Kaggle Kernels Guide for Beginners: A Step by Step Tutorial – KDnuggets

This is an attempt to hold the hands of a complete founder and walk them through the worldly concern of Kaggle Kernels — for them to get started. By Abdul Majed Raja, Analyst at Cisco

erstwhile back, I wrote an article titled “ Show off your Data Science skills with Kaggle Kernels ” and then late realized that even though the article made a good claim on how Kaggle Kernels could be a herculean portfolio for a Data scientist, it did nothing about how a arrant founder can get started with Kaggle Kernels.

This is an attempt to hold the hands of a accomplished novice and walk them through the worldly concern of Kaggle Kernels — for them to get started .

Register on Kaggle — https://www.kaggle.com/

If you don ’ t have a Kaggle Account account, the foremost footfall is to register on Kaggle. You can either use your Google Account or Facebook Account to create your raw Kaggle report and log in. If none of the above, you can enter your e-mail id and your prefer password and create your new score .

Logging in into Kaggle

If you have an account already or you just created one, Click the sign in release on the top-right corner of the page to initiate the login march. Again, you ’ ll be given an option to login with Google / Facebook / Yahoo or the last one, with the drug user name password that you entered while creating your account .

Kaggle Dashboard
After you login, You ’ ll be taken to the Kaggle Dashboard. ( It ’ s equitable the welcome page, I don ’ metric ton know what else to call, so called it a Dashboard ) .

This is how your down page appears immediately after you login ( if you had logged in from hypertext transfer protocol : //www.kaggle.com/ ). It has many components, few of them :

  • A feed of Kaggle Kernels that are recently updated or recommended to you by Kaggle
  • Profile summary (first of right sidebar)
  • Job Ad (right sidebar)
  • Your competitions (right sidebar — after scrolling down)
  • Your Kernels (right sidebar — after scrolling down)

Where we are heading adjacent is the top Kernels button in the seafaring bar .

Kaggle Kernels List (Hottest):

Once we click the top Kernels clitoris from any stead of the Kaggle Journey, we ’ ll bring on this screen .

This is the screen where everyone tries to see their Kernel because this is like the Front Page of Kernels which means your Kernel has more likelihood of getting a draw more visibility if it ends up here. The nonpayment kind order in the Kernels foliate is Hotness which is based on Kaggle ’ second privy sauce Algorithm to keep showing relevant Kernels but it besides has got other options ( to sort ) like New, Most Votes and so on. Kaggle besides uses this page to advertise if there ’ s any Kernel Contest happening / going to happen .
While we are hera, A Kernel Contest is a Kaggle Competition which doesn ’ metric ton fall under the Competition grade because of the nature of the contest where the output is a Kaggle Kernel and more much focused on Storytelling. Data Science for Good is one such series of Kernel contests where the Data Scientist / Kaggler is expected to help in a Social Problem ( for good ) using Data Science. For understanding more of it, You should check out the Kernels of Kernel Grandmaster Shivam Bansal who ’ s made a habit of winning them so many times .

Kaggle Kernels — New / Creation:

now, that we ’ ve understood the meta of Kaggle Kernels, we can jump veracious into universe of New Kernels. There are two primary ways a Kaggle Kernel can be created :

  1. From the Kaggle Kernels (front page) using New Kernel Button
  2. From a Dataset Page using New Kernel Button

Method #1: From the Kaggle Kernels (front page) using New Kernel Button
As you can see in the above screenshot, Clicking the New Kernel release from the Kernels foliate would enable you create a new Kernel. This method is estimable if you are trying to drill something of your own or you ’ re plan to input your own dataset. This method acting international relations and security network ’ thymine advisable ( in my opinion ), if you want to create a Kernel for a dataset that ’ s already existing on Kaggle .
Method #2: From a Dataset Page using New Kernel Button
This is one of the most popularly used method ( at least by me ) for creating new Kernels. You can open the dataset foliate of the dataset of your sake ( like the one in the screenshot below ) and then click New Kernel button in there. The advantage with this method is that unlike the Method # 1, in this method # 2 the Kaggle Dataset from which the Kernel is created comes attached with the Kernel ( by nonpayment ) frankincense making this bore march of inputting a dataset to your kernel easier, faster and aboveboard .

Kaggle Kernels — Kernel Type:

Irrespective of Method # 1 or # 2, once you click the New Kernel, you ’ ll be presented with this modal auxiliary verb shield to select the type of Kaggle Kernel you ’ d like to create .

broadly it ’ sulfur two categories — 1. Script vs 2. notebook.

