Lecture 29

IIT Kanpur July 201
20 followers
1 Video View·Feb 23, 2024
uh welcome back my dear students dear
friends and participants for this course
titled multicriteria decision making
under npil Muk series and my good name
is raghunandan s Gupta from the ime
department at aan
so as you know that we are in the set of
classes for the sixth week and um this
is a 12 week course for we spread over
60 lectures and each week we have five
classes each class being for half an
hour each so the broader overall
context based on which we have been
discussing on the slides and the
lectures are multicriterial decision
making multi-objective decision
making uh multi criteria decision making
is the broader umbrella under that you
have the multiobjective decision
making and then the multi-attribute
decision making there is a difference
which I keep repeating multiobjective is
more
quantitative continuous easy to
understand and
very easy to communicate multiattribute
is is more subjective qualitative in the
discrete
space and not that easy to
communicate solution techniques would
take Q from the concept of
multiobjective decision making and some
concepts of different qualitative
methodologies of
ranking and further on as we have
discussed about
um univariate utility Theory we will
consider the concept of
multi-attribute utility Theory also and
this is the 29th lecture as I mentioned
coverage would be par optimality we'll
again visit that with examples why I'll
come to that
later the concept of parro optimality in
more details with example and then going
into the concept of goal
programming now consider the simple
example it can have different flavors so
so let us consider this simple example
you have uh I'll just mark
it you
have a
three-dimensional problem because we
will be able to analyze it much
better and the coordinates are discrete
for
Simplicity they are given as X1 X2
X3 X1 can take only values discrete 0 1
2 3 X2 can only take discrete values 0 1
2 3 and similarly
X3 can take discrete values 0 1 2 3 the
functions to be analyzed are they are
not we need will need to be either
maximize or minimize I'll come to that
later but first let me say State the
function the functions are X1 2 + X2
into X3 so this it is technically
quadratic because
it is squared here and obviously X2 X3
is also concept of square for the
variables similarly the second function
FS2 is X2 2 + X1 into X3 and
finally the function 3 is f X3 2 + X1
into
X2 what are the constraint or what is
the constraint to make our life simple
for analysis I'm only going going to
take one constraint it can be expanded
also the constraints are X1 + X2 + X3 is
equal to 3 exactly 3 so obviously if I
consider these values few of the values
which would not be allowed and that's
very obvious if I take a value I'll few
examples which are not applicable so
consider X1 X2
X3 if I take X1 value of zero which is
applicable
X2 a value of say for example 1 and X3 a
value of three is that allowed no
because the sum is four not three so
this is not allowed I put a minus sign
this is a
minus not
allowed or let better would
be if I
put a
no consider the next value this this is
2 this is 0 and this is 2 X1 X2 X3 are
two 02 Allowed no because sum is
four consider the values of 1 1 1
allowed yes because sum is
three sum is exactly
three value of 1 0 1 one Allowed no sum
is
two consider the value
of 0
03 allowed yes
so we have different combinations which
are allowed different combination which
are not allowed and if you see I should
make this heading of X1 X2 X3 not in red
color red is no and green is yes so I'll
use the black color here
here this is the situation red is
no green is yes and we
see it should
basically follow the
norm number one and number two obviously
all these values X1 X2 3 are dis ret and
they are in the domain space of 0 1 2 3
discrete as I
said based on that I Mark all the
relevant
values so these are the relevant
values
so the value of X1 is 0 X2 is 0 X3 is 0
and X3 is three allowed yes because sum
is three X1 0 X2 1 X3 2 allow 2 + 1 is 3
third I'm not writing anything because I
have to plot it which I'll come later so
I'm not don't want to clutter it X1 0 X2
2 X3 1 allow 2 1 + 1 3 so I'll read the
values of X1 X2 X3 correspondingly 03 0
allowed one2 allowed 11 1 allowed loud
means yes 1 2 0 yes 2 0 1 yes 2 1 0 yes
3 0 0 yes all the sums are are exactly
equal to three and the values of X1 X2
X3 are either zero or 1 or 2 or three
and you can see it from column one so
there's no values which is not in that
set of 0 1 2 3 0 is is there one is
there two is there three is there for X2
similarly 0 1 2 3 is there and similarly
for X3 0 1 2 3 is
there now what are the functions if you
go back the functional value of
F1 is given by this I'll use a color C
exam valet so this
is X1 s + X2 into X3 so let us go
back
X1 S 0 X2 into X3 is 0 so 0 + 0 is
0 then 0 squ into 1 into 2 so sum them
up two which is written there in the
fourth column 0 2 2 into 1 2 so 0 S + 2
into 1 is 2 there 0 2 + 3 into 0 is 0 1
2 + 0 into 2 is 1 1 1 2 + 1 into 1 is 2
1 2 + 2 into 0 is 1 2 2 + 2 into 1 is 4
2 2 + 1 into 0 is 4 3 2 + 0 into 0 is 9
so fourth column which is F1 is taken
care
of I go to the second value which I will
highlight
using the light blue color
which is X2 s + X1 into
X3 so X2 s X2 is the middle value here
which is the second column so if I
consider 0 S + 3 into 0 is 0 1 2 + 2
into 0 is 1 2 2 + 1 0 into 1 is 4 3 S +
0 into 0 is 9 0 2 + 2 into 1 is 2 1 2 +
+ 1 into 1 is 2 2 2 + 1 into 0 is 4 0 2
