 extremes – 100% Rational & 100% Irrational Decision – don’t exist! A degree of “personal
element” in any decision is unavoidable. Also, Information needed for a decision is almost never 100%
complete/ accurate & even logic/ reasoning can have assumptions/ inferences which may not be perfect.
Rationality (& Irrationality) are therefore not absolute, but relative terms, based on specific views or frame of
reference about what’s rational/ irrational. The same decision considered Rational by one Firm or person can
be considered otherwise by another Firm or person.

3. When using Rational or Irrational with regard to Decisions, its important to interpret as “more on the Rational/
Irrational side”. Herbert Simon defines 5 levels of Rationality in decision-making situations:

Personally Organisationally Consciously Subjectively Objectively


Rational Rational Rational Rational Rational
Irrational Rational

Ultimately, a Decision being right/ wrong gets determined by whether it achieved the associated Objective/
Goal. Its not necessary that Rational Decisions are always right and Irrational Decisions are always wrong.
However, probability of Rational Decisions being right is certainly higher & more importantly, such
decisions are not specific person dependant, can be analyzed for improvements in future decision-making.
Decision-making quality is therefore governed more by Rationality, rather than the outcome/ result.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 3
Decision-making Process:
Herbert Simon Model
 In Business, numerous situations arise, when a Decision-Maker has to examine number of possible
alternative solutions or courses of action, before reaching a decision to adopt one of them. For example:
Surplus Cash Handling, Increasing Cash Inflow / Reducing Cash Outflow, Scheduling (Plant/ Support
Staff/ Customer Order Deliveries), Marketing Channel Mix, Product Mix, etc.
 These Decisions can use complex mathematical/ statistical analysis models (for example: Linear
Programming, Transportation Model, Correlation, Regression, etc.) to arrive at possible alternatives.
Inputs need to be very specific & carefully validated, including various conditions that need to be met.
 The Herbert Simon Model identifies a 3-stage iterative process for such Decisions. Requirements of
multiple iterations are based on changes/ corrections in the Input or Analytical model, after review.

Define a problem for decision-making, based on


INTELLIGENCE examination of processed Data (Information).

Develop models/ algorithms to generate alternative


DESIGN ways to solve the problem (decision alternatives).

Select one alternative as the Decision, based on


CHOICE evaluation criteria applied to the alternatives.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 4
Closed & Open Decision Systems
Decision-making Systems can be considered as Closed or Open, based on how much is known
about the environment surrounding the decision situation: available alternatives, value
generated by each, and the goal/ objective criteria on which the choice will be made.

Closed Decision-Making System Open Decision-Making System

The decision alternatives are limited & a known set; All decision alternatives cannot be practically known and
outcome value of each alternative is also fully known outcome value of each alternative is also probabilistic.

Rules/ algorithms/ models for generating, feasibility Models & algorithms can be flexible and there’s no
testing & ranking of alternatives is known/ available. certain ranking of alternatives possible.

Choice of one of the alternatives (the Decision) is No clear rule-based method of reaching the chosen
pretty much based on clear selection criteria linked alternative, since Objective/ Goal may not be defined with
to the objective/ goal associated with the Decision. certainty

Typical Examples: Scheduling, Loan Sanction, Typical Examples: Diversification options, Pricing of a
Performance Appraisal processing, Direct Marketing new product, Ranking of Customers, Market reach
Target Lists, etc. expansion, etc.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 5
Decision Types & Nature
1. When viewed from the perspective of the degree of knowledge/ confidence about the future outcome,
Decisions can be classified as: Certain, Under Risk & Under Uncertainty.
2. Decisions can also be classified on the basis of their nature, as “Programmable” & “Non-Programmable”,
depending on how fully automated the decision-making process can be. In a fully Programmed Decision, not
only is the Decision-Making process automated till the final choice/ Decision, but even the associated action is
automated. For instance: Sanction/ Rejection of a Loan & communicating the Decision; Reorder decision for an
item based on stock level & preparing/ sending the PO to the Supplier.
3. There’s an obvious link between Closed & Open Decision-Making Systems and the type/ nature of Decisions
they’re associated with. The classification of Decision-Making Systems & Decisions also has a link with the
Management Pyramid & MIS Level.

