Lecture 31

IIT Kanpur July 201
20 followers
1 Video View·Mar 1, 2024
[Music]
uh welcome back my dear students dear
participants uh a very good morning a
good afternoon and good evening to all
the participants and students who are
watching this lecture the title of of
this course is multicriteria decision
making and my good name is ragun Nandan
s Gupta from the IMA department at I
kpur so this is the 31st lecture out of
the total set of lectures which is 60
that is the whole course is spread over
12 weeks and each week we have five
lectures each lecture being for half an
hour the broader set of Concepts which
you are still considering is multi
criteria decision making multiobjective
decision making multi-attribute decision
making where we have been discussing
time and again the difference between
multiobjective and multiattribute
concept and then further cons consider
multi-attribute utility Theory its
implications how they can be
utilized the coverage for the 31st
lecture and uh
if we spill over it will be in the 32nd
lecture also which we'll discuss would
be further discussion about gold
programming very simply the concept of
parit optimality in 2D space for linear
programming and concept of parit opun in
3D space which solve
problems for a quadratic programming
again as I said that uh I assume and all
of you have a good background in
mathematics I statistics and operation
research so the ideas of how to program
quadratic programming linear programming
zer1 programming then integer
programming are you are all aware of
that so the concept of very simply of
trying to use uh those algorithms in the
perview or in the concept of
multiobjective programming is being
discussed and when I say
multiobjective I only I'm sticking to
the biob one because trying to convey
the information how the perative curves
will look like is much EAS easier in the
two- dimensional case and obviously it
can be extended for the higher Dimension
also like three dimension four
dimension
considering you have more than three or
more than three decision variables and
also the objective functions are also
more than than two so as we were
discussing this is the slide which we
have discussed and based on which we
ended the 30th
lecture the idea
was you
have the maximization concept
here and the minimization also is being
discussed so that means there are two
types of fun functions can be different
but there are two Notions of what you
want to do with these two
different functions which are there when
when I use the word functions it means
that when you see fi I can be more than
one when you che check GJ G can be more
than
one in one case I collect all the
functions FIS and I want to maximize
them and another case similarly
simultaneously I want to collect all the
functions and G and try to minimize them
now the question may immediately come up
which I did touch upon in earlier class
or earlier lecture but I'll also
highlight that this combinations which
you
see I'll write it
accordingly one case you can maximize
when I I'm writing the word maximization
is basically related to the first set of
functions consider there are only two
functions F1 I is one
g 1 J is 1 so let us make things Sim
simple to analyze that and then we can
extend
it so when I write Max it means
basically for the first function when I
is equal to
1 when I want to minimize I'm using the
green color and I'm writing it in the
next place which is in the second column
which is related to J equal to
2 then the other concept you can be I
want to
minimize the first
function so minimize F first function
means I equal to
1 and simultaneously
maxximize the second function so I'll
write it
as um let let me follow the same
procedure which we are using so it is
there's no confusion with what I display
and what I talk and what is written on
the
slides so when we are trying to
maximize so is maximization of F1 I'm
not writing x x can be in
multi-dimensional case also but it's a
function of X which is the decision
variable
and I want to minimize G
D1 similarly in the second case I want
to
minimize
the
function
fub1 I want to
maximize
G1 the other case can be I want to
maximize fub1
I want to
maximize
J1 and the last combination is I want to
minimize
fub1 I want to
minimize G1 so these four combinations
and if you
remember that in the space where there
is object function one and object
function two which you have drawn so
I'll mark objective function
o1 o1 can be considered is F1 so mark it
as
fub1 and it's on the
x-axis it is basically O2 which is F2 on
the Y
AIS and if you remember we had four
different combinations coloring scheme
I'll write
down
so they were like this the coloring
scheme what I'm writing down and I'll
highlight it accordingly which is
written there in Max F1 mean G1 minan F1
Max G1 Max F1 Max G1 minan F1 mean
G1
so so this was
one the other one
is
I'm using different colors so in order
to avoid
confusion so it should be run
properly so this is
one and the final one I use the blue
color so the light blue color that was a
little
bit dark
blue so I'll mark it
as
a so the curve as
a is
B uh
C so when I'm trying to
consider these four combinations a B CD
so let us see what is the overall
idea one by
one I won't erase this graph which I had
drawn A B C D and I won't erase the
equations
also so consider the first
one a
part so in the a part when it is
there