+ 2 into 1 is
2 1 square into 2 into 0 is 1 0 square
into 3 into 0 is
0 which is F2
done now come to
F3 I'll use
this a little bit darkish orange not
orange darkish yellow which is X2 X3 squ
+ X1 into X2 so we will concentrate on
F3 value it is not colored yet but I'll
come to the reason why I have not
colored it I'll come to
that so 3 2 + 0 into 0 is 9 2 2 + 1 into
0 is 4 1 2 + 2 into 0 is 1 0 2 into 3
into 0 is 0 2 2 into 1 into 0 is 4 1 2
into 1 into 1 is 2 0 2 + 2 into 1 is 2 1
2 into 2 into 0 is 1 1 0 S into 2 into 1
is = 2 0 S into 3 into 0 is
0 and
finally the last column which was not
mentioned in the earlier slide I have
taken the
sums why I'll come to that later so if I
take the sums so it is 0 + 0 + 9 is 9
then 2 1 4 7 2 4 1 I'm adding them up 7
09 09 1 2 4 7 2 2 2 6 1 4 27 4 2 1 7 41
27 9
09 now let us
concentrate on
the
columns Mark fub1 and
F2 which I have marked it here and what
I mean by par
optimity using this coloring scheme if
you see the coloring schemes are
yellow dark
yellow and blue and finally green so I
will
create and a cartisian this is two
Dimension create a cartisian coordinate
in order to explain
that so I'll uh add a blank file and
then discuss that so this is the blank
slide I have created
so let us consider only the Yellow Part
2112 and there this you are measuring F1
and FS2 so if I
consider I'll measure FS2
here fub1
here and this is so this is
0.12 this is this is 1 2 so mark it 1 2
1 2 now let us see the yellow Mark which
I said it's
2112 so
21 I should
you
21 is
here and one two is
here so what does this point set of
points mean if I
join a
line so these two
points
are on the same
level considering the net what if I
consider the sum of fub1 and FS2 they
are fixed but the individual values of
fub1 or FS2 have been changed so
F1 if I consider for point
a and for point
B F1 has changed from 2 to 1 and F2 has
changed from 1 to two sum remains three
but the initial points have changed so
if I consider this dotted line it is aen
to an Optimum paral Frontier or paral
line so as the value of the sum of fub1
+ F2 remain same individually they can
change now let us consider the next set
of points yellow we have already done so
let us go to the set of points which are
given by Green 14
41
14 4
One if I
join so so the set of points is 4
1
4 again the sum of the functions F1 and
F2 is 4 + 1 or 1 + 4 is 5 even
individually they are at
different values individually but the
collective value is
same again this Violet line is a Paro so
called line where all the set of points
have different coordinates but the sum
of F1 1 and F2 remain same next let us
come
to the setup points blue which is 2 4
42 2
4
42 so again if I
join this point is
42 this point is
24 again some is
six individually they are
different
finally I have the point as 9 0 and 09
which is in the dark yellow 90
09 I use the
color 9 would be somewhere here
consider and 09 here so again if I draw
it's out of the screen but I'll draw
again it's a straight
line what is and again this is Paro line
sum is nine individually are
different what is important to note is
I'll use the blue color if I
consider all these lines are parallel to
each
other they don't intersect the sum of
the value for each of these lines red
violet green and dark yellow is
increasing as I go up onto the right top
Corner that means the net value if you
are on along any of the line is fixed
the net value of the sum of F1 and FS2
increases so they are just like the
indifference curve which are basically
parallel to each other never
intersect and for the case when you want
to optimize maximize maximize both of
them obviously and obviously if you
remember those constraints were brought
into the picture in order to give a much
more practical feeling so where the
constraint was X1 + X2 +
X3 was three equal to three so these
four sets of the points are the par to
Frontier such that a a decision maker
based on the fact he can CH choose any
one of the points X1 X2 X3 such that
each R discrete 0 1 2 3 would obviously
lie on any one of these four lines which
are shown in red violet green and dark
yellow anything out is not possible all
of them are
parallel the overall net value along
each individual line is fixed and
difference Es are there as you move up
onto the right top Corner the values of
the sum
increases now I consider so we have
considered F1 and FS2 now let us
consider FS2 and F3 again the same
coloring scheme exactly the
same so here if you see the yellow part
I do the same analysis draw the same
curve the yellow part is as shown here
for the values of FS2 and F3
respectively as 2 1 1 2 sum is 2 + 1 or
1 + 2 is three fixed individually they
are different so if I draw them on the
cartisian coordinate they would be a
straight line and each individual point
along the straight line gives the same
sum but the individual values of FS1 and
FS2 are
different similarly when I move on to
the next set of points which is the
green one which is
1441 1441 are basically I'm talking
about F2 and F3 respectively sum is five
1 + 4 is 5 4 + 1 is
5 and individually they are different
coordinates but the sum remains same and
also if I compare with the yellow part
which the sum is three this is five that
means I moved from a value from 3 to
five corresponding to the concept that
the constraint is already
applicable when I come to the blue set
of cells which is highlighted here in
the
slide the fub1 and F3 FS2 and F3 values
are 2 4 4 6 individually they're
different but the sum is six so when I
