1
00:00:00,690 --> 00:00:06,000
In the last few lectures, we saw how to predict a binary response using a single predicted.

2
00:00:06,920 --> 00:00:10,860
Now let us see how to handle multiple predictors.

3
00:00:11,700 --> 00:00:14,070
So mathematically, the danger is very small.

4
00:00:16,320 --> 00:00:19,320
Earlier, we had only B does, you know, plus B, the one X.

5
00:00:19,830 --> 00:00:25,890
Now it will include all the B predictors that we are going to use and will have correspondingly B plus

6
00:00:25,890 --> 00:00:26,100
one.

7
00:00:26,100 --> 00:00:26,500
B does.

8
00:00:26,500 --> 00:00:27,330
So we predicted.

9
00:00:29,890 --> 00:00:32,410
With this small change, we will grindy model again.

10
00:00:32,680 --> 00:00:40,810
This is maximum likelihood criteria and we will get the values of all these redoes using the beta values,

11
00:00:40,930 --> 00:00:45,010
we can get the probability value for any set of observations.

12
00:00:45,520 --> 00:00:51,580
And if the probability is greater than the boundary limit, which we can set as but our own convenience

13
00:00:52,660 --> 00:00:57,720
for our house placing it, it will be using a body limit of zero point five.

14
00:00:59,200 --> 00:01:05,260
So if the program is coming out to be more than point crape, we will see that the house will be sold

15
00:01:05,320 --> 00:01:06,280
within three months.

16
00:01:06,670 --> 00:01:11,620
If it is coming less than point, we will say that it will not be sold within three months.

17
00:01:13,600 --> 00:01:16,300
So this is how we handle multiple predictors.

18
00:01:17,390 --> 00:01:23,410
Anything I want to talk about here is can we handle multiple classes in the response variable?

19
00:01:24,310 --> 00:01:30,640
So right now our dataset has only two classes and the response variable, that is whether it will be

20
00:01:30,640 --> 00:01:32,570
sold within three months or not.

21
00:01:33,850 --> 00:01:41,080
If our response variable had three classes, say, whether it will be sold within three months, within

22
00:01:41,080 --> 00:01:42,790
a year or more than a year.

23
00:01:43,840 --> 00:01:44,680
Can that be done?

24
00:01:46,090 --> 00:01:46,930
Answer is yes.

25
00:01:47,470 --> 00:01:54,010
Logistic regression can be applied for multi class response variable, and it can be done in most software

26
00:01:54,010 --> 00:01:56,110
packages like art and Python easily.

27
00:01:56,920 --> 00:02:02,080
But in practice, we do not use logistic regression in such a scenario.

28
00:02:03,160 --> 00:02:07,930
One of the popular techniques in such a scenario is linear discriminant analysis.

29
00:02:10,030 --> 00:02:15,790
So let us first train the model with binary classes and all the predictors we have in our dataset.

30
00:02:17,140 --> 00:02:21,640
Then I will tell you about linear discriminant analysis in the coming few lectures.