1
00:00:00,000 --> 00:00:04,122
[MUSIC]

2
00:00:04,122 --> 00:00:08,454
So we seen this one example, and
now let's look at a second house.

3
00:00:08,454 --> 00:00:17,230
So I'm going to now prediction for

4
00:00:17,230 --> 00:00:22,110
a second fancier house.

5
00:00:24,720 --> 00:00:26,170
So let's look at that.

6
00:00:26,170 --> 00:00:28,584
So let's call this house2.

7
00:00:28,584 --> 00:00:34,758
So, house2 from our sales data,
sales, is going to be the one,

8
00:00:34,758 --> 00:00:38,821
again, we're going to use a filter here.

9
00:00:38,821 --> 00:00:43,008
So it's the one

10
00:00:43,008 --> 00:00:48,162
whose id is equal to

11
00:00:48,162 --> 00:00:55,254
1925069082.

12
00:00:55,254 --> 00:00:59,940
So let's take a look at
what house2 looks like.

13
00:01:01,390 --> 00:01:05,843
House2 was sold in 2015 for $2.2 million.

14
00:01:05,843 --> 00:01:10,378
It has 5 bedrooms, 4 and
a quarter bathrooms, many more bathrooms,

15
00:01:10,378 --> 00:01:14,474
4,640 square feet of living space,
so it's much bigger.

16
00:01:14,474 --> 00:01:18,521
It's about 460 square meters.

17
00:01:18,521 --> 00:01:20,181
It's a pretty big house.

18
00:01:20,181 --> 00:01:24,258
So let's look at an image.

19
00:01:24,258 --> 00:01:27,930
I kinda downloaded also,
an image for this house.

20
00:01:27,930 --> 00:01:31,470
So here's my little cheat sheet for

21
00:01:31,470 --> 00:01:36,910
the image, and I'm going to insert it.

22
00:01:36,910 --> 00:01:42,400
So scan, so as before we're
gonna do an image whose source.

23
00:01:42,400 --> 00:01:45,767
So, this I downloaded to
the basic directory here.

24
00:01:45,767 --> 00:01:48,343
It's called

25
00:01:48,343 --> 00:01:57,945
house-1925069082.jog, close,

26
00:01:57,945 --> 00:02:01,930
not jog, but jpeg.

27
00:02:03,370 --> 00:02:04,490
So, that should work.

28
00:02:04,490 --> 00:02:08,534
So, remember that other
house cost $620,000.

29
00:02:08,534 --> 00:02:11,088
This one costs $2.2 million.

30
00:02:11,088 --> 00:02:12,880
Let's take a look at it.

31
00:02:12,880 --> 00:02:13,700
It looks fancier.

32
00:02:13,700 --> 00:02:15,170
It looks cool, has a bigger yard.

33
00:02:15,170 --> 00:02:16,440
But check this out.

34
00:02:16,440 --> 00:02:17,980
It's on the water.

35
00:02:17,980 --> 00:02:19,970
It's a waterfront house.

36
00:02:19,970 --> 00:02:22,630
Now that explains why it costs $2 million.

37
00:02:22,630 --> 00:02:24,830
It's fancier, much fancier.

38
00:02:24,830 --> 00:02:30,230
So, let's see now what our models
predict for the value of this property.

39
00:02:30,230 --> 00:02:32,510
So, it costs $2.2 million.

40
00:02:32,510 --> 00:02:37,190
Let's print what the square
foot model predicts.

41
00:02:38,200 --> 00:02:44,620
So, just like before,
we're going to do .predict on house2.

42
00:02:44,620 --> 00:02:48,287
And, it predicts it costs
only $1.25 million.

43
00:02:48,287 --> 00:02:50,621
So, it didn't do so well.

44
00:02:50,621 --> 00:02:56,717
So let's look at what
the more advanced models,

45
00:02:56,717 --> 00:03:01,790
so my_features model.predict does.

46
00:03:01,790 --> 00:03:03,790
And before I hit Enter here,

47
00:03:03,790 --> 00:03:07,140
it's good to start thinking
about what should happen here.

48
00:03:07,140 --> 00:03:10,861
The other house, the one above,
where the, the one with more features,

49
00:03:10,861 --> 00:03:14,704
more than one features didn't make
a big difference or actually did worse,

50
00:03:14,704 --> 00:03:16,822
was a pretty standard house for Seattle.

51
00:03:16,822 --> 00:03:19,843
Standard number of bedrooms,
standard number of bathrooms,

52
00:03:19,843 --> 00:03:23,510
it was kind of common, so you expected
both models to do about the same.

53
00:03:23,510 --> 00:03:24,560
But this house is crazy.

54
00:03:24,560 --> 00:03:27,580
It has features which are hard to
capture just because of the square feet.

55
00:03:27,580 --> 00:03:28,301
It has waterfront.

56
00:03:28,301 --> 00:03:30,182
It has lots of bathrooms,
lots of bedrooms.

57
00:03:30,182 --> 00:03:36,620
So we expect here for the more
advanced feature model to do better.

