Loading

Module 1: Managerial Support Systems

Notes
Study Reminders
Support
Text Version

Models in Decision Support Systems

Set your study reminders

We will email you at these times to remind you to study.
  • Monday

    -

    7am

    +

    Tuesday

    -

    7am

    +

    Wednesday

    -

    7am

    +

    Thursday

    -

    7am

    +

    Friday

    -

    7am

    +

    Saturday

    -

    7am

    +

    Sunday

    -

    7am

    +

Today, we are going to talk about ‘Models in Decision Support Systems’. We have already
discussed about models, we will slightly go into some more details.
(Refer Slide Time: 00:37)

The concepts that will be covered today will be the definition of a model, classification
of models, purpose of modeling in decision support system and we will slightly give you
an idea about certain terms which are widely used in DSS, those are optimization,
heuristics and simulation.

(Refer Slide Time: 01:07)

When we talk about models in decision support systems, we must know that the ability
to generate and compare various alternatives that have been formulated to solve the
problem under consideration is mainly supported by these models. Models they have the
ability to generate and compare various alternative solutions procedures. Models provide
such an ability and that is why models are considered to be an important component of
decision support systems.
Model, they basically processes the data which resides in the database of a DSS and then
this models they generate meaningful output which can influence the decisions. Because
it is the managers who will interpret this models, results of this models and then decide
that which solution is the one that they will accept.

(Refer Slide Time: 03:03)

A system A is said to be a model of another system B, if A explains easily or represents
some aspect of B by concentrating in itself those qualities of B which are essential for
the aspect under consideration. Once again I repeat a system A is said to be a model of
another system B. If A explains easily or represents some aspect of B by concentrating in
itself those qualities of B which are essential for the aspect under consideration.
For example, a road map of a city can be called a model of the city with respect to the
problem that is getting studied. If the map is meant to be used by a motorist then it shows
only the roads and their lengths in different regions of the city, but it will not concern
itself with the population density or the price of commodities because they are not
relevant with respect to the problem in hand. Thus abstraction is a very important feature
of any model.

(Refer Slide Time: 05:10)

The kind of abstraction which are used in a model will depend on the real system that is
getting represented by the model. And most important is the purpose for which the model
is built.
(Refer Slide Time: 05:57)

The models that are used in decision support systems particularly in the area of
management are usually logical or mathematical in nature. Models in the area of
management mainly consist of equations, constraints, initial values, final values that is
the boundary conditions, the operational rules and so on. Once again, models are made
up of equations, constraints, initial values, final values and the operational rules. So,
boundary conditions are very important.
(Refer Slide Time: 07:19)

Now, we discuss about the classification of models, the different classes of models that
we use in decision support systems. First according to function descriptive models. They
provide a picture of the real life situation that we are going to study, the completely
describe the different aspects of the real life situation and hence the name descriptive
models.
Predicted models they have the property or their primary objective is to predict the
behavior or the performance of the system under a given conditions given set of
conditions, ok. And normative models, they recommend a course of action and that
action is supposed to be optimal given the objective for a given situation; given a
situation, given the objective normative models will recommend the optimal course of
action.
So, descriptive models, they provide a picture of the situation being studied. Predictive
models predict the behavior or performance of the system under consideration. And
normative models recommend optimal course of action given the situation with clearly
specified objective.

(Refer Slide Time: 09:31)

Now, we classify models according to the structure. When a model retained some of the
physical characteristics of the system that is that we are going to study it is called an
iconic model. So, iconic model retains some of the physical characteristics of the system
that we are going to study.
An example for this type of model is a scaled version of an air plane. If some features of
the real object are represented in the model by some other features, then that type of
model is called an analog model. And symbolic models consists of expressions which
involve symbols to denote the various entities of the system. These are very much
important in the field of management, industrial dynamics problems. They basically use
symbolic models.

(Refer Slide Time: 11:01)

Then, we classify models according to time reference. Static models, they do not change
overtime, they do not refer to time. In dynamic models, time is explicitly incorporated as
an independent variable. An example of a dynamic model is a differential equation that
specifies the rate of change of sales volume as a function of current sales and advertising
expenditure.
So, according to time reference we have two kinds of models static models which do not
refer to time, dynamic models they refer to time and we have given examples of dynamic
models.

