Backward chaining, in which the inference engine begins with a goal, or hypothesis, and searches for data to support this goal. Since it is the goal that initiates the problem-solving process, backward chaining is sometimes referred to as goal-driven reasoning. In this approach, inference engine searches for a match between the consequent (the THEN part of the rule) and the system goal; when the match is found, the system continues to search the knowledge base until it finds evidence either to support or to reject the goal.
Definition