0:00

video i will discuss how to apply k

0:02
nearest neighbor algorithm to classify a
0:04
new instance
0:06
based on the given data set
0:09
this is the solved example number four
0:11
the link for other examples is given in
0:13
the description below
0:16
this is the data set given to us it has
0:18
four instances i1 i2 i3 and i4
0:22
x1 and x2 are the two features in this
0:24
case
0:25
output is the target variable
0:28
the output has two possibilities either
0:30
zero or one in this case
0:32
we have to calculate the or we need to
0:35
classify this new instance that is uh t1
0:37
is equivalent to three comma seven
0:40
into one of the classes by considering
0:42
the number of neighbors as three here
0:46
to do this one first what we do is we
0:48
will calculate the distance
0:50
from this particular new instance
0:53
to the existing data points
0:56
uh we use uh euclidean distance in this
0:58
case
0:59
the distance calculation looks something
1:02
like this
1:03
first what we do is we calculate the
1:04
distance from new instance to the
1:06
existing point that is always equivalent
1:09
to
1:10
for this particular first instance
1:13
square root of
1:14
7 minus 3 bracket square
1:17
plus 7 minus 7 bracket square here so
1:20
that is what i have written which is
1:21
equivalent to 4 in this case
1:23
similarly with respect to second
1:25
instance
1:26
7 minus 3 bracket square plus 4 minus 7
1:30
bracket squared so that is what i have
1:31
written here and the answer is 5 here
1:34
with respect to the third one
1:36
square root of 3 minus 3 bracket square
1:38
plus 4 minus 7 bracket square this is
1:40
what the case and the answer is 3 here
1:43
the last one is
1:44
square root of 1 minus 3 bracket square
1:46
plus 4 minus 7 bracket square
1:49
answer is 3.6
1:51
so once you calculate the
1:52
distance with respect to the new
1:55
instance
1:56
what we need to do is we need to assign
1:58
ranks to this particular training
2:00
instances
2:02
based on the distance so the minimum
2:04
distance in this case is 3 hence we
2:06
assign the rank to this one as one so
2:09
the third instance is rank one with
2:11
respect to any instance
2:13
similarly 3.6 is the second
2:16
smallest value so we assign 2 as the
2:18
rank to this one that is i4 is the
2:21
second nearest neighbor to the new
2:23
instance
2:24
similarly the i1 will be assigned a rank
2:27
3 i2 is assigned a rank 4 in this case
2:31
now once you assign the neighbor rank to
2:33
these particular instances we have to
2:35
check what is the value given to us that
2:37
is the
2:38
k value
2:39
so k value in this case is 3 here so we
2:43
need to identify the three nearest
2:45
neighbors so the first nearest neighbor
2:47
is this one i3 second nearest neighbor
2:49
is i4
2:50
third nearest neighbor is i1 because we
2:53
have assigned the rank over here
2:55
now we have to check the output variable
2:57
with respect to this one so the output
2:59
variable is
3:01
0 here 1 here one here and the majority
3:05
in this case is one hence the answer in
3:08
this case is one or you can say that the
3:10
new instance is classified as one in
3:13
this case
3:14
so this is a simple processor we follow
3:17
so that we can classify a new instance
3:19
into one of the classes here
3:21
but what has happened here is we have
3:23
calculated the distance
3:25
and we have
3:26
assigned the rank to these particular
3:28
instances but we haven't considered the
3:30
actual value of this particular the
3:32
distance
3:34
actually what should happen here is the
3:35
value or can say that if the distance is
3:37
very small
3:39
at that should be given the maximum
3:41
weightage
3:42
so for that reason what we do here is we
3:45
calculate something called as
3:47
the inverse square distance that will be
3:50
is considered as a weight for that
3:51
particular uh instance
3:54
that is also called as a distance
3:55
weighted nearest neighbor algorithm
3:58
so till here the procedure is same
4:01
we calculate the square distance first
4:03
that is 4 square phi square 3 square 3.6
4:06
packet square
4:07
which is equivalent to 16 25 9 and 12.96
4:12
and then we calculate the inverse of
4:14
this uh distance square that is 1 by 16
4:16
which is equivalent to 0.06
4:18
1 by 25 which is equivalent to 0.04
4:21
1 by 9 is equivalent to 0.11
4:24
1 divided by 12.96 which is equivalent
4:27
to 0.08
4:28
now if you'll notice here the minimum
4:31
distance is 3
4:32
and the value or the inverse square
4:34
distance value is more for this one the
4:36
meaning is minimum the distance maximum
4:39
the weight assigned to that particular
4:40
instance here
4:42
similarly 3.6 is the second minimum we
4:44
got the maximum
4:47
what you can say vote for this
4:48
particular instance also so once you get
4:51
this particular the vote for each and
4:54
every instance we need to assign rank to
4:56
this particular instances based on this
4:58
weight
4:59
the one which is having the maximum vote
5:02
that is nothing but the more nearest
5:04
neighbor here so in this case the point
5:06
one one is having the maximum value so
5:08
this will be assigned a rank one second
5:11
highest is point zero eight it will be
5:14
assigned rank two
5:15
third one is uh point zero six which
5:18
will be assigned rank three point zero
5:19
four is the minimum which will be
5:21
assigned a rank uh 4 in this case
5:24
again we have to
5:26
check the majority vote based on this
5:28
particular rank this is the first
5:30
nearest neighbor second nearest neighbor
5:33
third nearest neighbor because the k
5:34
value is equal to 3 again
5:37
so if you look at this particular
5:38
majority one is the maximum time it
5:41
appearing in this case so we classify
5:43
the new example again as one in this
5:45
case
5:46
so i hope you understood the procedure
5:48
how can we apply the k nearest neighbor
5:50
algorithm as well as a weighted
5:52
k-nearest neighbor algorithm
5:54
to find or can say that to classify a
5:56
new instance into one of the classes
5:58
here
5:59
if you like the video do like and share
6:01
with your friends
6:02
press the subscribe button for more
6:04
videos press the bell icon for regular
6:06
updates thank you for watching