1
00:00:01,220 --> 00:00:06,260
Now, let's start building as swim regression, Morden and Biton.

2
00:00:07,530 --> 00:00:14,520
I have provided you the links for oficial, a skill learned document of Astrium regulation.

3
00:00:15,570 --> 00:00:18,750
You can click on this link to open the documentation.

4
00:00:21,080 --> 00:00:27,830
Here you can find the syntax and the parameters of this function.

5
00:00:29,930 --> 00:00:32,960
The first parameter here is colonel here.

6
00:00:33,260 --> 00:00:37,220
You have to specify what Colonel you want to use by default.

7
00:00:37,250 --> 00:00:40,650
This is R, b, f, r, b, if it sense for radial.

8
00:00:41,970 --> 00:00:48,510
But you can also choose linear polynomial or sigmoid function, for our example.

9
00:00:48,600 --> 00:00:51,210
We will use the linear cardinal.

10
00:00:52,110 --> 00:00:58,370
The second parameter here is degree degrees sense for the degree of polynomial kernel.

11
00:00:59,250 --> 00:01:00,390
Then we have gamma.

12
00:01:00,840 --> 00:01:05,970
Gamma is only relevant and radial polynomial or sigmoid function.

13
00:01:07,170 --> 00:01:10,800
Again, we have another parameter which is specific to polynomial.

14
00:01:10,950 --> 00:01:12,270
That is coifs zero.

15
00:01:14,540 --> 00:01:23,090
And then we have see, this is the penalty parameter C, which we have already discussed during over

16
00:01:23,270 --> 00:01:24,200
two reflectors.

17
00:01:26,270 --> 00:01:32,510
And there are some other parameters as well, which we will not use while creating our modern.

18
00:01:35,790 --> 00:01:37,710
Now, these are the attributes.

19
00:01:38,830 --> 00:01:47,860
The parameters are the variables you need to provide value training, your model and attributes contain

20
00:01:47,860 --> 00:01:51,670
the information of the model that you have already trained.

21
00:01:52,630 --> 00:01:58,900
So, for example, a few write object name and you will provide DOD support underscored.

22
00:02:00,130 --> 00:02:03,580
It will give you the indices of the support vector.

23
00:02:04,840 --> 00:02:10,780
Similarly, if you want to view all the support vector, you can just right, you are model, object,

24
00:02:10,780 --> 00:02:12,670
name and dot support.

25
00:02:12,700 --> 00:02:14,620
Underscore vectors underscore.

26
00:02:16,600 --> 00:02:18,280
And it will give you all the support.

27
00:02:18,460 --> 00:02:18,960
But that's.

28
00:02:20,850 --> 00:02:25,850
So for each function, we have parameters and attributes in a scalar.

29
00:02:27,030 --> 00:02:29,130
Now let's go back to our notebook.

30
00:02:31,750 --> 00:02:37,140
So for every Skillern model, you have to follow the same astep.

31
00:02:37,780 --> 00:02:41,880
First, you have to import the function that we are doing here.

32
00:02:43,340 --> 00:02:45,470
Then we have to create an object.

33
00:02:46,250 --> 00:02:52,760
This can be your regression object or classification object using the function that you have imported.

34
00:02:53,570 --> 00:02:56,330
So here I am creating an object.

35
00:02:57,620 --> 00:03:04,460
By calling this function as we are calling to hyper parameters, we want linear kernel.

36
00:03:04,610 --> 00:03:09,470
That's why we are providing first parameter as cardinal, equal to limit.

37
00:03:10,280 --> 00:03:13,910
And we want our Seawell you panel do a load to be tosing.

38
00:03:14,390 --> 00:03:16,470
That's why we are providing sequel 2000.

39
00:03:17,720 --> 00:03:22,130
All the other hyper parameters will use there before you will lose.

40
00:03:23,000 --> 00:03:31,790
If you want to know the default value, you can open the documentation or you can just you can just

41
00:03:31,790 --> 00:03:35,030
press shift tab inside your function.

42
00:03:37,300 --> 00:03:38,890
First, we will run this.

43
00:03:39,760 --> 00:03:46,410
Now, if I press shift PEB inside my function, I will get all the parameters and their default value.

44
00:03:48,580 --> 00:03:49,690
In the next to step.

45
00:03:51,310 --> 00:03:56,560
We will use our object to fit graining data into that object.

46
00:03:58,300 --> 00:04:04,600
So since we have created our extreme standard variable, so we will use that variable and we will use

47
00:04:04,600 --> 00:04:06,610
our wide-screen variable as it is.

48
00:04:07,510 --> 00:04:09,430
So we will write as we are.

49
00:04:10,000 --> 00:04:16,990
This is our variable that we have created earlier, and we will use Daudt Fit Method to fit our model.

50
00:04:18,730 --> 00:04:23,500
If I run this now, I have fitted my model.

51
00:04:24,280 --> 00:04:29,220
I can use this model to predict the values of Y.

52
00:04:32,530 --> 00:04:38,920
Now I can predict Y values by using Daudt predict method of what object.

53
00:04:39,520 --> 00:04:43,750
So we want to assign our predicted values in y underscored test.

54
00:04:43,790 --> 00:04:45,370
Underscore red and white.

55
00:04:45,370 --> 00:04:47,210
Underscore train underscore spread.

56
00:04:47,710 --> 00:04:51,640
This will contain my predicted values on test data.

