Applying learned models to predict price of two fancy houses

Loading...

来自 University of Washington 的课程

机器学习基础：案例研究

7805 个评分

University of Washington

机器学习基础：案例研究

7805 个评分

课程 1（共 4 门，Specialization Machine Learning）

你是否好奇数据可以告诉你什么？你是否想在关于机器学习促进商业的核心方式上有深层次的理解？你是否想能同专家们讨论关于回归，分类，深度学习以及推荐系统的一切？在这门课上，你将会通过一系列实际案例学习来获取实践经历。在这门课结束的时候，

从本节课中

Regression: Predicting House Prices

This week you will build your first intelligent application that makes predictions from data.<p>We will explore this idea within the context of our first case study, predicting house prices, where you will create models that predict a continuous value (price) from input features (square footage, number of bedrooms and bathrooms,...). <p>This is just one of the many places where regression can be applied.Other applications range from predicting health outcomes in medicine, stock prices in finance, and power usage in high-performance computing, to analyzing which regulators are important for gene expression.</p>You will also examine how to analyze the performance of your predictive model and implement regression in practice using an iPython notebook.

Loading & exploring house sale data7:11

Splitting the data into training and test sets2:34

Learning a simple regression model to predict house prices from house size3:54

Evaluating error (RMSE) of the simple model2:29

Visualizing predictions of simple model with Matplotlib4:52

Inspecting the model coefficients learned1:18

Exploring other features of the data6:24

Learning a model to predict house prices from more features3:23

Applying learned models to predict price of an average house5:07

Applying learned models to predict price of two fancy houses7:20

与讲师见面

Carlos Guestrin
Amazon Professor of Machine Learning
Computer Science and Engineering
Emily Fox
Amazon Professor of Machine Learning
Statistics

[MUSIC]

So we seen this one example, and now let's look at a second house.

So I'm going to now prediction for

a second fancier house.

So let's look at that.

So let's call this house2.

So, house2 from our sales data, sales, is going to be the one,

again, we're going to use a filter here.

So it's the one

whose id is equal to

1925069082.

So let's take a look at what house2 looks like.

House2 was sold in 2015 for $2.2 million.

It has 5 bedrooms, 4 and a quarter bathrooms, many more bathrooms,

4,640 square feet of living space, so it's much bigger.

It's about 460 square meters.

It's a pretty big house.

So let's look at an image.

I kinda downloaded also, an image for this house.

So here's my little cheat sheet for

the image, and I'm going to insert it.

So scan, so as before we're gonna do an image whose source.

So, this I downloaded to the basic directory here.

It's called

house-1925069082.jog, close,

not jog, but jpeg.

So, that should work.

So, remember that other house cost $620,000.

This one costs $2.2 million.

Let's take a look at it.

It looks fancier.

It looks cool, has a bigger yard.

But check this out.

It's on the water.

It's a waterfront house.

Now that explains why it costs $2 million.

It's fancier, much fancier.

So, let's see now what our models predict for the value of this property.

So, it costs $2.2 million.

Let's print what the square foot model predicts.

So, just like before, we're going to do .predict on house2.

And, it predicts it costs only $1.25 million.

So, it didn't do so well.

So let's look at what the more advanced models,

so my_features model.predict does.

And before I hit Enter here,

it's good to start thinking about what should happen here.

The other house, the one above, where the, the one with more features,

more than one features didn't make a big difference or actually did worse,

was a pretty standard house for Seattle.

Standard number of bedrooms, standard number of bathrooms,

it was kind of common, so you expected both models to do about the same.

But this house is crazy.

It has features which are hard to capture just because of the square feet.

It has waterfront.

It has lots of bathrooms, lots of bedrooms.

So we expect here for the more advanced feature model to do better.

Oh, I misspelled something.

Oh, I forgot, here my features_model.

Sorry about that, that was really anticlimactic.

But if we run it, you see it's 1.38 million.

It's a little closer.

As you remember from before, the error difference was too big, but

here it's a little closer to the true price.

So it does a bit better on this house.

Now, just to conclude,

let's do a even fancier house, so

what I'm gonna look at, so a third house.

So the ##Last house is gonna be super fancy.

And in fact, what we're gonna do is take Bill Gates' house.

Bill Gates lives in the Seattle area, and try to predict what the house looks like.

So it has a lot of different properties.

And rather than typing them all in, I'm going to paste them in.

So I'm going to paste it here.

And this is what your Gates house would have if it were in the data set.

It has 8 bedrooms, 25 bathrooms.

Bill Gates really needs to go to the bathroom, apparently.

25 bathrooms, at least this is what it said online.

It has 50,000 square feet of housing.

That's about 5,000 square meters.

That's a big house, four floors, lots of stuff.

So that's what his house looks like.

We actually don't know what the house is worth because it's never been sold.

I guess he doesn't need the money.

[LAUGH] And let's see what that

house actually looks like.

So, I actually have a little cheat sheet here with a link to an image of his house.

So, this is a picture of, Bill Gates' house.

And here we go.

Oh, yes, sorry about that.

I forgot I had to do, tell it it was a markup.

And there we go.

So, let's see what the model predicts for the price of Bill Gates' house.

I'm gonna print what my_features_model.predict says for

Bill Gates' house.

Now these are GraphLab created models, and they only take SFrames as input.

So I have to take that dictionary that I just created to describe

Bill Gates' house, and convert it to an SFrame.

So I just type graphlab.SFrame(bill_gates), this was

a dictionary, which is what we defined above here that's called dictionary.

And then converts to SFrame, and I execute.

And it says it predicts it to be valued at $13 million.

I don't know.

It sounds like a lot, but the house might be worth a lot more than $13 million.

Who knows?

And that was exciting.

We built two models of house prices using county data.

We explored it, we use it, and we applied it to this task of predicting houses.

We even found out what we think Bill Gates' house might actually be valued at.

[MUSIC]

探索我们的目录

免费加入并获得个性化推荐、更新和优惠。

Coursera 致力于普及全世界最好的教育，它与全球一流大学和机构合作提供在线课程。

© 2018 Coursera Inc. 保留所有权利。

通过 App Store 下载

通过 Google Play 获取