Pergunta de entrevista da empresa Lyft
Predict a passenger will be churned or not, given some transaction data. Like a simple Kaggle data science competition, but much vague and no clear definition. Advantages and limitations of two classifiers.
Resposta da entrevista
Sigiloso
I thought about this problem from the perspective of sequential modeling with models like HMM or RNN, but the interviewer seems only know the simple classifier.... and if you cannot answer and clarify the problem from the perspective of simple classifier, he may feel you are wrong and give you the solution directly. For example, if you think of it from the perspective of RNN or transformer, then a lot of attributes could be used such as passenger ID, fare, prime rate.... however, if you think it as a simple classifier, many attributes will not be useful...