Current inductive learning algorithms have difficulties handling attributes with numerical output values. This paper presents FuzzySRI-II, a new fuzzy rule induction algorithm for the prediction of numerical outputs. FuzzySRI-II integrates the comprehensibility and ease of application of rule induction algorithms with the uncertainty handling and approximate reasoning capabilities of fuzzy sets. The performance of the proposed FuzzySRI-II algorithm in two simulated control applications involving numerical output values is demonstrated and compared to that of the recently developed RULES-F Plus fuzzy rule induction algorithm. Results show that the rules derived from FuzzySRI-II are simpler and yield higher accuracy than those from RULES-F Plus.