Ans. Forward (data driven) chaining is appropriate when there exist many equally acceptable goal states, a narrow body of relevant information (facts & rules) and a single initial state.

Backward (goal driven) chaining is appropriate when there exists a single goal state and a large amount of potentially relevant information.

In the forward chaining approach it is fundamental to note that the inference process proceeds exhaustively from the existing facts to a set of new facts.

The backward chaining seeks only to prove the validity of a chosen fact (whose truth value is not known) expression. It is computationally efficient than forward chaining, since it represents a goal directed strategy that may eliminate checking of many superfluous paths.

Data- driven inference is preferable when: (1) All or most of the required facts are in the initial database. (2) It is difficult to initially form a goal or hypothesis to be verified.

Goal - directed inference is preferable when: (1) Relevant data must be acquired as a part of the inference process. (2) Large numbers of applicable rules exist.