Rule induction plays an important role in knowledge discovery process. Rough set based rule induction algorithms are characterized by excellent accuracy, but they lack the abilities to deal with hybrid attributes such as numeric or fuzzy attributes. In real-world applications, data usually exists with hybrid formats, and thus a unified rule induction algorithm for hybrid data learning is desirable. We firstly model different types of attributes in equivalence relationship, and define the key concepts of block, minimal complex and local covering based on fuzzy rough sets model, then propose a rule induction algorithm for hybrid data learning. Furthermore, in order to estimate performance of the proposed method, we compare it with state-of-the-art methods for hybrid data learning. Comparative studies indicate that rule sets extracted by this method can not only achieve comparable accuracy, but also get more compact rule sets. It is therefore concluded that the proposed method is effective for hybrid data learning.