As a Data Scientist at a prestigious bank, your task is to develop a model that can classify transactions as either legitimate or fraudulent, based on the details of each transaction. You have been provided large-scale dataset with 100,000 samples having 90,000 legitimate and 10,000 fraudulent examples.

(a) (10 marks) You are getting a very high accuracy score both on training and test set. Can you deploy this trained model in real settings? What are the issues involved in evaluating the model based on only classication accuracy? What you will do to ensure the model is good enough to be integrated into the bank systems?

(b) (5 marks) You are concerned that your model is overtting or undertting the data. Briefly explain what these two terms mean and describe how you would nd out if your model had one

of these problems.