1
00:00:00,000 --> 00:00:04,085
[MUSIC]

2
00:00:04,085 --> 00:00:08,377
We've done now some simple regression,
some simple regression for

3
00:00:08,377 --> 00:00:11,440
our data just using square foot of living.

4
00:00:11,440 --> 00:00:15,650
But if you remember, our data set had many
other columns associated with that, or

5
00:00:15,650 --> 00:00:16,650
features.

6
00:00:16,650 --> 00:00:20,827
So what we're gonna do next is

7
00:00:20,827 --> 00:00:26,014
explore other features in the data.

8
00:00:28,350 --> 00:00:32,918
So let's do some exploration of other
possible features we might use.

9
00:00:34,749 --> 00:00:39,310
So I'm going to use a set of features.

10
00:00:39,310 --> 00:00:46,120
So let's create a list of the features
I'm going to explore, the code features.

11
00:00:46,120 --> 00:00:52,041
And these features
are going to be the number

12
00:00:52,041 --> 00:00:57,490
of bedrooms, the number of bathrooms.

13
00:01:03,118 --> 00:01:04,420
Let's see, what else?

14
00:01:04,420 --> 00:01:10,080
The square foot of living space,
which is what we've been exploring so far.

15
00:01:11,290 --> 00:01:15,690
In addition to this, I'm going to
include the square foot of the lot, so

16
00:01:15,690 --> 00:01:18,830
this is how much land
the house has all around it.

17
00:01:20,480 --> 00:01:24,390
The number of floors in the house.

18
00:01:26,210 --> 00:01:31,097
And finally, I'm gonna include
a variable called the ZIP code.

19
00:01:32,800 --> 00:01:36,480
And so the ZIP code in

20
00:01:36,480 --> 00:01:40,070
the United States is what other
countries call the postal code.

21
00:01:41,310 --> 00:01:44,090
In Brazil we call this SAPI.

22
00:01:44,090 --> 00:01:48,130
But there are many names in different
places, that's what we'll include.

23
00:01:48,130 --> 00:01:53,790
So, let's see, if I take the sales
data and I just select this column,

24
00:01:53,790 --> 00:01:59,880
just select the features column here,
why don't we call them just to be totally

25
00:01:59,880 --> 00:02:03,095
clear, instead of calling them features
I'm going to call them my_features.

26
00:02:04,990 --> 00:02:06,820
Shift+Enter.

27
00:02:06,820 --> 00:02:11,050
And then I'm gonna look at what
my_features columns look like.

28
00:02:11,050 --> 00:02:13,820
So I'm gonna type .show.

29
00:02:13,820 --> 00:02:17,020
Remember, we can type .show
anything with GraphLab Create.

30
00:02:17,020 --> 00:02:22,325
On the sframe for sales,
selecting the my_features.

31
00:02:23,710 --> 00:02:27,090
And now we're gonna have
a visualization of these features.

32
00:02:27,090 --> 00:02:29,530
So let me just walk you
through this visualization.

33
00:02:31,270 --> 00:02:32,490
Get my mouse here.

34
00:02:32,490 --> 00:02:35,640
So bedrooms, let's look at frequency.

35
00:02:35,640 --> 00:02:38,280
There are 13 different unique types.

36
00:02:38,280 --> 00:02:40,739
In fact,
there's some houses with ten bedrooms.

37
00:02:43,290 --> 00:02:45,730
Most houses have three bedrooms.

38
00:02:45,730 --> 00:02:49,400
Some have four, some have two,
some have five, and very few have more.

39
00:02:49,400 --> 00:02:50,580
With bathrooms,

40
00:02:50,580 --> 00:02:55,250
it turns out that house in the US,
you have fractional number of bathrooms.

41
00:02:55,250 --> 00:02:58,895
When you say, for example,
2.5 bathrooms is the most common number,

42
00:02:58,895 --> 00:03:04,860
it's because if you have a house
with a bathroom with a bath in it,

43
00:03:04,860 --> 00:03:06,380
it's called a full bath, it counts for 1.

44
00:03:06,380 --> 00:03:12,920
But if you have just a bathroom that has
a sink and a toilet, it's just worth 0.5.

45
00:03:12,920 --> 00:03:15,440
And so here you have 2.5.

