Given a large dataset of transaction logs, how would you optimize a SQL query to extract specific fraud patterns?
