LindaMac,
I guess, as a mathematical predictive tool, such models can be nearly accurate...perhaps 90% plus, till the kinks are ironed out, when they reach 99% plus accuracy, especially if the model can learn from mistakes.
The problem with what this model is trying to do is simply this: it assumes that there is a rhyme, reason, method in the madness... and therein lies its downfall. Mathematically, such models do not deal well with 'catasthrophes'... and the BCIS is surely that, in every sense of the word!