Research And Application Of Recommendation Algorithm Based On User Web Access Log Clustering

Nowadays,with the rapid development of Internet,the amount of data in the network has become more and more large.In the face of hundreds of millions of data,the problem is how to find the data of interest for the internet users.While for the Web site operators what they should consider is how to search to find users' interest and recommend to users with the geometric level of growth of network data.In the environment of big data,through a simple artificial screening has been unable to complete this task.The combination of the Internet and data mining is a good way to solve this problem.For users,the Internet can be for their choice of all kinds of resources,users need a technology to be able to read and understand the user's interest,and to predict the user's preferences in the future.In view of the above two requirements,this paper uses the access log of Internet users as the data source,after processing the user log data.Then based on the theory of data mining and personalized recommendation,this paper puts forward an improved clustering algorithm based on Hamming distance and based on the loglikelihood ratio of a collaborative filtering recommendation algorithm recommendation application mode.It mainly includes two processes:first,it finds the target users and users with similar interests through the clustering algorithm.Secondly,it tries to find the nearest neighbor N with a target user in the class of the masses through the TopN algorithm,then according to the access log context information of users for the target users to recommend appropriate cyber source.It is extremely easy to obtain the access log data of users.As long as the users browse the web according to their own interests,it can produce the recommended data source.The interest of a general user in a period of time is not changing too much,so the user can be taken offline after clustering experiments,real-time computation is not required to reduce the computational scale.The efficiency and accuracy of the recommendation can be improved by using the clustering algorithm to filter out a lot of users who are interested in different users.Because of the above research,this paper also compares the efficiency and accuracy of clustering and recommendation algorithms in the final application research.