| One common myth is that data mining is intended to provide automatic alerts of terrorist attacks. Such a goal is almost certainly not achievable because terrorist attacks are rare, and the population of such attacks is both diverse and quickly evolving. Tomorrow's attacks are unlikely to look like today's.
Instead, it makes more sense to direct data mining techniques toward predicting lower-level activities that are both more consistent and more common. Such activities include illegal immigration, money transfers, operating front businesses, and engaging in recruiting activities. If predictive models of these activities could be constructed, they could be used to direct analysts' attention to situations that warrant greater attention.
|
