Lecture 35

IIT Kanpur July 201
20 followers
Mar 1, 2024
[Music]
uh very good morning good afternoon good
evening to all my dear participants and
students for this
course and this the course title is
multicriteria decision making under nptl
M series
and my good name is raghunandan s Gupta
from the IM Department
atanur in
India so as you are aware that we are in
the 35th lecture out of the total 60
lecture for this course which means it's
spread over 12 weeks each week we have
five lectures and each lecture is for
half an
hour so the broader set of points we are
or the IDE ideas or the discussion
we are going through and which we had
for a long
time are as follows multicriterial
decision making we have discussed many
relevant points we are still in the
domain of multiobjective decision making
trying to formulate different problems
to understand how the maximization
minimization can be done for linear
programming programming or quadratic
programming
also and then we did discuss the two
very powerful Tools in multi
criteria where you give weights as
Lambda depending on the I as the number
of objective
function or also we considered as we
were considering in the last day 34th
lecture we'll continue doing so is how
to operate and try to optimize one of
objective function uh with the existing
constraint plus the other objective
functions being transferred into the
constraints part so whether it's a
maximization problem minimization
problem for any of the objective
function the idea Remains the
Same and obviously the other points
which we have discussed very briefly is
the utility function will go into the
multi-attribute utility Theory very
briefly about a little bit more
theoretical and then take the important
points from multi attribute utility
Theory and try to understand the
different concepts of other
multicriteria decision making uh which
we had mentioned long time back when we
were discussing the
syllabus those techniques being Electra
method topsis
method then the VOR method and all the
things
the examples which we'll consider in
this 34th lecture is the continuation of
the linear
programming and the concept in mcdm and
the problems which are there for the
linear programming can easily be
converted into the integer programming
there should not be any problem provided
the objective function and the
constraints are linear obviously if they
are nonlinear also we can consider that
the second
coverage hopefully we can start or else
it will be going into the the next
lecture is the idea of NLP where we'll
consider examples in the domain of
multicriteria decision making so the
problem which we are considering and I
did mention quite a few times that this
is a three-dimensional problem X1 X2 X3
why we are considering a three dimension
because it will be easy for us to
visualize even two Dimension is also
possible and all of them are
maximization fub1 F2 F3 and I told time
and again that fub1 would be considering
in red color F2 in violet color F3 in
blue
color and the constraints are all the
same for the case when we are trying to
solve them
individually so considering that we want
to solve them
individually the constraints are given
here so there are 1
2
3
4 five constraints the first objective
function was denoted by fub1 with the
red
color FS2 with the violet
color and F3 as the blue
color and we also Al
considered that for single objective F1
star would be the max value for
F1 F2 star would be the max value for
F2 and F3 star would be the max value
for F3 and for all these single
objective cases when we get fub1 star
FS2 star F3 star the constraints are the
same
as shown here these are the
constraints they will be applicable for
case one case 2 case
three without going to much details
the first case which is objective
Function One F1
maximization the points are 20 0 F1 star
is
40 we have discussed that for the second
case F2 star being found out the
decision variables are 020 F2 star is
40 third
case the decision variables or X values
vectors are 0 0 20 and F3 star is
40 so it's 20 0 0 0 20 0 0 0 20 and F1
star F2 star F3 star all the values are
40 each now we bring as we are
discussing we I'm repeating the problem
but please bear with me we want to
maximize
F1 this is
fub1 the constraints I'll just put tick
mark these sets this four
here and the fifth one is here and the
two object functions have been brought
into the constraints
one is this the one which I'm marking in
Violet which is related to the second
objective
function and another is we importing the
tick mark blue which is the third
objective function they are less than
because we are maximizing so obviously
it cannot be more than a maximum value
it has to be less and we'll tackle the
problem
accordingly the problem solution
depending which I told Delta 2 Delta 3
are the values which are to be
subtracted from F2 star and F3 star in
order to have a feasible
solution and Delta 2 Delta 3 are giving
given in column 1 where I put a violet
tick
mark in column two where I put a blue
tick mark the
corresponding X Vector which is X1 X2 X3
are
given
in the third
column and the corresponding values
which we get for F1 double star double
star is basically the other F1 star was
the max with a single objective so now
with the two other object is being in
the constant how does fub1 vary that we
are denoting by F1 double
star that is given in
the last column where I put two tick
marks and the values are there one
interesting thing is that I've just
noted down the values as they change
so a darkish
yellow then the value of 39 there are
two sets of 39 which I get depending on
Delta 2 Delta 3 so Delta 2 Delta 3
rather than that being important what is
important are what are the values of
decision variables X1 X2 X3
they are same for 39 if you
