1
00:00:00,480 --> 00:00:07,080
OK so in the last section we saw how to choose an estimate how or machine learning model for the problem

2
00:00:07,080 --> 00:00:08,400
that we're working on.

3
00:00:08,490 --> 00:00:09,720
Do we have a regression problem.

4
00:00:09,720 --> 00:00:11,700
Do we have a classification problem.

5
00:00:11,700 --> 00:00:16,470
And we followed this flowchart through and we tried out a bunch of different models such as linear SPC

6
00:00:16,800 --> 00:00:22,590
and ensemble classifiers a.k.a. the random ensemble classifier or the random forest classifier.

7
00:00:22,680 --> 00:00:26,470
That's in and now in this section we're going to have a look at actually.

8
00:00:26,580 --> 00:00:34,530
Let's revisit our what we're covering list so you've ticked off this and inside loan workflow Yep we've

9
00:00:34,620 --> 00:00:40,080
seen getting the data ready Yep we've seen choosing the right estimate a slash algorithm for our problems.

10
00:00:40,080 --> 00:00:40,980
Aha.

11
00:00:40,980 --> 00:00:46,930
Now we're up to here three fit the models slash algorithm and use it to make predictions on our data.

12
00:00:46,930 --> 00:00:51,410
Now we've kind of actually seen this Well we've seen the first half we've seen feeding the model.

13
00:00:51,860 --> 00:00:53,310
Well we're going to have a look at both.

14
00:00:53,310 --> 00:00:57,210
We're going to look at feeding the model and then use it to make predictions on our data.

15
00:00:57,240 --> 00:01:09,060
So what we might do is go three is feed the model slash algorithm on our data and use it to make predictions

16
00:01:10,190 --> 00:01:10,660
wonderful.

17
00:01:10,670 --> 00:01:17,750
And then under here will produce section three point one is fitting the model to the data.

18
00:01:17,750 --> 00:01:22,970
Now we've kind of got some code that will help us out for this already and again not in there.

19
00:01:23,030 --> 00:01:27,620
I'm not a fan of copying pasting code but in the interest of time because otherwise we'll just literally

20
00:01:27,620 --> 00:01:29,960
be re typing everything in this cell

21
00:01:32,970 --> 00:01:34,410
but I trust you might have already done that.

22
00:01:34,940 --> 00:01:36,780
So if not you can retype it out.

23
00:01:36,780 --> 00:01:38,210
I'm going to copy that in here.

24
00:01:38,350 --> 00:01:41,270
And the main line we're concerned about here is the fit.

25
00:01:41,280 --> 00:01:45,660
So if we type in here go fit the model to the data.

26
00:01:45,660 --> 00:01:51,210
So this re step through what we've got here because we didn't type it out we broke our rule but that's

27
00:01:51,210 --> 00:01:53,610
a right from SBA loan dot ensemble.

28
00:01:53,610 --> 00:02:00,180
Import random forest classifier This is saying hey SBA loan ensemble library import the random forest

29
00:02:00,180 --> 00:02:01,560
classifier model.

30
00:02:01,560 --> 00:02:03,350
We've set up a random seed.

31
00:02:03,540 --> 00:02:05,670
We've split our data into x and y.

32
00:02:05,760 --> 00:02:13,170
So X is a feature variables or the features or the data they're different names so I might put in here

33
00:02:13,690 --> 00:02:24,730
or actually maybe up here x equals features feature variables.

34
00:02:25,090 --> 00:02:25,570
Data.

35
00:02:25,570 --> 00:02:33,970
They're just different names for X and Y equals labels or targets or target variables.

36
00:02:33,970 --> 00:02:42,720
So different names for this is just so if you see where you're out and about see different names for

37
00:02:42,720 --> 00:02:46,680
these two X and Y common variable names that you'll see in python.

38
00:02:46,800 --> 00:02:50,780
But these are different like names that you'll be they'll be referred to right.

39
00:02:50,820 --> 00:02:55,890
And so what is fit doing when it goes through here.

40
00:02:55,890 --> 00:02:58,410
So let's have a look at X within visual here.

41
00:02:58,410 --> 00:02:59,850
Right.

42
00:02:59,850 --> 00:03:01,620
Not getting stuck in the theory of things.

43
00:03:01,620 --> 00:03:05,940
So why not head this is x and y and what are we passing and why we're passing it.

44
00:03:05,940 --> 00:03:10,220
The training data and training X and training y to our classifier.

45
00:03:10,230 --> 00:03:11,310
So this is the fit function.

46
00:03:11,310 --> 00:03:16,340
This is what we're concerned about fitting the model to the data what does fit actually do.

