Here's how you can create a Data Science Problem -
Go to your dashboard and click on the (More Options) button next to your profile.
Select the Library option to view, edit, and create questions in the library.
- To create a new problem, click on the blue button ( ). This will open up a modal for you.
- In the opened modal, enter a name for the problem (see image below) and select the Data Science option from the drop-down list under problem type.
- Assign a difficulty level for your problem. (Easy, Medium, Hard).
- Click on the CREATE button to go ahead.
- Define your problem in the description section and follow the checklist for your guidance.
- Next, you can select the allowed programming language(s) for the candidates to attempt the problem.
Note: Only languages R and Python can be used.
- To organise your problems better, make use of Tags. (E.g. Data Science tag can be used to organise all the Data Science problems which you have created).
- You need to provide three datasets for every problem. You should upload the following datasets as zip archives -
This dataset will be available to the developers for training their model.
It will available at
This dataset will be used to test the developer's solution. It will be available at
This dataset will be available to the developers for validating their modal before they submit for evaluation. It will also be available at
/data/test/will contain the sample datasets when the developer runs sample test cases.
- Click on the ADD NEW button in test cases section. This will open a new modal.
Test cases are programs which will be used to evaluate your developers' submissions. You have the freedom to write the test cases in
RThe correctness of solution is determined by the exit status of the evaluation code. Evaluation code should exit with a zero only if the submission is correct.
Sample test cases are used for self-validation only and will not be used for final evaluation.
The evaluation code should write error metrics to the File
SSE 2.45 RMSE 0.7
If you need to show plots in the reports, you can write the plots to
output = read.csv("/code/prediction.csv") png(filename = "/code/output/1.png") plot(output$estimate) dev.off().
Once the test cases have been written, click on the ADD TESTCASE button.
Note: To attach a boilerplate code for a solution, you can click on the ADD NEW button in the stub section.
- Finally, click on SAVE PROBLEM to save all your hard work. :)
- Once you have saved a problem, a blue tick mark ( ) appears next to it. Your problem is now ready for use.