Non-Programmable Under Risk, Uncertainty Open

Decision-
Nature of Type of
Making
Decision Decision
System

Programmable Certain Closed

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 6
Evaluating Decision Alternatives
In Decision-Making Systems, level of support/ automation provided by MIS depends on how Closed/ Open the
System is, certainty/ uncertainty of Decision Alternatives & their outcome values & extent to which final choice is
“rule/ procedure” based. Many models/ algorithms/ methods are available for evaluating Decision Alternatives &
aiding final choice/ Decision, for different types of Decision-Making situations:

Decision Alternatives Brief Description


Evaluation Models
a) Optimization Also called Operational Research (OR) models, they apply when decision-making system, even if
Models complex/ laborious, is Closed/ Deterministic/ Procedural. Linear Programming & all variants; Queuing,
Inventory & Project Mgt. Models etc. are examples. They find the optimum plan or resource allocation,
to achieve a goal/ objective, based on constraints/ restrictions. Eg: Product Mix, Scheduling,
Transportation & Assignment Models, Inventory planning & control Models. Queuing Models, Project
network & crashing etc.
b) Payoff Analysis Useful in risk/ uncertainty situations. Uses Payoff Matrix (rows are Decision Alternatives, columns are
possible outcomes/ events/ states of nature along with probabilities if available. Payoff or value from
each alternative-outcome combo is found & criteria like MaxiMax, MiniMax, Regret, Hurvicz Alpha &
Laplace are used for Choice. Examples: Price Change, How much to Stock, Product to focus on,
Marketing Campaign method choice, etc.
c) Decision Tree Useful when the decision-making situation and various alternatives are better depicted & understood
Analysis as a flow/ sequence of decision paths & points, for each alternative, with probabilistic outcome values
apparent at the end of the flow/ sequence. Compliments Payoff analysis when alternatives are many.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 7
DSS- Concept & Understanding
1. DSS (Decision Support System), like many terms/ concepts in
Management & IT, is subject to many interpretations/ views and
there exist various definitions of the term, based on differences in
the perception framework. A proper definition of DSS is based on
the perception framework that an IT System/ Application S/w is
used to support (not necessarily in entirety) the decision-making
process (Herbert Simon Model) beyond the 1st step (Intelligence).

2. A DSS can be defined as “an IT/ Application S/w System that supports
(partially/ fully, as appropriate), the Decision-Making Process (relating to
resolution of a problem/ situation/ issue) of generating & evaluating
alternative solutions (Decision Alternatives) & choosing one, as the
optimum/ best/ most suited solution or the Decision”.

3. DSS are a subset of the holistic MIS concept, one of the 3 forms that MIS can assume, with
others being Traditional MIS, EIS/ BI Systems

4. Often, the impression is that DSS are for higher level managers, which is incorrect. DSS work
at all levels in a business (Operations, Control & Strategy) & across managerial levels. DSS
have many forms & can be classified in a number of ways.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 8
DSS- Concept & Understanding
4. Complexity of a decision-making process, the extent to which it can be “automated” is defined
by how “deterministic” or “probabilistic/ uncertain” the problem & outcome of the decision are
or in other words, the type/ nature of the Decision (structured/ semi-structured/ unstructured or
under certainty/ risk/ uncertainty) and not by the complexity of computations/ processing
involved in generating/ evaluating Decision Alternatives and reaching the Choice.

Simple/ Deterministic/ Structured/ High: Even implementation of Choice


Under Certainty can be automated
Automation
in DSS

Type/ Nature
of Decision
Complex/ Semi & Unstructured/ Low: Significant “human element” in
Under Risk & Uncertainty evaluation of alternatives & Choice
Lower Middle Senior/Top
Managerial Level

Its important to remember that, in most decision situations, particularly at higher levels
of Mgt., “human element” in the form of experience & judgment is the key, despite the
best of enabling technology being used in the decision process through DSS!

