Read the Uber case and other readings (linked in the syllabus) for the session a
Read the Uber case and other readings (linked in the syllabus) for the session and answer the following questions:
1) Develop a list of hypotheses Uber could use to predict a rider’s pickup location with information such as the rider’s previous trips and current destination, as well as historical patterns related to the pickup location. Augment the case information with your personal Uber experiences to suggest potential hypotheses.
2) Create a quantitative pickup quality metric using attributes derived from the passive, active and third-party signals available to Uber. Discuss why your selected attributes represent a robust pickup quality metric. What weights would you assign to the features you choose for your pickup model?
3) Based on your pickup quality metric, what actions can Uber operators take to improve the pickup experience?
4) Discuss the steps involved in setting up an ML model for automating pickups at scale. Use the framework of the seven-step model in the case (Exhibit 7) to elaborate on how Uber should apply this framework to the ML model. Hint: you can create a table and list the tasks under each step.
5) Is there a role for unstructured data for the issues we have examined in this case? For which prediction problem is it likely to make the biggest impact and for which is it likely to make the least impact? Explain why.
6) Where is human domain knowledge likely to be helpful for the prediction problems above? Where do you think such knowledge is unlikely to be helpful? Explain.