>It also discussed programs done by anti-terrorists and anti-fraud units which check for suspicious withdrawals. >Everything gets tracked. I haven't worked for a bank IS, but it could be interesting to develop such programs.
Banks, at least in Canada, have been running DSS/AI/Anti-Fraud/Terrorist detection for years. But, they have put in arbitrary thresholds, such as $1000, or the like. The biggest issue is the number of false positives. A similar issue showed up with the scanning of e-mails for violent/terrorist lamguage on the INTERNET. Every teenage kid playing World of Warcraft got flagged. I got flagged once, at work, for using a very vile word in an e-mail. I didn't. I was just discussing Soccer and a town that ended in "thorpe'. The word was pulled out of the middle of a larger word, without delimeters. Another example, not financial, is at a company I used to work for. The service provider had introduced a SPAM filtering package that kept suspected SPAM away from the recipient, so the intended recipient could not verify that it was SPAM. So, there was no Human Intervention and approval. But, the provider was using the percentage rejected as a performance metric. When I asked about false positives, they told me I didn't understand the issue. The whole point is, any AI algorithm needs a human overseer. It's not good enough on its own, yet. - Too busy driving to stop for gas! ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