46
00:03:15,440 --> 00:03:19,925
In fact, if you have a bathroom
with a sink, a toilet, and

47
00:03:19,925 --> 00:03:24,334
a shower, but no tub,
that's worth 0.75 in the US.

48
00:03:24,334 --> 00:03:25,840
So there you go.

49
00:03:25,840 --> 00:03:29,220
That's where you can have one bathroom
is the second most common, and

50
00:03:29,220 --> 00:03:33,090
then you have this 1.75 bathroom,
which is probably

51
00:03:33,090 --> 00:03:37,350
going to be one full bath with a bath and
a bathroom with a shower.

52
00:03:37,350 --> 00:03:39,160
And you can see the distribution here.

53
00:03:40,475 --> 00:03:44,148
Similarly for other things,
like square foot of living, and

54
00:03:44,148 --> 00:03:50,990
number of floors,
most houses have one floor, some have two.

55
00:03:50,990 --> 00:03:52,620
And then the ZIP code.

56
00:03:52,620 --> 00:03:55,990
The most common zipcode is 98103,

57
00:03:55,990 --> 00:04:01,170
which is an area where a lot
of people live in Seattle.

58
00:04:01,170 --> 00:04:02,080
Okay, so

59
00:04:02,080 --> 00:04:07,230
we've seen a high level visualization
of the different columns of the data.

60
00:04:07,230 --> 00:04:09,140
Now let's look at some other
relationships of the data.

61
00:04:09,140 --> 00:04:11,720
Let's do some fun visualizations here.

62
00:04:11,720 --> 00:04:15,745
So I'm gonna take the sales table and
I'm gonna type show.

63
00:04:15,745 --> 00:04:20,460
But the view that I'm gonna do,
it's not gonna be scatterplot.

64
00:04:20,460 --> 00:04:25,171
But it's gonna be what is
called a box whisker plot, and

65
00:04:25,171 --> 00:04:31,940
this box whisker plot is gonna relate
two variables that we looked at.

66
00:04:31,940 --> 00:04:37,560
On the x side, I'm gonna use the ZIP code,
so this is the postal code

67
00:04:37,560 --> 00:04:44,306
that we discussed, and on the y-axis,
I'm going to plot the price.

68
00:04:44,306 --> 00:04:48,810
So what we're gonna see is
the relationship between the location,

69
00:04:48,810 --> 00:04:51,780
the ZIP code, where the house is,
and the price.

70
00:04:51,780 --> 00:04:55,130
And we're gonna see that with
what's called a box whisker plot.

71
00:04:55,130 --> 00:04:59,240
So I'm gonna press Shift+Enter and
we're gonna plot it.

72
00:04:59,240 --> 00:05:00,194
And here's what we'll see.

73
00:05:00,194 --> 00:05:03,955
So you will see, for example,

74
00:05:03,955 --> 00:05:08,296
this area's zip code, post code,

75
00:05:08,296 --> 00:05:14,770
98003, has a significantly lower price.

76
00:05:14,770 --> 00:05:19,785
So the average price is low, that's the
red line, and not a lot of variability.

77
00:05:19,785 --> 00:05:24,780
While this other zip code 98004,
so this is 003,

78
00:05:24,780 --> 00:05:28,810
98004, has a highest average price,
much higher,

79
00:05:28,810 --> 00:05:33,620
1 point something million,
1.1 million, and a huge variability.

80
00:05:33,620 --> 00:05:41,068
So the houses range from something like,
what is this?

81
00:05:41,068 --> 00:05:44,970
$800,000, it's almost $4 million.

82
00:05:44,970 --> 00:05:46,790
But here I'm just showing a few ZIP codes.

83
00:05:46,790 --> 00:05:51,030
If I drag down here,
you'll see more and more ZIP codes.

84
00:05:52,030 --> 00:05:55,430
And wow, what is this one over here?

85
00:05:55,430 --> 00:05:59,320
There's one that's astronomical,
it goes out of the scale.

86
00:05:59,320 --> 00:06:04,794
Some houses here cost like, what is this?

87
00:06:04,794 --> 00:06:07,990
$7 million or something.

88
00:06:07,990 --> 00:06:12,547
And this post code is 98039.

89
00:06:12,547 --> 00:06:18,053
Remember 98039, we'll come back to
it at the end of this notebook.

90
00:06:18,053 --> 00:06:19,011
It's kind of funny.

91
00:06:19,011 --> 00:06:24,099
[MUSIC]