1
00:00:00,650 --> 00:00:03,440
Step five modeling part two.

2
00:00:03,560 --> 00:00:07,330
Choosing mentioned in the last listen.

3
00:00:07,340 --> 00:00:09,100
There were three parts to modeling.

4
00:00:09,170 --> 00:00:13,310
Choosing a model choosing a model and comparing models.

5
00:00:13,310 --> 00:00:20,990
Once you've got your data split into training validation and test sets you can start to go through each

6
00:00:20,990 --> 00:00:22,390
of these steps.

7
00:00:22,490 --> 00:00:28,700
In this lesson we're going to cover some points on choosing a model which is you choose a model and

8
00:00:28,700 --> 00:00:30,280
train it on your training data.

9
00:00:31,100 --> 00:00:36,620
Unlike creating your own algorithms from scratch there are many pre-built machine learning models which

10
00:00:36,710 --> 00:00:40,260
you can take advantage of when you first begin.

11
00:00:40,400 --> 00:00:46,430
Your main goal will be knowing what kind of machine learning algorithm to use with what kind of problem

12
00:00:46,910 --> 00:00:52,560
this is because some algorithms work better than others on different types of data.

13
00:00:52,610 --> 00:00:56,630
We'll have a look at this specifically when we get hands on with our projects.

14
00:00:56,780 --> 00:01:03,920
But for now a tidbit to remember is if you're working with structured data decision trees such as random

15
00:01:03,920 --> 00:01:11,060
forest and gradient boosting algorithms like cat boost in x g boost tend to work best and if you're

16
00:01:11,060 --> 00:01:18,790
working with unstructured data deep learning neural networks and transfer learning tend to work best.

17
00:01:18,860 --> 00:01:22,140
Once you've chosen a model your next step is to train.

18
00:01:22,520 --> 00:01:27,680
The main goal here will be to line up the inputs and outputs.

19
00:01:27,770 --> 00:01:34,520
For example in our heart disease problem we want our model to look at the feature variables the inputs

20
00:01:35,090 --> 00:01:40,500
and then find the patterns and use them to predict the target variable.

21
00:01:40,500 --> 00:01:47,060
So remember from previous lesson these variables here are the feature variables we want to use these

22
00:01:47,060 --> 00:01:49,580
to predict the target variables.

23
00:01:49,580 --> 00:01:58,600
Another common naming setting is to use x which is the data to predict Y which is the labels different

24
00:01:58,600 --> 00:02:01,450
machine learning algorithms have different ways of doing this.

25
00:02:01,900 --> 00:02:06,340
We'll see how to do it for a handful of useful ones in future projects.

26
00:02:06,580 --> 00:02:11,020
And remember training a model takes place on the training data split.

27
00:02:11,050 --> 00:02:14,490
This is where your model learns the course material.

28
00:02:14,560 --> 00:02:19,780
We don't want to let our models cheat and see the final exam before they do their study.

29
00:02:19,900 --> 00:02:25,790
Depending on how much data you have and how complex your model is training may take a while.

30
00:02:25,870 --> 00:02:32,710
One of your biggest goals when training a model is to minimize the times between experiments.

31
00:02:32,770 --> 00:02:38,420
So sometimes this will mean to use a small portion of your data first.

32
00:02:38,590 --> 00:02:45,340
For example if your training dataset had 100000 examples you might start training a model with only

33
00:02:45,340 --> 00:02:48,770
the first 10000 and see how it goes.

34
00:02:48,790 --> 00:02:55,240
You might also decide to use a less complicated model to begin with deep model such as no networks generally

35
00:02:55,360 --> 00:02:58,350
take longer to train than other kinds of models.

36
00:02:58,350 --> 00:03:02,590
Now this is something worth considering when it comes to training your own models.

37
00:03:02,590 --> 00:03:09,730
For example if an experiment takes you three hours or even up to a couple of days for a small percentage

38
00:03:09,730 --> 00:03:15,050
abuse and performance of your model you might consider is this experiment actually worth it.

39
00:03:15,280 --> 00:03:18,160
Because machine learning is highly iterative.

40
00:03:18,160 --> 00:03:25,090
We want to minimize this experimentation time that we can go from step 1 to step 2 to step 3.

41
00:03:25,090 --> 00:03:32,930
But again if this looks confusing we'll see it in practice in the hands on projects and things to remember.

42
00:03:33,100 --> 00:03:36,280
Some models work better than others and different problems.

43
00:03:36,280 --> 00:03:38,730
Don't be afraid to try things and this is really important.

44
00:03:38,730 --> 00:03:41,910
Machine learning is as we said before a highly iterative process.

45
00:03:41,920 --> 00:03:48,760
So some things that you try out may not work the first time and that means you just neglect that thing

46
00:03:48,760 --> 00:03:51,240
going forward and try something else.

47
00:03:51,280 --> 00:03:54,910
Start small and build up add complexity as you need.

48
00:03:54,910 --> 00:03:59,620
What this means is that for example if you had one hundred thousand examples we're going to start on

49
00:03:59,620 --> 00:04:04,870
ten thousand we're going to use a simple model first to begin with rather than going to the biggest

50
00:04:04,870 --> 00:04:10,690
and latest and greatest model because what we're after is practical results not something that's that's

51
00:04:10,700 --> 00:04:12,190
the best on paper.

52
00:04:12,190 --> 00:04:15,180
We want to see something that can actually be used in the real world.

53
00:04:16,030 --> 00:04:21,850
And part of your experiments will involve tuning a model which shows good initial results to get better

54
00:04:21,910 --> 00:04:22,660
results.

55
00:04:22,660 --> 00:04:27,730
Like tuning a car if your car does well on on one track it might not do well on another track.

56
00:04:27,940 --> 00:04:29,950
So you turn it up.

57
00:04:29,950 --> 00:04:34,510
Let's have a look at how to do that but instead of four cars we'll do it for machine learning models.