As we all know about a Notebook ( cell-based layout ), It ’ second fair what a Jupyter Notebook is and the Script is what you ’ five hundred code on probably — Pycharm or Sublime Text or RStudio. additionally for R users, the handwriting is the Kernel type for RMarkdown — the beautiful way to programmatically generate a report from R .
To summarize the types of Kernels :

  • Script
    * Python
    * R
    * RMarkdown
  • Notebook

Kaggle Kernels — Kernel Language:

This second level of Kernel Language survival happens only after the first charge of Kernel Type Selection .
Kaggle GIF
As in the above GIF of a Kaggle Kernel of Type Script, The speech of the Kernel can be changed by going into Settings and then selecting desired Language — R / Py / RMarkdown. The lapp settings besides provide option to make your Kernel Sharing Public ( which by default is Private unless made Public ). individual Kernels are normally used if you ’ rhenium working on your university assignment or self-learning where you didn ’ t want to reveal the code. secret Kernels are besides used by Kagglers who participate in rival to leverage Kaggle ’ s calculation power but not reveal their code / approach .

Notebook Kernel:

Similar to the above GIF, where Kernel Type Script is selected, You can besides select Notebook to create a Notebook Kernel .

RMarkdown Kernel — (Kernel Type: Script > RMarkdown)

RMarkdown uses a combination of R and Markdown in generating analytic Reports with synergistic visualizations embedded on it. While this is the most simplistic way of explaining what ’ s RMarkdown, its uses and potential turn much far and beyond the definition .
fortunately, Kaggle Kernel Script supports Rmarkdown which means it can help create interactional documentation and much more that wouldn ’ thymine be possible in a Notebook-based scenario. here ’ s a full-fledged Interactive Dashboard built on Kaggle Kernel by Saba Tavoosi, that illustrates the electric potential of Kaggle Kernels not just for building Machine Learning models but besides for interactional storytelling at its best form. Checkout this course, If you ’ re concerned in learning how to build dashboards with flexdashboard .

Copy and Edit (formerly, Forking)

Similar to the Fork choice in Github, If you would like to take an existing Kaggle Kernel and use it in your own distance — to modify or give your own touch — you ’ five hundred need to use the top right blasphemous button Copy and Edit. In fact, In a lot of Machine Learning competitions on Kaggle Competitions track, many high scoring populace kernels are normally forks of forks forks where one Kaggler would improve upon the model that was already built by some other Kaggler and made them available as a Public Kernel .

Public / Private Kernel

As we saw above in another section, Access setting of a Kaggle Kernel can be either Public or Private. A Public Kernel ( as obviously the name suggests ) is available and visible for everyone ( including Kagglers and Non-Kagglers ). A private Kernel is available for only the owner ( one who created it ) and those with whom the owner shared the Kernel with. A populace Kernel can be besides built on Private Dataset. Let ’ s say, It ’ s a Machine Learning contest and you ’ ve done some feature engineer with some 3rd Party data and you wouldn ’ thymine want to reveal the data during the period of the competition. This is a typical scenario where Kagglers normally keep their dataset private, so far make the Kernel a Public one so others can see their approach and learn from it .

The screenshot above illustrates how an existing Kernel ’ randomness access determine could be changed to either Private or Public. All newly created Kernels are Private by default ( at this meter of writing ) and the owner then changes it Public if required .

TL;DR — How to create a New Kaggle Kernel

If everything above seemed a spot excessively heavy to grasp at the first glance, this is the section to help you with creating your first base Kaggle Kernel .

Steps :

  1. Login to Kaggle using your Credentials
  2. Go to any Public Kaggle Dataset
  3. Click New Kernel on the top right (blue-colored button)
  4. Select Notebook/Script of your interest
  5. If Python is your language of choice leave it as it’s, If R, then go to the Settings at the right side and click to expand the items where you can see Python next to the Language which you can click to change to R
  6. Go to the Editor section / Pane (left-side) of the screen and Start writing your beautiful code (the above GIF also illustrates how you can use the dataset from where you created the Kernel)
  7. Once your code is complete, click Commit on the top right (blue-colored button)
  8. If your Kernel execution is successful (without any errors), Make your Kernel Public (either by editing the Kernel Settings > Sharing (Public)or by opening the Kernel again and clicking the Access button on the top)
  9. At this stage, your first Kaggle Kernel must be ready for being shared with your friends across your network!

Check out this Kaggle Video for aid .


For a lot of Kagglers, Competition Track has been their fun ride but for me, Kaggle Kernels Track has been my matter which gives us the enormous electric potential of completing a full-stack datum skill journey from Data Prep to Data Visualization — Machine Learning Modelling to Storytelling. Hope you ’ d like it excessively. good Luck on your Kaggle Kernel Journey .
Check out my Kaggle Kernels at My Kaggle Profile and share your feedback with me at My Linkedin Profile. The videos/GIFs/screenshots used in this tutorial is available on my github .

Bio: AbdulMajedRaja is an analyst at Cisco .
Original. Reposted with permission .

source : https://coinselected
Category : coin 4u

Leave a Reply

Your email address will not be published.