see they are same 191 0 1910 similarly
the green one is related to The
Objective function F1 star being
37.5 and the values are I'm not marking
it here but you can see is 17.5
2.50 then we consider 36.5 which is
there the values are 16.5 3.5 0o and for
a value of 35 when I to talk about the
value 35 this F1 star double star the
values of X1 X2 X3 are 15 5 0 now I
again I go
to the second objetive function which is
FS2
maximize and all the constraints remain
same this four and this fifth one so
1
2
3 4 and
five and the two object
function pushed into the constraints are
corresponding to the blue one which is
objective
function
two and the red
one is objective function
one again less than type like fub1 Star
minus Del Delta 1 and F3 star minus
Delta 3 so I again take different
combinations of Delta 3 Delta 1 which
are given
in column
1 column
2 and the corresponding values of X1 X2
X3 are given in column 3 and the and the
corresponding values of F2 double
star is given in the last column where I
put a double tick mark black
again the combinations of color like
dark yellow green blue are done here in
the same way and only change which you
see is the decision variables so for 39
it was if you remember when the 39 value
was for
f double star F double star means F1
double star here so the values were
1910 here also you get 39 for F2 double
star the values are
091 so the other combinations are just
flipped and the values come out to be
exactly the same and again this Delta 3
Delta 1 which is there in the First
Column second column doesn't make much
sense only our concentration should be
on the decision variables X1 X2
X3 these are again replication of the
same way for
f235
the decision variables are 0 15 5 now we
come to maximizing the third objective
which is
F3 subject to the same set of initial
conditions which are marked here as 1 2
3
4 and
five
and the first objective there is there
in the constraint which I put a red tick
mark and second objective in the
constraint which I put a bullet tick
mark again blue vet tick mark again they
are of the less than type so I have to
subtract a value of delta 1 Delta 3 in
order to obtain feasible
Solutions the table which is shown here
we do the exactly the same analysis
First Column is related to Delta 1
values which are to be subtracted from
fub1
Star then the second column are the
values of Delta 2 which are to be
subtracted from F2
star and the third column and fourth
column where I put a single tick mark
and a double tick mark are the
corresponding values of X1 X2 X3 the
decision variables which is in the third
column and the fourth column are the
corresponding values of F3 double star
and if you see fub1 double star F2
double star F3 double star if the values
come out to be same the X Vector which
is X1 X2 X3 obviously
changes so in the first case when it is
39 for F1 double star it was
1910 again going
back 191 0 here
I will highlight with say for
example green
color 191 0 for
39 then for 39 when I'm interested for
F2 double star it is 091 where I'm
circling the decision variables and 39
is also circled so this is the
repetition for the other dark yellow
which I'm not
repeating and for F3 double star again
it's
1019 this 39 is F3 star again a
repetition here so the green blue and
the yellow which you are seeing in the
slide had been already
discussed and in the same way we can
obtain that we have already obtained
when we are trying to maximize objective
function
one with second and third objective in
the
constraint then maximizing object
function two with first and third
objective in the constraint and this one
which you're seeing is maximizing object
function three with objective function
one and two in the constraint so these
are the
repetion
now
I take all the values from this
combinations and make a table and I have
marked it with different colors with a
specific reason and I'll highlight that
first concentrate on Row one only if you
see Row
one it has X1 X2 X3 given in red color
and the corresponding F values fub1
double star F2
F3 are given in the fourth column where
I put a double tick mark in
Red so the First Column would go hand in
hand in trying to
analyze with the fourth
column if I
consider and and obviously that becomes
very obvious if you see the heading of
the First Column which is X1 X2 X3 is in
red similarly The Heading which is in
the fourth color fub1 double star is in
red FS2 and F3 are in their usual
black font size or font color
if you see the second
column and the fifth
one this one the second column is
corresponding to when I'm only
optimizing the second objective function
with the first and the third are in the
constraints that's why it's shown in
Violet X1 X2 X3 Violet and the
corresponding heading of the the fifth
column is given as F1 is black font
color F2 double star which is violet and
F3 is black font color finally so I'm
just mentioning what are the headings
then it will be easy for us to go
through
finally you have the blue
one which is the third and the last
column and it basically corresponds to
optimizing the third objective function
with the first and the second in the
constraints so that's why it's shown in
blue so X1 X2 X3 third column blue and
in the last column is F1 black font
color F2 black font color and F3 star is
blue as it should
be now let us see why did we make
different coloring schemes for the
values which are there in the table
let us consider the part corresponding
to the yellow one so if I consider the
yellow
one the corresponding values of X1 X2 X3
corresponding to F1 being
optimized in the objective function are
1910 and the values for fub1 which is
first objective being optimized are 39 9
18 18 18 18 are the
values for the constraints corresponding
to the second and the third which was
basically 40 because it was Delta 2
