1
00:00:00,640 --> 00:00:01,450
Welcome back.

2
00:00:02,020 --> 00:00:04,320
And I couldn't be more excited for this section.

3
00:00:04,360 --> 00:00:05,260
You know why.

4
00:00:05,260 --> 00:00:11,710
Because we're going to finally start getting hands on with an end to end project putting all the tools

5
00:00:11,710 --> 00:00:15,120
that we've learned so far on the processes in the frameworks and whatnot.

6
00:00:15,160 --> 00:00:16,510
We're gonna put them all together.

7
00:00:16,600 --> 00:00:21,400
We're gonna make something of it we're gonna work on a project specifically with structured data or

8
00:00:21,400 --> 00:00:23,050
a couple of projects actually.

9
00:00:23,050 --> 00:00:27,250
And so you might be asking what is structured data.

10
00:00:27,310 --> 00:00:33,890
And while we've actually been hands on with structured data throughout what we've learned so far basically

11
00:00:34,010 --> 00:00:39,610
structured data is whatever you can fit into this sort of structure more structure is this.

12
00:00:39,680 --> 00:00:46,670
Well this is rows and columns or something like a panda's data frame or something like an Excel spreadsheet

13
00:00:46,670 --> 00:00:49,070
or maybe a Google Sheets document.

14
00:00:49,070 --> 00:00:55,130
And so without heart disease data set that we've been working with as kind of a demo on the left hand

15
00:00:55,130 --> 00:01:00,020
side you could have something like patient I.D. and in the middle you might have some feature variables

16
00:01:00,500 --> 00:01:07,610
such as details about the patient their weight their sex heart rate chest pain and then on the right

17
00:01:07,610 --> 00:01:10,480
hand side or somewhere in the data set.

18
00:01:10,490 --> 00:01:12,080
You may have a target variable.

19
00:01:12,230 --> 00:01:16,970
The thing you want to predict and that's in essence what we're trying to do with machine learning we're

20
00:01:16,970 --> 00:01:23,540
trying to learn some patterns in feature variables and how those patterns relate to a target variable.

21
00:01:23,750 --> 00:01:28,820
And that's what we're gonna do throughout this section is work on structured data projects which take

22
00:01:28,820 --> 00:01:34,850
this sort of approach getting a machine learning model to learn patterns within feature variables and

23
00:01:34,850 --> 00:01:40,750
then using those patterns to predict some target variable now how are we going to do that.

24
00:01:41,200 --> 00:01:44,080
Well we're going to put this bad boy into play.

25
00:01:44,140 --> 00:01:46,170
We've seen it a few times already.

26
00:01:46,180 --> 00:01:48,130
We'll go through it step by step.

27
00:01:48,130 --> 00:01:51,520
The data modeling section of a machine learning project.

28
00:01:51,520 --> 00:01:53,800
So this is where we already have some data.

29
00:01:53,980 --> 00:01:56,860
We're going to go through we're going to define what our problem is.

30
00:01:56,860 --> 00:01:58,300
What problem are we trying to solve.

31
00:01:58,300 --> 00:01:59,920
Is it classification is it regression.

32
00:01:59,920 --> 00:02:01,440
Is it something else.

33
00:02:01,480 --> 00:02:02,950
What data do we have.

34
00:02:02,950 --> 00:02:06,420
We'll look at the dataset where might that actually come from.

35
00:02:06,460 --> 00:02:07,840
What defines success for us.

36
00:02:07,840 --> 00:02:12,700
So if we're working through a machine learning project and we're working on machine learning proof of

37
00:02:12,700 --> 00:02:18,370
concept what level of success do we have to get to before we can say Okay this may move from proof of

38
00:02:18,370 --> 00:02:22,820
concept into actually being deployed what features should we model.

39
00:02:23,170 --> 00:02:27,820
So we have a look again back at the data and are there certain features of the data that should take

40
00:02:27,820 --> 00:02:29,500
more precedence over others.

41
00:02:29,530 --> 00:02:30,850
What kind of model should we use.

42
00:02:30,850 --> 00:02:35,260
So we looked at this a little bit in psychic learn of how to choose a different machine learning model

43
00:02:35,260 --> 00:02:39,580
for different problems and then finally what experiments could we try.

44
00:02:39,580 --> 00:02:41,070
So what have we tried.

45
00:02:41,080 --> 00:02:42,910
What else can we try and the whole thing here.

46
00:02:42,910 --> 00:02:45,870
This is an iterative process that's the main thing to remember.

47
00:02:46,030 --> 00:02:51,970
And then again if we go here we've got some tools that we can use for each step of the pipeline so if

48
00:02:51,980 --> 00:02:59,360
this whole overall pipeline is data science all the tools we're going to focus on are Panda's map plot

49
00:02:59,360 --> 00:03:03,160
lib num pi psychic line and Jupiter.