57
00:04:51,820 --> 00:04:55,840
And this will contain projected values on our train data.

58
00:04:57,280 --> 00:05:01,770
We can radically use as we are dort predict method and in bracket.

59
00:05:02,110 --> 00:05:04,330
We have to mention the X variable.

60
00:05:04,660 --> 00:05:12,040
So for our test data I am providing X underscored tests under the gold standard and for train data.

61
00:05:12,340 --> 00:05:15,790
We are providing X underscore train underscore standard.

62
00:05:16,840 --> 00:05:17,920
Let's run this.

63
00:05:20,470 --> 00:05:27,730
Now I have my predicted values in these two variables to view our predicted values.

64
00:05:28,180 --> 00:05:33,790
You can just state any of these two variables and executed.

65
00:05:37,220 --> 00:05:40,190
You can see this are over dedicated values.

66
00:05:41,770 --> 00:05:45,460
Now we want to know how our model is performing.

67
00:05:47,080 --> 00:05:51,790
So we have already predicted our test and Kreen y values.

68
00:05:52,300 --> 00:06:00,230
We will use these values to find meaning squared error and are square scored with a Skillern.

69
00:06:00,610 --> 00:06:02,350
We get this to escort's.

70
00:06:03,550 --> 00:06:10,660
We will import DESPUES scores and we will use our credit card values to find this escort's.

71
00:06:13,600 --> 00:06:22,120
Now, if you're not aware, what I mean is squared error means it mean that it is the mean of the square

72
00:06:22,150 --> 00:06:26,590
of difference between the word predicted values and actual values.

73
00:06:28,860 --> 00:06:38,200
And if you want to know about what parameters do we need to execute this, you can just press shift

74
00:06:38,250 --> 00:06:40,500
Tebbe here.

75
00:06:40,620 --> 00:06:44,980
You can see the first parameter is via Google News.

76
00:06:45,960 --> 00:06:51,310
Since our actual values are in via underscore test, we will.

77
00:06:51,320 --> 00:06:51,650
Right.

78
00:06:51,690 --> 00:06:52,980
Why underscore test?

79
00:06:53,500 --> 00:06:57,830
And the word second parameter here is predicted values.

80
00:06:58,390 --> 00:07:04,770
Since we have predicted our values and why underscore tests underscore bread, we will use this with

81
00:07:04,770 --> 00:07:08,070
Able and define Londis.

82
00:07:09,300 --> 00:07:14,970
The masc value on our test dataset is 160 million.

83
00:07:16,090 --> 00:07:23,780
Now, the thing with MASC is you can only compare model performances on the same data.

84
00:07:24,220 --> 00:07:26,150
So this are absolute number.

85
00:07:26,770 --> 00:07:32,560
So you can only compare it with the masc value of some other model on this data.

86
00:07:33,340 --> 00:07:37,180
You do not get any information by just looking at this number.

87
00:07:38,620 --> 00:07:45,700
So the better option here is to use artist square values to evaluate the performance of more than.

88
00:07:47,550 --> 00:07:54,570
Now, again, to look at the parameters that we need for our art to underscore its core function.

89
00:07:55,380 --> 00:07:57,340
Just click shift tab.

90
00:07:58,290 --> 00:08:04,380
And here also we need to put away the true value and the predicted value as our parameter.

91
00:08:07,350 --> 00:08:08,190
We'll run this.

92
00:08:09,060 --> 00:08:12,780
The oddest, good value on our training site, this point, seven one.

93
00:08:14,100 --> 00:08:15,230
This is quite good.

94
00:08:17,700 --> 00:08:25,080
Now, if you don't know about our a squirrel, you artist squirrel loose ranges from minus one to one,

95
00:08:26,010 --> 00:08:32,220
one or minus one means perfect fit and zero means no fate.

96
00:08:33,150 --> 00:08:41,760
In other words, one means your model is able to convey all the variations in your wide dataset, and

97
00:08:41,760 --> 00:08:47,940
zero means your model is not able to convey any information for our Y variable.

98
00:08:48,930 --> 00:08:55,770
So ideally you are discussing value should be between zero point five two zero point nine for a good

99
00:08:55,770 --> 00:08:56,160
model.

100
00:08:57,630 --> 00:09:02,980
For our test dataset, the artist good value is point five again.

101
00:09:03,120 --> 00:09:03,810
This is good.

102
00:09:04,350 --> 00:09:11,430
We can v.c of a sea value to increase our artist good value.

103
00:09:12,600 --> 00:09:21,720
So if you want, you can just provide another Seawell, you suppose three thousand and creadon this

104
00:09:23,430 --> 00:09:24,420
whole code?

105
00:09:30,180 --> 00:09:36,150
You can see now our artists goodwell you on our training despond, so Anwan and artists get a value

106
00:09:36,540 --> 00:09:38,840
on Ortez Sirtis point for me.

107
00:09:41,040 --> 00:09:45,840
So I recommend you to try the value of see as 500.

108
00:09:48,140 --> 00:09:51,950
And try to find out the artists good value for a sequel to 500.

109
00:09:53,980 --> 00:09:57,010
From the next lecture, we'll move on to classification.

110
00:09:57,970 --> 00:10:01,090
And there we will see advanced techniques.

111
00:10:03,230 --> 00:10:07,430
To optimize this value of C to get the best results.