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 9
DSS- Generic Model & Elements
When a DSS operates with significant Data volume,
its obviously preferable to prepare required Data in
the Data Warehouse, rather than enter online.
INTERNAL Data from Application Databases
& files, from various sources- ERP, Other
TPS & Stand Alone Systems

ETL

DSS S/w Tools:


• Operational Research Models
ETL: • Data Mining Tools
Extract, Transform, Load DATA WAREHOUSE • Decision Table/ Workbook Tools
(eg. Spreadsheets)

ETL

EXTERNAL Data
from various sources

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 10
DSS Classification/ Types
1. Classification based on the process/ managerial level:

Type Brief Description & Examples

Enable Staff, Supervisors & Managers responsible for day-to-day operational processes, to quickly reach a rule/
logic-based, sound & optimal decision. These DSS are highly automated or “programmable”, since there’s little
or uncertainty in problem definition, inputs & outcome of the decision. When appropriate, they are fully
Operations automated (actions relating to implementing the decision are also executed by the DSS itself). Some examples:
Level DSS - Deciding on items to re-order & how much, based on stock below reorder level & if required, generating PO.
- Deciding on marks/ grade/ result in an objective test/ exam and if required, initiating appropriate actions.
-Deciding on whether an applicant should be sanctioned loan or not & if yes, to what extent?
-Plant schedule based on confirmed Work Orders, Shipment plan based on deliveries to be executed, etc.
For decisions relating to longer timeframe (week/ month/ quarter), based on problems identified from various
Performance Reports or Opportunities. Risk/ uncertainty and lack of complete structure in the decision-making
creeps up. The DSS enables quick examination of alternatives and may/ may not suggest Choice, depending on
Tactical / level of risk/ uncertainty. Typical examples:
Control -Decisions relating to taking up a special one-time order or leasing out spare manufacturing capacity.
Level DSS -Tactics to increase closing deals (Price decrease, Value addition, Internal incentives…)
-Deciding target list for Direct Marketing, based on campaign objectives
-Tactics relating to “temporal demand”- how much to purchase & stock
-Revising Project plans, based on over-runs, change in scope etc.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 11
DSS Classification/ Types
Type Brief Description & Examples

Pertain to long-term plans & strategies (annual, 3-year, 5-year…). Can be highly unstructured with significant
uncertainty in all aspects- problem & goal definition, alternatives and choice. DSS only makes the process lot
more efficient, with rapid computations/ processing/ analysis, as needed. Decision is almost entirely dependant
Planning &
on the Senior/ Top Manager. Very often, the decision-maker uses a well-designed & modeled “What-If,
Strategy
Sensitivity, Goal Seeking & Searching canvas & has complete flexibility in the process. Some examples:
Level DSS
-Annual Budgeting & Resource Planning (Manpower, Finance)
-Long range forecasting, market strategies, product strategies, expansion & divestment decisions.
-Corporate Image Repositioning, Diversification decisions etc.

2. Classification based on the nature/ type of Decision:

Easily automated, since the entire decision-making process is without any risk & uncertainty; a clear step-by-
step process flow exists to reach the decision with optimum “value”. Most Operational Level DSS fall in this
Structured/ category. Though computations/ processing/ analysis may be complex (not at all a problem with IT), the overall
Programm- decision-making process is straightforward/ deterministic. Such DSS may also exist at Control & Strategic levels
able/ Rule- in cases where the decision-making is under certainty. For instance: Make-or-Buy decision based on fairly
Based DSS accurate costs & benefits, Optimum Annual Product Mix analysis, under the assumption that minimum &
maximum demand for each products and capacity/ profitability inputs for each product are reliable. Often, these
DSS, particularly at Operational Level are also automated for implementation of the Decision.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 12
DSS Classification/ Types
Type Brief Description & Examples

In many situations, inputs for a decision are probabilistic (eg. Demand for Product X is 10000 units with 60%
probability, 12000 with 25% and 15000 with 10%. Or, a couple of new projects with >90% probability of
finalization need 200 additional staff with certain skill sets & the probability of all selected candidates actually
Semi-
join is 70%). Such decision-making is “under risk” & the process is not cut & dry. Its also infeasible to figure out
Structured
& evaluate all alternative solutions. While the DSS helps with a recommended solution based on inputs, its upto
DSS
the decision-maker to evaluate more scenarios, use experience & judgment to reach a final decision. Typical
examples are Materials & Inventory Planning based on demand forecast, Bench strength planning in S/w
companies, Marketing Mix decisions etc.

