How many Non-frauds would be blocked to stop one Fraud based on the given confusion matrix?

Question

One of the questions in a recent online test that I couldn't answer is below(I couldn't copy paste exact question, so this one is totally from memory. Kindly excuse in case of poor wording.)

We have a fraud classification model whose confusion matrix is shown below. How many non-frauds would the model block to stop one fraud transaction?

Answer options -

1. 19
2. 5
3. 3.5
4. 1.5

I got a bit confused with this and guessed 19. But I request anyone to kindly explain in case you know the answer. Thanks in Advance.

Answer 1

Answer from Bard -

The confusion matrix shows the following:

There were 200 actual fraud transactions, and the model correctly predicted 200 of them as fraud (true positives). There were also 300 actual non-fraud transactions, and the model incorrectly predicted 300 of them as fraud (false positives). To stop one fraud transaction, the model will block 5 non-fraud transactions. This is calculated as follows:

Number of non-frauds blocked / Number of frauds stopped = 300 / 200 = 1.5 The answer choices are:

19: This is the number of non-frauds blocked if the model predicts all transactions as fraud. This is not a good measure, because it will block a lot of legitimate transactions. 5: This is the correct answer. It is the number of non-frauds blocked per one fraud stopped. 3.5: This is the average of 19 and 5. 1.5: This is the number of non-frauds blocked per one actual fraud. This is not a good measure, because it does not take into account the false positives.