Research case studies
You are asked to "predict demand" for a bike-share operator. What do you say first?
- AThat you would start with gradient boosting on the historical data
- BWhat decision this supports, and what exactly is being predicted
- CThat you would need at least three years of history first
- DThat demand forecasting is a well-studied problem already
Try it
Pick an answer
Next free question: Which statement about the central limit theorem is correct?
Show the answer and worked solution
Answer: B – What decision this supports, and what exactly is being predicted
Ask what decision the forecast supports, then define the target precisely. Hires per station per hour informs a dispatcher moving bikes around tonight; city-wide daily demand informs a planner deciding how many bikes to buy. They differ in horizon, granularity, loss function and data, so nearly every downstream choice follows from settling this. Naming a model first signals that you brought a favourite tool rather than analysed the problem, and it is the single most common way candidates lose a case in the opening minute.
- A. A model choice before the target is defined tells an interviewer you have a favourite tool.
- B. Correct. Hires per station per hour for a dispatcher is a different problem from city-wide daily demand for a planner.
- C. Possibly true, but it is a data question and it comes after the target.
- D. True and unhelpful.
Takeaway: Define the objective and the target before choosing a model.
Report an issue with this question
Learn the method
Of the kind reported at
More research case studies questions
- Why does an interviewer ask for a baseline before you describe your model?Applied · Free solution
- Your evaluation metric is mean absolute error and the target is heavily skewed.Foundation · Premium
- You forecast demand to decide inventory, where being short costs four times as…Applied · Premium
- You are asked to predict house prices, which range from tens of thousands to…Applied · Premium
- A target takes values 1,2,3,4,100.Applied · Premium
- How many independent observations are needed to detect an information…Advanced · Premium