In these situations, the probabilities of some or all inputs are not known. This renders decision-making to be
under uncertainty, rather than under risk. Semi and Unstructured DSS differ only with respect to the degree of
Un- risk and the level of confidence in the inputs & decision outcomes. There are number of techniques/ models for
structured decisions under uncertainty, which the decision-maker may need to apply & rely heavily on experiential
DSS judgment for the Choice. Significant % of decision-making at the highest levels of Management tend to be
Unstructured. Building some kind of “defined method” can be highly challenging and may happen only over a
period of time after actual experiences with previous decisions, based on certain approaches & techniques.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 13
DSS Classification/ Types
3. Classification based on Decision-Making Techniques and/ or Models:

Type Brief Description & Examples

Operational Research (OR) is the use of mathematical techniques to analyze & solve business problems,
through “optimal allocation of resources, based on constraints, to achieve a desired goal/ objective”. OR is most
applied in situations which are quite deterministic (decision-making under certainty or low risk, regarding inputs
OR-based & probability of outcome. OR-based DSS tools use number of “Models/ Algorithms”:
DSS -Linear Programming & all its variants such as Transportation Model, Assignment Model, for scheduling.
-Inventory Control, based on Reorder Level & EOQ (Economic Order Quantity).
-Queuing & Simulation models
-PERT (Project Evaluation & Review Techniques)
Can be applied for all types of decisions- under certainty/ risk/ uncertainty. The DSS tool computes Expected
value (Monetary or a Rating/ Score) for each alternative, based on appropriate procedures, formulae & rules,
using a decision table (along with a Decision Tree, if required) called “Payoff Matrix”. Choice is made on the
basis of maximum EMV/ EV (under certainty or risk) or criteria like MaxiMax, MiniMax, Regret, Hurvicz Alpha,
EV/ EMV-
Laplace etc. These DSS find wide-ranging use at all Managerial levels. Break-even analysis, Planning,
based DSS
Budgeting & ROI models to reach the most suitable plan, Make or Buy decision, Strategies/ tactics to achieve
specific goals such as revenue/ market share increase, cost reduction, response to specific opportunities etc.,
Customers to focus on, Supplier rating, Selection of personnel for recruitment/ promotion, Inventory item
classification such as ABC/ HML criticality etc. are some good examples.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 14
DSS Classification/ Types
Type Brief Description & Examples

Use various Data Mining techniques (Classification, Regression, Clustering & Association) & Models, for
Predictive Analysis in various Business Functions like Marketing & Sales, HR, Manufacturing, Customer Service
& Support etc. Demand forecasting, Qualification Criteria & Critical Factors determination are highly used
analysis areas. Outputs from these DSS may result in:
-Process improvements (Eg. Defining a Firm’s prospective customer, Qualifying Customer for upto Rs. 1 Lac
DM-based
loan without collateral, Inspection/ monitoring of manufacturing processes, Qualify/ disqualifying a candidate for
DSS
recruitment, Corrective actions to reduce employee turnover etc.) or
-Specific actions to achieve goals/ objectives (eg. Sending Direct Marketing communication as per Target List
recommended, Changes in customer focus based on future revenue forecasts, Revised items set for pickup
near billing counter at a Retail Store or for display to influence additional online shopping etc.) or
-Inputs for other DSS (Eg. Demand forecast is input for materials planning, inventory control, budgeting)

