Abstract: We present an ILP approach for mining the user access patterns with DL-safe rules.This method defines the log ontology based on events, and combines log ontology and DL-safe rules into a decidable hybrid knowledge base.The useful user access-patterns can be generated from the knowledge base by ILP learning method.The semantics of the relations between events in log ontology can improve the results and efficiency of ontology learning effectively.The results of the method provide more decision-making for improving the structure of web sites.The experimental results show that this method is decidable and it is quite feasible to solve practical problems.