[MUSIC] Hi, everyone. In this video we'll learn
how to use Kaggle for participation in data
science competitions. Let's open kaggle.com. On the Competitions page, we can see
a list of currently running competitions. Every competition has a page which
consists of title, short description, price budget, number of participating
teams, and time before the end. Information involves all
previously running competitions, we can find if we click to All. Let's select some challenge and
see how it organized. Here, we see several tabs which we'll
explore, and let's start with Overview. In the Description section we see
an introduction provided by organizers. In the Description, there is a short
story about company and tasks, sometimes with illustration. At the Evaluation page, we see
the description of the target metric. In this challenge, target metric
is the Mean Absolute Error between the logarithmic transform predictions and
ground truth values. This page also contains example of sample
submission file, which is typical for such kind of competitions. Now let's move to the Prize page. In the Prize,
page we can find information about prizes. Take notice that in the title we have
information about the whole money budget, and this page,
we see how it will be split among winners. I want to highlight that
in order to get money, you need not only be in top three teams,
but also beat a Zillow benchmark model. Now let's see, Timeline page, which
contains all the information about dates. For example,
when competition starts, ends, when will the Team Merger deadline and
then what month. All the details about competition,
we can find in the Rules. So we need to check really the rules. Here we can find that team
limit is three individual, that we have maximum of five
submissions per day, that you, for example, should be at least
18 years old to participate. And that, find it, that external data are not allowed. I strongly suggest you to read the rules
carefully before joining the competition. And after reading, you should accept it,
but I already accepted it. Now, let's check this, Data. Here we have data provided by
the organizers, several files which we can download, and sample submission among
them, and the description of the data. Here we have description of files,
description of data fields, and more importantly a description
of train and test split. This is quite useful information in
order to set up right validation scheme. If you have any question about data or
other questions to ask, or insights to share, you can go to the forum,
which we can find under Discussion tab. Usually it contain a lot of topics or
threads, like Welcome, questions about validations, questions about train and
test data, and so on and so on. Every topic have title,
number of comments, and number of reports. Let's see some of them. Here we have main message,
a lot of comments, in this particular we
have only one comments. Each we can up vote or down vote and
reply to by click the reply button. That was a brief overview on forum and
now we switch to the Kernels. Usually, I run my code locally,
but sometimes it would be handy to check an idea quickly or share code
with other participants or teammates. This is what Kernels are for. You can think of Kernel as a small virtual
machine in which you write your code, execute it, and share it. Let's take a look at some Kernel,
for example for this one. This show explanatory data analysis
on the Zillow competition. It took quite long, contain a lot of
pictures, and I believe it very useful. Here we can see comments for
this, different versions. And in order,
if you want to make a copy and edit it, we need to Fork this Notebook. It doesn't matter how your
predictions were produced, locally or by Kernel, you should submit
them through a specialized form. So go back to the competition. Go to submissions. I already submit sample submission,
you can do the same. Click submit predictions,
and drag and drop file here. Let's look at my submission. After submission,
you will see it on the leaderboard. This is my sample submission. Leaderboard contains information
about all the teams. So here we have team name or just name
in case of single competition team. Score which we produced,
number of submissions, time since the last submissions, and
position data over seven last days. For example, this means that this guy
drops 19 positions during the last week. That was a brief overview
of Kaggle interface. Further, I will tell some extra
information about the platform. So let's move to Overview
page at the bottom. And here,
we see information about points and tiers. As mentioned here, the competition will be
counting towards ranking points an tiers. If you participate,
it will be beneficial for your rating. Sometimes, especially in educational
competitions, it's not like that. Information about Kaggle Progression
System we can find if we click this link, where we can read information about
tiers like novice, contributor, master, grandmaster. About medals and ranking points. This ranking points, I use for
global User Ranking. Let's check it. So, we have user ranking page, and we see all the users ranked, and
with links to their profile. Let's check some profile,
for example mine. And here we have photo,
name, some information, geo information, information about
past competitions, medals, and so on. In addition, I want to say a few words
about ability to host competition. Kaggle has this ability. Click Host competition, and
there is special Kaggle in class. At in class, everyone can host
their own competition for free and invite people to participate. This option is quite often used in
various educational competitions. So this was a brief overview
of Kaggle platform. Thank for your attention. [MUSIC]