join these points 2 4 42 as I have done
in the previous figure the sum is six
they are the all the set of points along
these um the straight line they give the
same sum of FS2 plus fs3 individually
are different and also yellow with sum
was three green sum was five blue the
sum is six which means I have moved
parall one step from 3 to five then 5 to
six six means 3 5 6 are the sum of f2+
F3 finally when I go to the dark yellow
part which is basically equ
09 and 9 0 which are the respective
values of F2 and
F3 the sum is 9 for 0 + 9 is 9 9 + 0 is
9 individual values are different so if
I join this coordinate 09 9 0 is a
straight line all the set of points
along the straight line gives you the
same sum evenually are different and
also the sum of 9 is more than the sum
of six that is 2 2 + 4 is 6 and 4 + 2 is
6 that means doesn't have been a
parallel shap so the we of trying to
draw FS2 and F3 would be exactly equal
to this only change would
be this would be
F2 and this would be
F3
then let us come to I'm going a repetion
but I'm going slow when I consider F1
and
F3 because why I'm taking two
dimensional because easy for us to
understand again the same coloring
scheme yellow green blue orange yellow
being for the values of 2 1 1 2 2 1
being for fub1 F3 1 2 being for for F1
F3 so I'll first say the F1 value then
the F3 value the sum is 1 2 + 1 1 + 2 is
same three individually are different
when you draw this join this two points
is a straight line straight line like
this the red part here
2112 but in this case let me change it
beforehand this will be F3
find fub1
when I come to the green
highlights the respective values of F1
F3 is 1 4 4 1 sum is five individually
are different when I join these points
they give the same satisfaction level of
sum of FS2 plus FS1 plus F3 individually
as I said different so the sets of
points are the Violet one as you see 41
14 when I come to the blue cells 2 4 42
for F1 and
F3 the sum is six individually are the
different when I join these two points
again in a straight line all the set of
points along the straight line gives me
the same value of FS1 +
F3 and when I draw
it this is the green line
4224 finally when I take the orange uh
this dark yellow cells values of fub1 F3
09 9 0 sum is 9 that is 0 + 9 is 9 9 + 0
is 9 individually the values of F1 F3
are different because one is one
coordinate is 09 another is
90 the sum of points joining on line
joining them have give me the same
utility or same value of fub1 +
F3 and if I draw it this is the dark
yellow and again all the lines for
example one example two example three
are
parallel the values along any line is
fixed and the values increase as I go
onto the right top Corner this is a
increasing increasing case
now I
consider the example as given here and
I'll highlight it
now when I consider the three cases
collectively fub1 FS2 fs3 and the right
column is basically the sum of fub1 F2
F3 and if you see the values are given
09 is sum is 9 21
47 2 41 7 09 09 1 2
47 22 2 6 1 427 4 217 4127 9 9
9 now consider these
movements and consider is
a minimization I want to do individually
for F1 F2
F3 and check whether the if they are
individually being minimized so
technically I would think that
individual sum should also be minimized
whether it's possible or not let us
consider now concentrate on this
dark yellow set of
cells consider F3 is to be minimized so
I I should go from two I should jump to
a lower value one so let us do that
which is shown in the dotted blue part
so it is decrease I'm very
happy but see here two
things I have decreased for F3 fine
individually but the sum which was six
initially has now increased is to 7
which is not done because collectively
I'm basically compromising why this
compromise is there because if I move
from 2 to 1 for
F3 when I move for only consider FS1 it
remains as two so this dashed Circle
part this one for F1 remains in the
orange cell and the green cell same two
which is fine I'm saying and the same
level for
F1 F3 has decreased from 2 to 1 which
I'm happy but
interestingly the value for FS2 which
was initially two has now increased to
four so which is increasing if I want to
op minimize so this should not have been
the case so in this case the value of 2
goes to 4 for F2 hence the total sum
rather than me six has increased to to 7
so which is not a way of trying to
handle the concept of the parit
optimality and subsequent problem
because this is a
compromise if I
consider the idea of C for example
trying to
maximize so if I consider the concept of
maximization in initially consider the
minimization and
maximization consider on this concept of
F1 F1 is increase 2 to 4 fantastic I'm
happy so I'm considering this dark
yellow and the bottom green
part F2 remains same fine no
problem but
interestingly it has been maximized six
has gone to
seven it is a maximization case but the
idea of trying to increase F3 is not
happening because it has decreased from
2 to one collectively fine I have
increased but individually I have
compromised in the second example in the
first example individually one is
benefiting collectively there's a
disbenefit so these two examples were
collectively disbenefit for movement
collectively benefit for the movement
but individually they are being
compromised it will be considered
further on in further examples this
example has not finished I'll consider
more more such um different combinations
of the example later on and with this
I'll I'll end this class have a nice day
and thank you very
much
