With the ever expanding Web and the information published on it, effective tools for managing such data and presenting information to users based on their needs are becoming necessary. In this paper, we propose a new algorithm named "EIN-WUM" for Web usage mining based on artificial immune system metaphor. This algorithm introduces several novelties such as using danger theory, directed mutation and an enhanced immune network model. Experimental results show that The EIN-WUM algorithm can properly learn the frequent trends in noisy, sparse and huge Web usage data in single pass.