Ant Clustering Algorithm And Its Application To Recommender System

The paper concludes the current clustering technologies and the research production of the collaboration filtering and recommender system. According to the research production of the ant-based clustering applying to the data clustering analysis, we study the ant-based clustering algorithm and implement clustering based on the similar interesting users in the recommender system.First, an improved ant clustering algorithm based on outlier is proposed and compared with related work. The improved algorithm can control the ant's behaviors randomly to decide the next load node quickly, and shorten the executive time. These results are compared with those obtained using established clustering techniques and we find evidence that the improved ant-based clustering is a robust and speed-up the convergence. The algorithm works by identifying groups of users who appear to have similar preference. It be applied as a "first step" for shrinking the candidate set in a nearest neighbor algorithm and balanced the between accuracy and throughput. Then, this paper constructed a collaborating recommendation model based on Agent. It putforward the time limit of users' interests and discus how to learn and renew the users' interests. The useful information will be push to the user with the memory-based algorithm. The system achieves the prospective goal through some practical instances, and the correctness of the way of thinking in this thesis is also validated.