Delta 3 so the depending on the values
of Delta 2 Delta 3 we obtained FS2 minus
Delta 2 F3 minus Delta 3 those values
are
189 now when we consider the values
yellow in the second
cell it is
091 which is for second objective being
optimized first and uh third in the
constraint and the corresponding values
of fub1 F2 star double star and F3 are
18398 so here it was done by fub1 minus
Delta 1 and F3 - Delta 3 so
corresponding to Delta 1 and Delta 3 as
22 it was 40 - 22 40 - 22 gave us the
value of 1818 as you see so I'm not
writing Deltas here I'm writing 40 40
was basically the max value which we
obtained for optimizing fub1 FS2 F3 as
single objective so we are writing
that difference of fub1 minus Delta FS2
- Delta F3 minus Delta fub1 F2 F3 are
basically the max values fub1 start F2
start F3 start as you want to
see similarly for the third
cell which is the heading is shown in
blue for X1 X2 X3 the values of decision
variables are 1 0 and 19 and the
corresponding values of fub1 FS2 F3 star
R1 1818 F3 double star R1 181
1839 39 is the value which you are
getting which you already seen in the
slides if you go back to the slides you
will understand
and these 1818 are values of fub1 star
minus Delta 1 F2 star minus Delta 2 so
before moving on to the other color like
the
dark yellow green pink blue and
subsequently so the values which you
have seen in the yellow one and what I
was talking about this 18 being obtained
from 40 minus 22 it will become clear
here if you consider the first one it
was basically for the so consider
here so this is basically 22 which is
Delta
2 and if I subtract 40 - 22 this is
18 here
also 40 -
22 is equal 18 and the objective
function F1 star was
39 so it was 39 1818 which we
saw and the values of X1 X2 X3 which
were red in color are 19 0 1 0 here it
is 39 18 18 19 1 0 so this
one
right
now when I I go to F2 double
star so again it is D3 D1 is 1818 so
This Is
40 - 22 is
18 here also is 40 - 22 is
18 and this is 39 so this was basically
oh I should basically use the sorry my
Miss I should use the violet color
because we are following a uh coloring
scheme which is much easier for us to
discuss so if
I so this was 4 -
22
18 40 - 22
18 so the values
of fub1 FS2 double star and F3 would be
18 39
18 and the values of X1 X2 X3
is
0191 so here you
see
0191 and 18 39 18
091 091 which you see and 18 39 18
finally when we see this
case are the values here so for for F3
dou star let us go back to F3 dou
star so I'll remove this and go
to so this is equal to 40 -
22 is =
18 Again 42 - 22 is =
18 so this value is 39
and the values of X1 X2 X3 are
109 so actually the value for the
decision variables are 1 19 and fub1 F2
and F3 double stars are 18 18
39 so this is here
1019 and
18839 now with the same logic I repeat
now before I repeat an interesting point
to be noted if I consider the sum of F1
F2 F3 as the overall objective function
then very interestingly you
see the following so I'll remove the
colors not to make it too
cluttered here it
is so if I
consider the objective function fub1
double star f F2 F3 if you consider the
total sum as uh a case on the
multiobjective not singularly as a case
for
multiobjective so the
value
is 8 + 18 18 36 36 is 5
or 36
75 here also
75 here also 75
and the all the objective functions are
same and these points which I'm marking
as in red as
one Violet as one y coloring scheme is
the same thing to differentiate blue as
one then the red violet blue are the par
points with different points for the
decision variable the object function if
if I consider as a sum of the objective
functions of fub1 FS2 F3 which double
star being for the first case second
case third case and the sum being 75 is
the case or a nice way of trying to
explain that for different paror points
we get the same objective
function now if we repeat the same
Concepts considering the sum of the
objectives F1 F2 F3 as the
multiobjective case then let us see for
the case when you
have the value of fub1 double star or
FS2 double star or F3 double star as
38.5 then this
dark
yellow set of values would corresponding
correspond to the
case where you would
have the Paro points this is number the
one of
them second for
them and third for
them and what is the objective function
add them up so you will have it so here
the objective function if I denote
it would
be 0.5 and this is 33
33 71.5
71.5
71.5 now if I go to the two ones which
are marked in light and the the C green
or light green color the last one others
can be repeated for the pink one and the
green one we can repeat the same thing
but I'm going to the last
one if you
see the setup points very
interestingly
1550 which I will mark it as red one why
red one because we are trying to
optimize
F1 then the Violet
one and the blue
one so I should Mark it
sorry I should Mark it
as
and the corresponding values which I get
is very interesting even though that
would not give the exact idea of par
optimal points because the indifference
concept don't cross each other so if I
consider the sum for the second last
column so this would
be is
8 50 8 58 for
all what of object function which is sum
of F1 F2 F3 which is 58 the paror points
are given the last two columns as 1 1 1
which is 1 15 15 5 0 0 15 5 and 5 0
15 but for the problems as we have
considered very
interestingly some of the last
values is given
as
8 +
3 3 1 4 5 53 here 53 here 53 so value of
our 53 and 58 we get same sets of points
theoretically it may not be possible
because technically then in that that
case for the same set of points if
you're considering the utility you
getting different values we will take
this problem later
on so with this I will end this uh
lecture and continue discussing more
about it in the subsequent thank you
very much and have a nice
[Music]
day
[Music]
[Music]
