These are chat archives for FreeCodeCamp/DataScience
discussion on how we can use statistical methods to measure and improve the efficacy of http://freeCodeCamp.com
@evaristoc maybe the sample (to apply either of what's described below) could be reduced by finding what 'real' progress (not necessarily an average) is -i.e., somebody completes the entire course in a week should raise a flag
Who is REALLY interested in (simple!) fraud detection algorithms? There is a possibility that FCC would implement an academic honesty enforcement workflow policy.
---I have suggested similarity analysis, but because an exhaustive comparison could be expensive, it would be better to implement an heuristic. I don't remember which the basic is, I think it had something to do with naive Bayesian, I am not sure.
@qmikew1 although the core team is managing time of completion as a variable, I also agree with @jameswinegar that time is not necessarily the best measure. Also as he mentions we must be wise when to use these techniques and what for. So far the people who have been claiming certifications have been honest with their work. It is possible that the experiment we are going into is more for making the evaluation more efficient.
@qmikew1 I am adding to @jameswinegar post above that the idea is to keep it simple for the time being, something that allows a first scan under certain circunstances. @ChristianBertram is interested in working on a first approach. If @luishendrix is still interested in making an sketch of a comparison tool, perhaps Christian can join us in preparing the exercise.
Again: it is demonstrative. It is usually how it works: we make a demo, and ask for feedback.