In this paper, an inductive learning algorithm CAP2 is described, which is a binary tree classification method, with the division probability of state space as heuristic information, based on the provided example set. CAP2 can satisfy the requirement of precision of users to induce the rule set,of the proper precision,and it can deal with continuous data, noisy data. Background knowledge provided by users can be used by CAP2. CAP2 has been successfully applied in real world. Also its comparison with ID3 is given.