50
00:03:03,320 --> 00:03:07,850
At least for the first project we got some green boxes highlighting the ones that we're going to be

51
00:03:07,850 --> 00:03:12,230
focused on and we've already been hands on with some of these and we've seen what they can do.

52
00:03:12,230 --> 00:03:18,380
But now it's time to combine them all in an overall project setting and there's one more diagram we

53
00:03:18,380 --> 00:03:19,690
have to look at.

54
00:03:19,730 --> 00:03:23,660
Well this is the steps we're going to take to set up a new project the steps you could take with almost

55
00:03:23,660 --> 00:03:26,290
any machine learning project.

56
00:03:26,290 --> 00:03:30,470
And so far we've got our computer We've downloaded installed many conduct.

57
00:03:30,590 --> 00:03:34,070
We followed this but this year we're up to now.

58
00:03:34,100 --> 00:03:39,320
So this is if we have our computer and we have Condor or mini Condor we're gonna start a new project

59
00:03:39,800 --> 00:03:44,560
and then create a project folder and within that folder will leave some data.

60
00:03:44,870 --> 00:03:48,630
We're going to create an environment similar to the one we've already been working in.

61
00:03:48,720 --> 00:03:53,840
And then within that environment we can create our workspace in the form of a Jupiter notebook.

62
00:03:54,260 --> 00:04:00,320
And then within that workspace we can perform data analysis and manipulation using tools like num pi

63
00:04:00,350 --> 00:04:05,630
panders and map plot lib and then we'll start to look at how we can build or use machine learning models

64
00:04:05,630 --> 00:04:10,640
within psychic learn so psychic line has those pre-built machine learning models that we can use.

65
00:04:10,810 --> 00:04:14,370
We're going to see how we can apply those to our problem.

66
00:04:14,390 --> 00:04:17,520
So you might be wondering of course where can you get help.

67
00:04:17,670 --> 00:04:22,400
And the first step is to because we're going to be working through this project together is to follow

68
00:04:22,400 --> 00:04:23,900
along with the code if you can.

69
00:04:23,900 --> 00:04:28,790
If you fall behind don't worry it will be available to you through some sort of resource through some

70
00:04:28,790 --> 00:04:32,620
sort of link in the place where you get those extra resources.

71
00:04:32,960 --> 00:04:36,740
If you are following along try it for yourself or if you're aren't going along with it at a different

72
00:04:36,740 --> 00:04:37,490
time.

73
00:04:37,640 --> 00:04:39,710
Make sure if in doubt run the code right.

74
00:04:39,710 --> 00:04:40,620
That's the motto here.

75
00:04:40,630 --> 00:04:44,900
Find out run the code if you're curious about what a function does.

76
00:04:44,900 --> 00:04:50,570
Remember you can press shift tab within the brackets of a function to read the dock string that will

77
00:04:50,570 --> 00:04:55,490
give you a brief overview of what the function does and then don't be afraid if these three steps don't

78
00:04:55,490 --> 00:04:56,050
work.

79
00:04:56,240 --> 00:04:58,270
Don't be afraid to search for it right.

80
00:04:58,310 --> 00:05:03,650
You'll probably end up in places like stack overflow or the documentation I put both documentations

81
00:05:03,650 --> 00:05:07,580
here for cyclone and panders because we're gonna be using a few tools here.

82
00:05:07,700 --> 00:05:11,630
So you might even have to look up the documentation for num Pi or something like that.

83
00:05:11,630 --> 00:05:14,080
Another tool that we're using maybe even that plot lib.

84
00:05:14,180 --> 00:05:17,150
But wherever you get stuck Don't be afraid to search for it.

85
00:05:17,190 --> 00:05:23,100
What separates a good data scientist from a from a poor one is one that doesn't ask questions.

86
00:05:23,150 --> 00:05:28,550
So a good data scientist is just the one that keeps asking more and more questions a good machine learning

87
00:05:28,550 --> 00:05:32,510
engineer is just the same as a as a normal machine learning engineer.

88
00:05:32,540 --> 00:05:35,890
But they ask more questions and this is where the try again comes in.

89
00:05:35,890 --> 00:05:40,520
So once you've searched for it once you've figured out something once you've read something don't be

90
00:05:40,520 --> 00:05:42,620
afraid to take it back to your notebook and try it.

91
00:05:43,100 --> 00:05:48,950
And then finally as I said before what separates a poor machine learning engineer from a good machine

92
00:05:48,950 --> 00:05:54,080
learning engineer is the questions that they are had never be afraid to ask a question you might do

93
00:05:54,080 --> 00:05:59,350
this in the forum space or in the stack overflow section wherever you see fit.

94
00:05:59,510 --> 00:06:02,810
So without any further ado let's jump in.

95
00:06:02,810 --> 00:06:04,420
I'm really excited I hope you are.

96
00:06:04,430 --> 00:06:08,990
Let's get our hands on our first end to end machine learning project.