technically
the issue would be that we would try
to and this is object objetive
function 2 which is FS2 this is
fub1 and
technically
for
F2 what we actually would be doing is
that we would try to re is the corner
Point here so Corner point on the left
top Corner what does it mean it means
that I'm trying to increase FS2 go up
and decrease F1 go
left so if I'm considering fub1 and FS2
so fub1 is being minimized so
minimization color I'm have been using
is basically
green so I'll move in this direction
and the maximization color I've been
using is red which is in this
direction so maximization has been there
for FS2 minimization there is for fub1
maximization for f FS2 means basically G
so also let me re remove that confusion
also I used o1 O2 because it was based
on the diagrams which we had
done so I'll
write so when I'm using this so this is
minimization of F1 and maximization of
of G1 minimization of fub1 and
maximization G1 basically this
diagram would correspond
to this point I'm using the same color
here as I'm using
here so I'm trying to minimize F1 go to
the left maximize G1 go
top so this is the two Dimension case so
obviously there's three dimension means
first function f second function G and
third function h now don't confuse the
functional form of the objective
function H based on what is written on
the constraints constraints I've already
discussed here it was basically written
as two objectives with I and G can be
more than one also so H what I just
mentioned is basically if there the
objective function was denoted by H so
it the three dimension
case when
I
use the B1 b means B which is drawn here
so this is like
this you're moving the bottom left
corner going
down which
means fub1 is
being
minimized G1 is being
minimized so if I consider both of
them this is this
part both are being
minimized now when I come to
C again I have deleted it but H can also
be obtained or added as a third
objective function and the diagrams
would be considered accordingly so but
but only you have to remember that H can
also be maximized and minimized so for
two functions f and g you have four
combinations Max Min Min Max Max Max Min
Min then when you bring H obviously age
having two combination it means that now
I will have basically six
combinations so one extreme being all
the maximization happening for f g h and
the Other Extreme being all minimization
being happening for
fgh and so there are
technically accordingly uh two for these
two for these and two for these so it
will be 2 into 4 4 2 4 to 8 sorry not
six eight eight combinations would be
resulting from
this now when I come to
C which is
this so I'm going to the bottom right
condition so
here
G1
is being
decreased and F1 is being
increased so if we consider
this F1
being um increased and G1 being
decreased so it will
be this
combination
and obviously H can be also considered
as a third
function so if we remove this I'll
remove this color also so to in order to
Mark so I just rot Max mean for the
color
combination when I finally come
to the D
curve which is
here I increase both of
them which
means increase of fub1 increase of
FS2 so this would be become become the
third
part so
correspondingly I'll
Mark which I already done but just
for a is for
this B
part is for
this C
part is for
this and D
part use for this using this we
analyze
now considering that we have already
discussed I'll erase this and come to
the constraints constraints I already
discussed but I'll still highlight
them so the constraints are so there are
one is greater than type as you
see the one is equal to sign
and the third one is less than sign I'm
using different
colors and the function is H1 H2
H3 and the subscripts K1 K2 K3 means the
total number of such constraints are
capital K in
number where capital K is equal to small
K1 or Capital
K1 which is for greater than type plus
means Union
here K2 me for the equality type again
plus means Union and finally K3 being
for the L than typ and X
is of in the real line of n
Dimension now what I have explained is
is written here so it will be easy for
us to follow where we want to maximize
fub1 I equal to 1 to Capital I while on
the other hand we want to minimize GJ G
equal to 1 to capit J so here for
Simplicity I considered as I mentioned I
and J as
1 subject to capital K number of
constraint which I just mentioned few
seconds back and they
are again I'm repeating greater than
equal to less than and on the right hand
side greater than equal to less than
have been given by the constants they
are deterministic they are not
stochastic that means
constants it is B suffix K1 for equality
sign is B suffix
K2 and the last one when it is less than
sign it is B suffix
K3 here so here the suffix K1 K2 K3
which are written in the the start
bullet point which I'm putting a tick
mark so small K1 goes to 1 to Capital K1
small K2 goes from 1 to Capital K2 small
K3 goes from 1 to Capital K3 and here
what I wrot the plus is basically they
add up to the total number of
constraints and obviously we consider
the intersection of the greater than
equal to less than is a Nite because
they are distinct and we can for
Simplicity we can differentiate them as
constraints of the greater than type
less than type equal to type here x are
the structure decision
variables fi and here GJ are the set of
objective
functions and these constraints are
