Reference no: EM133143555

You are a manager in customer service department in Amazon. Daily, your team takes care of 100,000 customer emails. You need to develop a model, that scans those emails into two general categories: complaints and approvals. Complaints must be sent to a specific department to take care of them, either by refunding the customer or giving them some promotions ...

Approvals need to be sent to another department, to try to encourage happy customers more to purchase more using a variety of techniques. Approval is labelled as positive and complaints are labelled as negative. Surely, you understand the importance of reading those emails by a machine and classifying them into correct category. Your model is not perfect, though:

For every TN, in average, we make $10, for every TP we make, $8, and for every FP we will lose $3 and for every FN we will lose $12. For your model, TN=20, TP=90, FN=15, FP=2. All numbers are in millions.

Per each email, you are paying $0.1 to your data scientist to do the prediction. Based on EV you have, shall you fire her or promote her? Options are below:

1. Fire her

2. Keep her, and promote her

3. Not enough data to decide