This afternoon I went to the local Panera and paid by credit card. My bank declined my charge of $4.82. I figured it was the magnetic strip on the card which had failed or that the new trainee using the cash register may have made a mistake. She ran the card three more times and it was rejected. Then I got four text messages from the bank saying that they are rejected my charges. To text me, they used my phone number.

I called. They had put a hold on my card because they had some questions about my charges from the prior few days. The red flag event was that I had made an earlier charge of $9.65 at Panera about eight hours before. Their computer program was not smart enough to figure out that it was not unreasonable for someone to have breakfast at 6:30am at a Panera in Durham and then walk into a Panera in Chapel Hill later in the day with 30 minutes to kill and had a coffee (and a Danish I probably should not have had) while I played with my iPad on their free wireless connection. The computer also questioned the $1 charge at a gas station this afternoon (which the human representative immediately recognized as the established practice of gas stations opening charge lines with their automated payment systems of $1 when you swipe your card and then next day putting a $92 charge on the card for filling the tank). I was also asked if the payment made on the account was one I had made (I asked the customer service rep if she thought that if someone had paid a bill for me that I would tell her it was an erroneous transaction and she laughed for a long time) as well as a $71 charge to a software company outside the US.

They had freaked out because they could not reach me by phone at three numbers that were old ones not active (I know they have my current number because they sent me texts at it and same bank sometimes calls about my other accounts at the cell phone I never turn off and which has a voice mailbox). Of course, if they did not have a no reply text address, I could have responded to the four texts they sent.

Predictive models have been around for a decade or more in banks as they attempt to identify fraud and protect themselves. The episodes I have with my bank about every 2-3 months illustrate what happens when somebody blindly runs predictive analytic programs through big datasets without using some commonsense to guide the modeling process. Just because anyone can buy a $100,000 program from IBM or others for developing predictive analytics does not mean that the model that comes out of the Big Data and expensive program makes any sense at all.

Or that the NSA or FBI or CIA or Google or Amazon models make much sense as they probe your private information.

If a computer predictive system is going to think that somebody is committing credit card fraud because they purchase two cups of coffee at the same national restaurant chain in a day, we are in big trouble.

The bottom line is that Big Data models are going to have to be regulated before some idiot accidentally turns on Sky Net.

Or maybe the problem is that the NSA or FBI or CIA or Google has done it already.