basic the goals you cannot exceed them
cannot violate them so H1 H2 H3 where
the number of such constraints for H1 is
capital K1 K1 in number for H2 is
capital K2 in number for H3 is capital
K3 in number bk1 bk2 bk3 are the target
values or aspiration levels associated
with the respective
goals
collectively these constraints greater
than plus equal to and plus the less
than type are called the constraints
total constraints so we will consider a
problem and expand it accordingly so
this this is a linear programming
problem suppose one rich farmer in India
wishes to grow the following four crops
which which are accordingly rice weed
Basra maze there are four
crops the selling price per kg of this
rice wheat bajra and Maze are
respectively
28
25 30 and 24 rupees per
kg the cost price per acre for these
four crops which are crops are again I'm
repeating ice wheat basa and Maze the
cost price like you have to use pump set
to draw water you have to pay the person
who is stealing the land use a tractor
or a bulock bulock whatever way you are
trying to uh till the land you have to
use seeds maybe you have to use
organic uh ways of trying to
um do the farming or somebody use
fertilizer whatever it is the c burning
cost there would be many costs the cost
per acre for these four crops are
respectively
33,200 for rice 30,500 for wheat
35750 for bazra and Maze is 27,000
respectively corresponding with the
average output per per acre so there are
different outputs different cost also
different selling points also throughout
India but I'm taking a Aver very average
value
output per acreage are for this four
crops rice wheat B me are
1,250 kgs
1,150 kgs for the second one 1,200 kgs
for the third one and 1,100 KS for the
fourth one the total irrigated land
available with the farmer is about 1,000
acres and for each crop at least 100
acres and at most 400 Acres of the total
irrigated land should have to be used
which means for rice or wheat or Basra
on maze you have to use 100 acres and
cannot exceed 400 acres per uh the crops
you have furthermore due to water usage
restriction total AC of rice cannot
exceed that because rice needs a huge
amount of water rice cannot exceed that
of the combined acreage of Wheat and
basda taken
together and also as wheat consider the
person is irrigating in the part of
India where wheat is the staple
diet so considering wheat is a stable
diet hence it production should be more
than 50% of the combined production of
rice bajra and Maze because the person
has to supply to the
market finally for the movement of the D
caor vehicles and other things you have
say for example the person wants to lead
out on fencing there are different type
of animals like which come and
destroy the
irrigated land eat up these crops so the
person has to put fence and has to use
the pathwave to take the tractors or the
Bullocks whatever consider that part of
the land is not utilized
for
irrigation so here it mentions finally
for movement the tractor and the vehicle
Etc to be used after crop
Harvest 2% of the total irrigated land
cannot be
utilized so let us see the problem
accordingly as evident the decision
variables for these problems are related
to the acreage to be given
for Rice wheat basa and Maze which we
consider as decision variables X1 X2 X3
X4 now there are Revenue cost based on
selling and there are cost also for
irrigation for tilling the land and for
production so if I consider the
corresponding the revenues so if I
consider the two corresponding revenues
it is the output per acreage for these
four crops which I'll
mark are 1250 1 150 120 0 1 1 0 0 for
wheat
brra we Bry we bajra Maze and if X1 X2
X3 X4 are aage for this four cross
respectively and if I consider the
selling price given as
2530 28 sorry 28 25 30 and 24 so
obviously you have to multiply the
corresponding factors to find out the
total revenues which we are
doing which is 28 for the first one
which is
here 25 for the second I'm again
repeating it please don't mind 30 and 24
so when I multiply is 28 into 1250 into
X1 which is the total eage for the first
drop price 25 into 11
1,550 into X2 for the second 30 into
1200 for the 3D into X3 and 24 into
1,11 100 into X4 for the 4th so this FX
gives you the total
revenues now when I come to
the cost function so if you note down
the revenues are marked here in blue
cost are also blue so I marked the
revenues
so now as per the concept given
here the total cost price for theage for
these four uh crops are given
as 3,00 33,200 30,500
35750 257,000
and per acreage so acreage for these
crops are X1 X2 X3 X4 so when I find out
so so this is 35,000 X1 +
28,750 into X2 + 36,000 into X3 + 26,400
into X4 so this is
the cost function and so this is if I
consider G so now if I consider so this
was sorry sorry sorry sorry this
was I just missed it so my mistake
so this was
basically the cost function given
accordingly my apologies so when I
multiply them for
the revenues they were coming as
35,000 X1 + 28,750 plus into X2 + 36,000
to X3 + 26,400 for X4 and the
corresponding um
equation for the cost R which I have not
read I will just mention it 33,200 X1 +
30,500 X2 +
35750 X3 and 27,0
X4 so there would be analysis for them
problem which we will do in the
subsequent class I'll just show this and
then continue to the discussion for this
lecture thank you very much and have a
nice
[Music]
[Music]
[Music]
day
