Given:
0.1% of all transactions are fraudulent
99% correct identification whether a transaction is fraudulent or not.
Scanned 5,000,000 transactions.
5,000,000 x 0.1% = 5,000 fraudulent transactions.
For me, there are 5,000 fraudulent transactions. This is based on the 0.1% rather than the 99%. Because the problem clearly states that the 0.1% of ALL transaction is identified as fraudulent. The 99% of the computer program only deals with the correct identification of the transaction as either fraudulent or not. For me, it is not a clear measure of the true number of fraudulent transactions.