Study On Real-Time Personalized Commend System Based On GA

Service is a solution for the "Information Overload" and "Information Labyrinth" problems in the context of the "Information Explosion". It will be the main mode of information service in the future.In this paper, I present a personalization e-commerce site system based on web usage mining. This system adopts genetic algorithm and clustering to mining user visiting history records. Then we classify user's visiting by visiting mode. And last, we can recommend personalization information to user according the visiting records of similar visiting mode.Nowadays, most clustering methods based on genetic algorithm adopt fixed length chromosome .The amount of clustering result generated by those methods is pre-seted. In this paper, on the work of fixed length chromosome Genetic Algorithm, I present a clustering method based on variable length chromosome Genetic Algorithm, which can dynamic output the results of clustering without inputting some parameters (such as clustering amount, initial clustering centers, the minimal distance between clustering centers, etc.). Thus, setting wrong parameters won't influence on clustering results. The variable length chromosome GA can cooperate with fixed length chromosome GA. When we can determine the number of clustering by experience, we can use the fixed length chromosome GA. And when we can't determine it, we use variable length chromosome GA. Then, this system can have good efficiency and good effect.This system's strong suits as follows:(1) This system adopts clustering method based on Genetic Algorithm to mining user visiting modes. Clustering based on GA doesn't need input initial clustering centers, the minimal distance between clustering centers, etc. So, clustering will not be sensitive to such parameters.(2) The clustering method is based on fixed length chromosome GA and variable length chromosome GA. It can improve the efficiency and the quality of data mining.(3) It is a real-time commending system. So it can reflect user's changing interesting. System captures user's visiting sequences real-time, and compartmentalizes it to a visiting mode. (4) This paper's researching background is Web Shopping Site. This system commends products based on both static and dynamic pages, so it can be adopted insome dynamic sites like e-business. (5) This system doesn't need user's participation. The system can commend personalized info to users with neither registering information nor logging in site. It is 'zero input' personalization.And this system focus on product pages, it wouldn't be effected by other middle pages. Moreover, it considered the sequence characteristic of visiting, which reflects users' interesting.