58
00:03:36,620 --> 00:03:38,460
Oh, I misspelled something.

59
00:03:39,650 --> 00:03:42,300
Oh, I forgot, here my features_model.

60
00:03:42,300 --> 00:03:44,044
Sorry about that,
that was really anticlimactic.

61
00:03:44,044 --> 00:03:49,328
But if we run it,
you see it's 1.38 million.

62
00:03:49,328 --> 00:03:50,831
It's a little closer.

63
00:03:50,831 --> 00:03:53,640
As you remember from before,
the error difference was too big, but

64
00:03:53,640 --> 00:03:55,522
here it's a little closer
to the true price.

65
00:03:55,522 --> 00:03:57,892
So it does a bit better on this house.

66
00:04:01,133 --> 00:04:04,518
Now, just to conclude,

67
00:04:04,518 --> 00:04:09,193
let's do a even fancier house, so

68
00:04:09,193 --> 00:04:15,010
what I'm gonna look at, so a third house.

69
00:04:19,352 --> 00:04:24,706
So the ##Last house is
gonna be super fancy.

70
00:04:24,706 --> 00:04:30,083
And in fact, what we're gonna
do is take Bill Gates' house.

71
00:04:30,083 --> 00:04:36,360
Bill Gates lives in the Seattle area, and
try to predict what the house looks like.

72
00:04:36,360 --> 00:04:41,400
So it has a lot of different properties.

73
00:04:41,400 --> 00:04:46,910
And rather than typing them all in,
I'm going to paste them in.

74
00:04:48,540 --> 00:04:49,810
So I'm going to paste it here.

75
00:04:51,040 --> 00:04:55,700
And this is what your Gates house
would have if it were in the data set.

76
00:04:55,700 --> 00:04:59,350
It has 8 bedrooms, 25 bathrooms.

77
00:04:59,350 --> 00:05:01,868
Bill Gates really needs to go
to the bathroom, apparently.

78
00:05:01,868 --> 00:05:04,628
25 bathrooms,
at least this is what it said online.

79
00:05:04,628 --> 00:05:08,150
It has 50,000 square feet of housing.

80
00:05:08,150 --> 00:05:10,419
That's about 5,000 square meters.

81
00:05:10,419 --> 00:05:14,459
That's a big house,
four floors, lots of stuff.

82
00:05:14,459 --> 00:05:16,420
So that's what his house looks like.

83
00:05:16,420 --> 00:05:19,500
We actually don't know what the house
is worth because it's never been sold.

84
00:05:19,500 --> 00:05:21,137
I guess he doesn't need the money.

85
00:05:21,137 --> 00:05:25,952
[LAUGH] And let's see what that

86
00:05:25,952 --> 00:05:30,590
house actually looks like.

87
00:05:30,590 --> 00:05:38,021
So, I actually have a little cheat sheet
here with a link to an image of his house.

88
00:05:38,021 --> 00:05:44,860
So, this is a picture of,
Bill Gates' house.

89
00:05:50,595 --> 00:05:52,080
And here we go.

90
00:05:52,080 --> 00:05:55,490
Oh, yes, sorry about that.

91
00:05:55,490 --> 00:05:59,170
I forgot I had to do,
tell it it was a markup.

92
00:06:01,020 --> 00:06:02,460
And there we go.

93
00:06:02,460 --> 00:06:08,799
So, let's see what the model predicts for
the price of Bill Gates' house.

94
00:06:08,799 --> 00:06:15,735
I'm gonna print what
my_features_model.predict says for

95
00:06:15,735 --> 00:06:19,020
Bill Gates' house.

96
00:06:19,020 --> 00:06:24,293
Now these are GraphLab created models,
and they only take SFrames as input.

97
00:06:24,293 --> 00:06:28,676
So I have to take that dictionary
that I just created to describe

98
00:06:28,676 --> 00:06:32,160
Bill Gates' house, and
convert it to an SFrame.

99
00:06:32,160 --> 00:06:36,928
So I just type
graphlab.SFrame(bill_gates), this was

100
00:06:36,928 --> 00:06:43,197
a dictionary, which is what we defined
above here that's called dictionary.

101
00:06:43,197 --> 00:06:46,409
And then converts to SFrame,
and I execute.

102
00:06:46,409 --> 00:06:52,911
And it says it predicts it
to be valued at $13 million.

103
00:06:52,911 --> 00:06:53,968
I don't know.

104
00:06:53,968 --> 00:06:57,710
It sounds like a lot, but the house might
be worth a lot more than $13 million.

105
00:06:57,710 --> 00:06:58,210
Who knows?

106
00:07:00,180 --> 00:07:01,360
And that was exciting.

107
00:07:01,360 --> 00:07:05,707
We built two models of house
prices using county data.

108
00:07:05,707 --> 00:07:10,720
We explored it, we use it, and we applied
it to this task of predicting houses.

109
00:07:10,720 --> 00:07:15,793
We even found out what we think Bill
Gates' house might actually be valued at.

110
00:07:15,793 --> 00:07:20,199
[MUSIC]