TABLE OF CONTENTS Of the many types of information technology potentially useful for risk management in e-business, we are interested in decision support systems that help generate business rules that manage risks, while fulfilling other business objectives. This issue can be naturally thought of as the problem of deciding which actions to take in order to minimize the expected risk (or cost). Numerous offthe-shelf statistical and data-mining methods can be used to assess the risks associated with possible actions. Viewed this way, risk management becomes a traditional decision theoretic problem, and as such there is a large body of work in related fields such as decision theory, statistics, machine learning, and data mining that can provide a solid theoretical basis for it.
One aspect of the risk management problem that has not received proper attention is the intrinsically sequential nature of this problem. That is, it is hardly satisfactory to consider the risk of a single action in isolation, but rather it is necessary to consider the long-term risks of sequences of actions taken over time. For example, consider the problem of deciding to whom to make a certain campaign offer for the purpose of churn detention. In the isolated problem setting, one might conclude that the optimal decision is to make the offer just in case the expected profit resulting from that mailing exceeds the mailing cost. However, it may very well be the case that that customer...
