Questions d'entretien de Data scientist partagées par les candidats
How to reverse client churning thanks to anayltics?
I spoke about classical methods, from the simplest one (logistic regression) to deep neural networks for prediction. I add that machine learning won't bring any insight regarding the cause for churning, claimimg it should be better to build a 360°data model with semantics and graph databases: quid? what is a graph database?....
Develop an iOS App with the welcome screen asking the user to fill in an airline 2-digit code, after which the App will return the name of the airline. Note that the reference file contains airline codes and names.
Develop a webpage/web app with the welcome screen asking the user to fill in an airline 2-digit code, after which the app will return the name of the airline. Note that the reference file contains airline codes and names.
What other features (e.g. destination temperature, currency, country GDP) do you think would be interesting in this data set? Can you find sources (e.g. weather API, CIA world fact book, Wiki) for these features and add them to the data set? Show the results during the interview.
How would you store the data from the reference file if it would have been significantly larger? And with that assumption, store the data appropriately using tools, techniques and software, as you see fit. Prepare an environment that you could show on the interview.
Use the results from previous exercises, add other data sources if necessary, and if necessary add dummy generated data to ensure you have training data set and test data. With that use ML, supervised and unsupervised, and deep learning to create a predictive model and show this during the interview.