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 15
DSS Benefits
1. Using the power of IT, DSS make it possible for levels of analysis that Managers may tend to stay
away from (since its time-consuming & laborious, or due to lack of knowledge about possible
analytical techniques/ models), resorting to purely experiential & judgmental decision-making instead
– very subjective/ person-dependant.
2. With the advanced S/w tools available today, incorporating a wide range of techniques & multiple
models/ algorithms within each, the Manager has access to sound decision-making logic in areas,
where risk & uncertainty are involved. The focus shifts to ensuring that problem definition, inputs &
assumptions are proper & well-defined, which in turn improves quality of decisions.
3. Decisions can be changed pretty quickly, if outcome is not as expected, since evaluating other
alternatives and reaching a new Choice is much faster.
4. DSS which cater to structured/ programmable decisions can be completely automated, right upto
implementing the Choice.
5. Decision-making using DSS improves the overall Knowledge Base of the Organisation, based on
decision outcomes, resultant changes in the techniques/ models etc., all of which can be captured as
“Knowledge” for the future. In similar future situations, not much time gets spent in figuring out the
analysis model to be used.
6. Most importantly, DSS take the concept of MIS to a new level altogether, beyond just providing
Information, to the level of effectively assisting the actual solution process for a problem identified
through the other forms of MIS such as Traditional, EIS/ESS/BI.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 16
Data Mining Basics
Data Mining (also called KDD or “Knowledge Discovery from Data”) is the
use of various mathematical (& statistical) models/ algorithms, to analyze
huge repositories of historical data and the influencers &/or patterns
therein, to derive Information in the form of predictions about a desired
“Target Variable” or high impact “Data Clusters” / “Data Associations”, for
future decisions/ actions to achieve desired objectives & goals.

& TOOLS
Define Objective/ Goal &
Target Variable (if
applicable)

Run & Implement; Establish the Data


Review results Warehouse (DW)

Perform trial/ test runs & Identity Influencing


select desired DM Variables & Data Mining
Technique/ Model Techniques & Models

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 17
Data Mining Basics
Common DM Techniques/ Methods & Models/ Algorithms
1. Classification: Used when Target Variable is “categorical” & is classified into states (Low/ High or Low/ Medium/
High). Eg. Risk exposure at a level of Credit, Customer Relevance, Repurchase Likelihood etc. The purpose is to
find variables that influence the target variable & the strength of influence; reach a scheme to predict the target
variable. Some Classification Models used are C&RT, CHAID, Boosted Trees and Random Forests.
2. Regression: Similar to Classification, with difference being that Target Variable has a continuous value (Eg.
Revenue expected, Demand Forecast, Probability of purchase etc.). C&RT, MARSplines and Neural Networks
are common Regression Models.
3. Clustering: There’s no Target Variable as such. Instead, the objective is to find “meaningful clusters”, based on
particular outcome (Eg: Are customers who purchase a particular product clustered most around age/ gender/
income/ marital statusH? And within that, are there sub-clusters.. and so on. K Means, EM Algorithm and
Kohonen Network Clustering are some common Clustering Models.
4. Association: The objective here is find a list of well associated or correlated data based on a particular input,
using Association Rules & Models such as “6 5 5 5”, Apriori Algorithm etc. Eg. Customer buying or showing
interest in product X also tends to buy/ be interested in A, D, HH; A person with occupation as Plumber is
unlikely to be a Graduate or Post-graduate or Doctorate or with > x years of education, <certain age...

Data Mining S/w Tools:


Numerous S/w packages/ applications are available for Data Mining, which offer all the common techniques and
multiple models within each. SSAS, Microsoft SQL Server, Oracle Data Miner, Statistica, Prism, R, AdvancedMiner
etc.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 18
Data Mining Basics
Some Data Mining applications in Business:
1.Demand Forecasting over a period of time, for a particular product or class of products,
based on past sales data and other related inputs. Sales forecasting, territory-wise,
Customer-wise etc.
2.Credit worthiness of / Credit Risk with Customers & Suppliers.
3.Likelihood of machine breakdowns in a manufacturing plant, Cause Analysis of equipment
breakdowns in general for future preventing planning.
4.Target list of Customers for a Direct Marketing campaign, based on Clustering & analysis,
RFM (Recency, Frequency & Monetary Value) Analysis.
5.Media mix for Advertising in a particular territory or for a particular campaign objective.
6.Influencing interest & purchase behaviour of Customers in Retail Stores, through
Association Lists at billing counters, during e-Purchases & e-Searches.
7.Identifying potentially erroneous or suspicious data, which is not obvious, during online or
web-based form filling. This can improve data quality significantly.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 19
Expert Systems- An Overview
1. The Artificial Intelligence (AI) field evolved in the 90s & has picked up pace rapidly since then. AI
attempts to develop computer-based systems (both H/w & S/w) which can behave in some way, like
human beings & create “Intelligent Machines”, able to perform certain tasks independently, once the
initial set up is done.

