siesgstarena / labs Goto Github PK

Idea for recommendation:
The recommendation will be independent of the type of problem, it will be completely dependent on the language in which the user has solved the previous problem.
The recommendation system (success score) is based on formula:- 1/(1+TS-PS);
where TS=successful submissions/total submissions,
PS=successful submissions for a particular language/total submissions for particular language
The dataset received in json form is converted into csv format first. The only columns useful here are the column of problem_id and the submission language. Then the formula decided is applied over the data, then the data needs to be standardized, a new csv of success score is generated.
How actual recommendation will work:
For training the model ,KNN algorithm is used.
Based on the calculated success score for a particular problem, the user will be recommended a problem having a similar success score. So the user will be recommended with problems having a similar success score to that of the previous problem solved based on language used, irrespective of type of problem.

Short description
#2 work has has motivated us to bring in a structure for future contributions. Thank you to author of #2 for that :)

I would suggest you to create a README template for all projects which follow some basic sections so that all experiments are easier to follow e.g. README

# Title

## About

## Setup 

### Installation  if any 

### Training/Testing/Running if any 

## Results/Visualizations

## Acknowledgements if any 

## Licenses if any

Now this README template can have more sections which are common to many people, please feel free to add them and create a final README template. You will find many references on GitHub.

Expected behavior
Any new contributor when raising a PR should have the README with well-written guide for using new experiments.

Additional context
Not a bug, but raising as a bug so that it can be fixed quickly as its blocking some PRs.

Is your experiment related to a problem? Please describe.
Taking up a new problem based on just the problem statement is difficult. The framing can be confusing and may not be conducive to you reaching the solution. This relies on a number of factors :

• Type of the problem (string manipulation, math, logical).
• Language dependency(some problems can be comparatively easier to solve with language specific libraries/functions).
• Liking for that problem type(you may not like graph based problems and like array handling and sequencing).
• Performance on similarly solved previous problems.

Describe the how to are going to take the experiment forward
I had in mind a prediction engine, that shows a score of 'how likely' one is to successfully solve that problem based on the users' previous results for similar problems.
Every problem will have a Success Score tailored based on the users' performances in previous problem sets. This can be a better judgement factor and will statistically increase success rates.