A Study Of M-commerce Personalized Recommendation System Based On The Ant Colony Algorithm

Nowadays, building M-commerce personalized recommendation system is a very important topic in the development of mobile ecommerce.Based on the differences between mobile commerce and desktop e-commerce, and the problems of traditional personalized recommendation algorithm, especially Collaborative Filtering technology, an Ant Collaborative Filtering (ACF) is proposed in this paper, the study applies one simulated evolutionary algorithm called Ant Colony Algorithm. In addition, it is structured that a personalized recommendation system fits for mobile commerce. Its fundamental idea is that the system can use Ants'behaviors of searching for foods to simulate people targeting products, and distinction between two objectives can be used as expected heuristic factor. Moreover, this paper raises users'timely ratings, which are composited with users'historical ratings as pheromones, by which people could collaborated with each other. On the basis of this analysis, the algorithm can get the possible products that meet uses'interests on the next step, which is output of the recommendation system.With the blossom of internet technologies, M-commerce grows at full speed, and that of information explosion, people have more choices, and are more focused on the timeliness and accuracy of information hunting, such as recommendation systems. At present, collaborative filtering algorithm, along with the other algorithms, is most widely used in recommendation systems. However, it still has some problems, such as the less accuracy of users'similarities, cold-start, etc., all of which can do harm to M-commerce. It can be helpful to get rid of the redundancies in the mass information to use Ant Colony Algorithm, which can greatly improve the recommendation speed and validity. Therefore, Ant Colony Algorithm proposes new solution thinking in hunting useful information.In this study, the nature of ACF is analyzed by simulation experiment. With the experiment result, it can effectively improve the cold-start problem in recommendation system, and it also can significantly raise the recommendation quality in the situation of extreme sparseness of user rating data.