2. AI Systems capture, codify & store very specific areas of human expertise, knowledge & reasoning, to
automate (to extent feasible) physical & mental tasks/ processes done by humans, for benefits such as:
- Immensely higher speed
- Execution unhampered by emotion, fatigue, worry, stress, environmental dangers etc.
- Eliminating dissatisfaction & loss of efficiency, due to routine/ boredom/ repetition.
- Enhancing the overall Knowledge Base; reducing dependence on specific individuals.

3. Expert Systems constitute one of the 4 streams into which the field of AI is broadly divided:

While all forms of AI Systems are widely used, Expert Systems are relevant from the MIS perspective,
since they can be considered as a form of DSS & hence, as inclusive under the overall MIS concept.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 20
Expert Systems- An Overview
An Expert System (ES) is an AI System in which a set of criteria & thumb rules (Knowledge) a Human
“expert” uses as the process flow or method, to reach a particular decision in a specific domain/
area, is captured & stored in a Knowledge Base as a Set of Rules to be applied in a defined logical
flow & accordingly programmed/ coded into an Application S/w package (ES S/w), which can then be
independently used to reach the same decision.
• Inference Engine contains sequence/ flow in which
rules from Knowledge Base must be applied &
accordingly captures inputs from the User, through
the Interface, through questions in proper order.
• Each User response triggers appropriate actions
(another query to User for additional inputs or
communicating a decision/ advice to User based on
inputs captured thus far & concluding the process).
• For an ES to be effective, its domain or “area of expertise” must be
carefully “limited”/ “narrowed”. Rule Set (IF-THEN constructs) can be
large, but outcome of every Rule application must be certain.
• If used properly, ES can significantly increase efficiency of many business
processes in Sales, Customer Service, HR etc., by reducing involvement
of a Human Specialist & even eliminating it in many transactions.
Expert System Shell
• Many ES also use Fuzzy Logic in Rule application, for > effectiveness

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 21
Expert Systems- An Overview
Examples of Simplified Expert Systems- Qualifying for Loan & Granting Discount to Distributor on Bulk Order
A-B-X B-C-X C-D-X
D
If Income>50000 If Car Payment<10% of Income If Mortgage Payment <20% of Income
Grant Credit Line, Ask
Ask about Car Payments Ask about Mortgage Payment Do D
about years employed
Else Do X Else Do X Else Do X

F E-F-G
X Credit Limit 10000 If Yrs >=4, do F
G-F-H
Decline Credit & Exit Else Ask about other Debt
H If Other Debt<5% of Income do F
Credit Limit 3000 ELSE Do H

A-B-X E-F-I H
If Order>20L, Do B, Else Do X If Qty of Item Z>250, Do H Else Do F Grant 15% discount on Order Value

B-C-X F-I-G I
X If Order within 7 days, Do C, Else Do X If Qty of Item Z>150, Do J Else Do G Grant 12% discount on Order Value
Decline
Discount C-D-X G-J-K J
If Qty of Item X>500, Do D, Else Do X If Qty of Item Z>75, Do K Else Do L Grant 9% discount on Order Value
D-E-X K
If Qty of Item Y>300, Do E, Else Do X Grant 6% discount on Order Value

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 22
Expert Systems- An Overview
Some examples of areas where ES have been or can be applied:

1. Preliminary medical diagnostics, for a person to get advice on action to take next, or to
save a Doctor’s time, prior to a patient actually presenting himself/ herself for diagnostics.
2. Preliminary qualification of an applicant, for a loan and the amount entitled to.
3. Fault diagnosis in products/ equipment: PCs, Appliances, Cars, Aircraft etc.)
4. Identifying consignments to be inspected closely; identifying low-risk consignments for
passing without inspection.
5. Deciding on suitability of a candidate for further steps in the recruitment process, based
on certain preliminary qualification criteria/ rules.
6. Granting discount level for bulk orders from Distributors, based on order value & product
mix criteria.
7. Suggesting a particular product/ model to a Customer, from numerous options, based on a
flow of requirement inputs gradually narrowing down the choice to the optimum one.

Presentation content & material developed for exclusive use by V. Mohan. All rights reserved. No unauthorised copying or usage permitted. 23

También podría gustarte