A recommender system based on artificial immunity

The recommender systems encounter a series of challenges as e-commerce widens its scale and scope. The dramatic increase of information quantity reduces the efficiency of the recommender systems. Collaborative filtering is the most widely used personalized recommendation techniques. However, its related issues, such as accuracy, cold start, and extensibility, affect the recommendation quality. This paper explores the current E-commerce recommended algorithms and proposes a personalized recommended approach based on immune learning, clonal selection and self-adaption of the natural immune system. Our approach first clusters initialized antibody of the immune network. Then it applies self-adaptive aiNet algorithm on cluster centers for clonal variation. Differentiation attribute of antibodies could reduce the data sparsity. The inhibiting attribute of immune network could increase the extensibility. Therefore, the way avoids the local minimum. Compared to collaborative filtering, our approach provides a more accurate prediction on users´ interest and improves the quality of the recommender systems. Our experiment verifies its effectiveness and feasibility in the real recommender systems.