47
00:03:16,590 --> 00:03:23,360
Well when we pass x and y to fit it will cause the model to go through all of the examples in X..

48
00:03:23,450 --> 00:03:30,810
The training data in our case and see what their corresponding y label is and try to figure out the

49
00:03:30,810 --> 00:03:34,520
patterns that lead to different combinations of numbers here.

50
00:03:34,650 --> 00:03:40,810
What leads to having a one is a label or a zero as a label in the case of the bottom end of this.

51
00:03:40,830 --> 00:03:46,020
Why not tail yes or a zero as a label that is fit in a nutshell right.

52
00:03:46,110 --> 00:03:52,260
And how the machine learning model in our case a random forest classifier how this does is is going

53
00:03:52,260 --> 00:03:57,300
to be different depending on the model you use and explaining the detail.

54
00:03:57,300 --> 00:04:03,450
So remember all of these green ones here a different machine learning models and explaining the details

55
00:04:03,450 --> 00:04:09,410
of each and every one would require an entire textbook or a much longer video course.

56
00:04:09,480 --> 00:04:09,960
Right.

57
00:04:10,140 --> 00:04:14,700
For now to become a practitioner to start applying machine learning code.

58
00:04:14,700 --> 00:04:21,270
The most important thing you can remember is that it's kind of similar to how you would figure out patterns

59
00:04:21,300 --> 00:04:23,330
if you had enough time.

60
00:04:23,520 --> 00:04:29,580
You'd look at the feature variables in X the age the sex the cholesterol I'm not sure actually what

61
00:04:29,580 --> 00:04:30,490
the rest of these are.

62
00:04:30,480 --> 00:04:35,700
They're just different health attributes and you'd see what different values of these lead to different

63
00:04:35,700 --> 00:04:41,020
values in y one for heart disease or zero for not heart disease.

64
00:04:41,070 --> 00:04:46,590
This concept regardless of the problem is similar throughout all of machine learning.

65
00:04:46,590 --> 00:04:50,510
So during training which is what the fit function is another term for fit.

66
00:04:51,280 --> 00:04:53,100
So we want training.

67
00:04:53,490 --> 00:04:53,840
Okay.

68
00:04:53,850 --> 00:05:00,780
That's why we're using training training machine learning model.

69
00:05:00,780 --> 00:05:06,300
The model is going to find patterns here and then during testing here.

70
00:05:06,510 --> 00:05:12,180
Or when a machine learning model is in production it's going to use the patterns.

71
00:05:12,180 --> 00:05:22,100
So use the patterns the model has learned that is the crux of all of machine learning right.

72
00:05:22,250 --> 00:05:27,380
The way we become practitioners of this is starting off by first diagnosing what problem we're working

73
00:05:27,380 --> 00:05:30,060
with and then coming in finding a model.

74
00:05:30,110 --> 00:05:35,540
Now you can dive in as deep as you want to actually figure out the nuts and bolts of what a random forest

75
00:05:35,540 --> 00:05:38,210
classifier does or what a random forest regressive does.

76
00:05:38,210 --> 00:05:44,430
That may be a bit of an extension we work on in the future but for now that's the crux of it right.

77
00:05:44,450 --> 00:05:49,340
If you want to figure out what a machine learning model does it take some data.

78
00:05:49,340 --> 00:05:54,710
It finds patterns and then it figure out or in supervised learning it figures out how those patterns

79
00:05:54,710 --> 00:05:56,540
relate to labels.

80
00:05:56,540 --> 00:05:57,740
That's in a nutshell.

81
00:05:57,740 --> 00:06:02,420
If you wanted to you can dive as deep as you like into each of these models not focusing on that for

82
00:06:02,420 --> 00:06:02,980
this cause.

83
00:06:03,000 --> 00:06:08,030
Remember our focus is to write machine learning code rather than dive into the theory if you like I'll

84
00:06:08,030 --> 00:06:12,980
leave some extra resources on some deep dives into the random forest model which is the one we've been

85
00:06:12,980 --> 00:06:13,680
using.

86
00:06:13,790 --> 00:06:20,240
But for now we've finally seen how we've heard what the crux of machine learning is finding patterns

87
00:06:20,240 --> 00:06:21,010
in data.

88
00:06:21,170 --> 00:06:31,070
We're gonna see how we can this is three point to make predictions using a machine learning model.

89
00:06:31,070 --> 00:06:38,060
So now that our models have learned patterns in the data by calling the fit function how can we use

90
00:06:38,060 --> 00:06:41,000
what it's loaned to make some predictions on data.

91
00:06:41,010 --> 00:06:44,140
The model hasn't seen we'll have a look at the next video.