The Design And Implementation Of Personalized Recommendation System On Web Mining

With the development of Internet technology,people gradually from the lack of information era into the era of information overload.In this case,both consumers and producers of information are faced with severe challenges,the consumer that cannot quickly find what they are interested in from the mass,while the producers,how to push their own information to users might be interested in it as a very difficult thing.We need an automated tool that can analyze the historical behavior of users,tap the potential interest of the user,and then find the information the user may be interested in from the huge pool of Web resources,this tool is personalized recommendation system.Recommendation system is a bridge of users and information,which help users find the information,but also to better explore the long tail of information,so as to realize the "win-win" of information consumers and producers.The basic process of the recommendation system is to track user online behavior,in-depth analysis of user's history preferences then forecast information that users might be interested.On the one hand,what to collect and how to collect user behavioral data is the problem that we must first solved,only high-quality user behavior have guidance on the recommendation.On the other hand,huge amounts of data make the personalized recommendation system puts forward a new challenge.We should know how to deploy the system in order to ensure the reliability of the user behavior data storage,scalability and efficiency of analysis.Finally,the validity accurate recommendation result is a direct measure of the recommendation system,and only novel,diverse and accurate results in order to improve the user experience.In this paper,we put forward an integrated solution for the challenges faced by the recommendation system.Integrate of the explicit feedback and implicit feedback data effectively,explore the data may be hidden and then analyzes theuser's preferences.In the data storage and analysis module,the system selected Apache Hadoop cluster,HDFS ensured the reliability of information storage and scalability,MapReduce can ensure the efficient analysis.In recommendation engine module,the system will be combine Mahout and Apache Hadoop cluster to ensure high throughput,high concurrency features.We study of the Taste recommendation engine to resolve some issues such as cold-start and so on.At last,through a large number of experiments to verify the performance of the system,and recommended the accurate effectiveness of the result.