Machine Learning (Introduction)

Which model is the best choice for fraud detection?

• Precision = 0.8, Recall = 0.4

• Precision = 0.4, Recall = 0.8

• Precision = 0.7, Recall = 0.7

For the application of telesales, suppose we have a million customers in database, a company want to make a phone call for offering special privileges to customers who are likely to purchase. Calling to all of them is impossible. Which measures are the most important to identify possible customers?

• Precision

• Recall

• Sensitivity

• Specificity

What is the concept of reinforcement learning?

• Learning from labeled data

• Learning from unlabeled data

• Learning from environment

Why do we need to set hyper-parameters?

• Model configuration

• Model validation

• Model deployment

What is the correct for the red dot for the application of diabetes prediction?

• Setting threshold = 0, All samples have diabetes

• Setting threshold = 1, All samples do not have diabetes

• Setting threshold = 0, All samples do not have diabetes

• Setting threshold = 1, All samples have diabetes

What is the correct for the red dot for the application of diabetes prediction?

• Setting threshold = 0, All samples have diabetes

• Setting threshold = 1, All samples do not have diabetes

• Setting threshold = 0, All samples do not have diabetes

• Setting threshold = 1, All samples have diabetes

What is True (You may have multiple answer)

• Training accuracy = 0.90, Testing accuracy = 0.89, your model is good enough to deploy for diabetes prediction

• Area under curve of ROC is 0.9, your model is good enough to deploy for lung cancer prediction

• Training accuracy = 0.90, Testing accuracy = 0.89, your model is not overfit

• Area under curve of PR Curve is 0.9, your model is good enough to deploy for lung cancer prediction

Which model is good enough for fraud detection?

• Precision = 0.9, Recall = 0.9

• Precision = 0.9, Recall = 0.6

• Precision = 0.6, Recall = 0.9

• Precision = 0.6, Recall = 0.6

What is limitation of over sampling?

• It could make model underfit

• It could make model overfit

What is limitation of under sampling

• It could make model underfit

• It could make model overfit