(Refer Slide Time: 12:13)

Next classification of models according to the degree of uncertainty. In deterministic
models all the variables are known and they are deterministic in nature, whereas in
probabilistic models some or all of the variables are random in nature.
Next classification comes when we talk about the degree of generality that is there in the
model. General models like say linear programming, they have application in several
functional areas of an organization, whereas special models can be used only for solving
specific problems.
(Refer Slide Time: 13:32)

We have already said that models are used in DSS, mainly to solve problems where we
need to forecast or project the trends of a given activity by finding patterns in historical
data. When we will talk about forecasting DSS, this kind of models will be highly
relevant and used.
And also models are used to answer what if questions that is I already said sensitivity
analysis type of problems, such as what happens to next year sales if the price is
increased by 10 percent. And these kind of models are highly used in every organization
through the use of spreadsheets. Spreadsheet models they are very effective in solving
sensitivity analysis type of problems.
(Refer Slide Time: 14:49)

Models are used in DSS to analyze a proposed solution in terms of its feasibility and its
effectiveness to suggest several alternative solutions for a given problem and to
recommend a best course of action among the various alternatives.
We have already mentioned about all of these points. Again I am repeating because one
must be very-very thorough about the usefulness of the model; how to construct the
model, the properties of the model; because next to managers, models are the most
important components of this effective decision support system.

(Refer Slide Time: 15:45)

Several problems encountered in practice tend to have too many alternative feasible
solutions. And it will be impossible in most cases to generate and examine each
alternative. Therefore, techniques have been developed which generate only a small
subset of the numerous alternatives such that the best solution is contained in the
solutions generated. And these kind of solution techniques are basically from
optimization, heuristics and simulation.
(Refer Slide Time: 16:58)

Simplex method for linear solving linear programming type of problem is one such
technique which generates optimal solutions. So is branch and bound method for solving
integer solution, integer programming problems. And these techniques they generate the
best solution on a given constraint and they are basically these techniques are known as
optimization techniques. Some problems in real life are very complex.
(Refer Slide Time: 17:44)

And even the optimization procedures, in spite of their tendency to examine only a
subset of the feasible solutions prove too costly and time consuming for solving these
kind of complex problems. In such cases, the managers they satisfy themselves by
generating a feasible solution which is effective and hopefully very close to the optimum
solution by deploying some rules of thumb and these rules of thumb are called heuristics.
So, when the problem in hand is very complex, then we have observed that optimization
techniques, will take lot of time and it is also very costly to solve those kind of problems
using those optimization techniques, so in industries managers they are more happy to
use these kinds of heuristics. That is, they imply certain rules of thumb to generate a
feasible solution which is an effective solution and hopefully that solution may be very
close to an optimal solution. Even if it is not an optimal solution, the managers they are
more satisfied with the effectiveness of that solution and that is what is most important in
any decision support systems.

(Refer Slide Time: 20:27)

So, by deploying heuristics the managers they save time and effort. They may likely miss
the best solution, but the emphasis here is to generate an effective solution. For example,
job shop sequencing or even scheduling is a very complex problem. If we try to find an
optimal solution we will be left nowhere. So, in production shop such kind of job shop
sequencing problems are often solved using heuristics.
In the job shop sequencing problem what is the problem? There is there will be a set of
machines and set of jobs and each job has to pass through a different route through the
machines, which job should be loaded and which job should follow which route, this is
the problem.

(Refer Slide Time: 21:57)

The machine sequence for each job and the processing times for those jobs in the
respective machines are already known. The problem is basically to develop a schedule
for those jobs for each machine such that some criteria like say minimizing the in process
inventory or minimizing the total lateness of jobs is met. Different heuristics will result
in different performance of the system. And depending on what the managers want the
rules of thumb will be deployed.
(Refer Slide Time: 22:59)

For example, whenever a machine becomes free, examine the jobs that are lined up at
that machine and load the job which has the shortest processing time on the machine.
This is also known as shortest imminent processing time rule. This is one kind of
heuristic.
We can define different heuristics by varying the rule for the selection of the job to be
processed next. For example, we can choose a job with the earliest due date or with the
smallest number of remaining operations. So, there are different kinds of heuristics and
depending on what we want to achieve those kind of heuristics are used in job shop
sequencing problem. Heuristics are very-very popular in solving job shop sequencing
problems and also in scheduling.
(Refer Slide Time: 24:19)

Next, we talk about simulation. Simulation is another technique, often employed to deal
with systems that are complex, so complex that if we need to find an optimal solution we
will be left nowhere. Here the system is described by a set of logical and mathematical
relationship and such a description is translated into a computer program. Then a set of
pre-formulated policies are given as inputs to the program, and output of the program
gives the performance of the system under the different policies. The policies are
embedded as part of the computer program.

(Refer Slide Time: 25:36)

Simulation have been used in a variety of situations. Examples are queuing systems; for
example, computer networks, railways. Examples include ecological and economic
systems, and corporate investment analysis.
The beauty of simulation techniques is that simulation models can deal with complexity
more easily and in a less costly manner compared to use of optimization techniques. And
hence, heuristics and simulation they are very popular in designing decision support
systems. Of course, optimization techniques have also been used in various DSS. It
depends on the situation; depends on the complexity of the problem; depends on the
affordability; and things like that.