Research On Personal Recommender Algorithm Based On Artificial Immune Network In E-commerce

Nowadays,not only e-commerce policies,but also logistics,credit ,electronic payment and other support systems become more and more perfect.With the restructuring of the market economy,consumer will be in more need of e-commerce in their daily life.Their awareness of using e-commerce will become stronger step by step.You will find that e-commerce is used in almost all the areas of production, circulation and consumption in the future.The e-servies group will emerge gradually.This will become a new growth point of national economy.Statistics show that the volume of business transactions in e-commerce market keeps growing since 2005.The investment markets of China will show a good state of development in the next three years.In this condition,in order to promote their business,many e-commerce sites have developed recommended system for providing personalization service to consumers.Recommended system has became an important research part in e-commerce.Recently,with the scale of e-commerce getting larger and larger,it facs a series of challenges. The rapid growth of information affects the efficiency of the recommend system.As a personalized recommended technology,collaborative filtering can be used in most areas.However, collaborative filtering algorithm has limitation of it's own,such as sparsity, cold start, and scalability problems.etc.In this paper,we make a useful exploration and research on recommendation algorithm in e-commerce.As natural immune system has the characteristic of immune learning, clonal selection and adaptive,we develope a hybrid recommendation system called collaborative filtering & clustering & immune recommendation system. Firstly,the clustering algorithm is utilized to cluster initial antibody into several classes,then centre users wil be cloned according to aiNet adaptive algorithm to obtain a good neighbor set. The algorithm can take the advantage of the clone and mutation of the artificial immune network to reduce the data sparsity and use the clone suppression and network suppression to reduce the data dimension to improve scalability,so can avoid falling into local optima.Compare with collaborative filtering algorithm,this algorithm will be more accurate in prediction users'interest,so the quality of the recommendation system can be effectively improved.Simulation results shows that the presented algorithm is effective and feasible in improving the performance of CF systems in the end.This paper is mainly focused on following aspects:(1)The recommendation system based on artificial immune This paper draws on the interaction between antigens and antibodies, antibodies and antibodies, the mechanism of antibody cloning and variation.We introduce the idea of the adaptive aiNet to select neighbors for collaborative filtering algorithms.This algorithm can improve the accuracy of the selection of nearest neighbors.(2)The recommendation of the immune system based on clusteringThis paper makes the clustering algorithm into use, divides registered users in web basing on their interest. and then combines the idea of artificial immune,uses immune-based clustering to select neighbor users.(3)The calculation of the user's ratingThis paper considers density of neighbors when predicts the interest of target user. So it can improve the prediction accuracy indeed.