Research On News Recommendation Algorithm Based On Weighted Sequential Pattern

Currently,there are various news recommendation systems on the internet,and the vast majority of them are based on content filtering and collaborative filtering.However,Content filtering based news recommendation merely takes into consideration the preferences of single user,and depends on keyword tags seriously.Although collaborative filtering news recommendation takes into consideration the group behaviors,it still might cluster readers to a narrow viewpoint to counter these drawbacks,some scholars recognize that reader's news reading is often a sequential pattern,so they introduced sequential pattern into news recommendation algorithms.Sequential pattern mines user's reading sequences and discovers the relationship between sequences,predicting articles that users might be interested in based on reader's previous reading sequence.First,the improving strategies of weighted sequential pattern algorithm are analyzed,due to the lack of consideration of sequence weight of WSpan algorithm,the improved algorithm takes into consideration of sequence weight,and maintain the downward closure property by a minimum support,then prune the candidate sequences multiple times to decrease the set of sequences.Second,considering the dynamic characteristic of sequence database,this dissertation proposed a update method to update the frequent sequences set,the proposed update method exploited the already computed results,thus make the update process efficient.Third,this dissertation proposed a news recommendation algorithm based on the improved weighted sequential pattern algorithm,the recommendation algorithm computes user similarity based on the whole set of sequences and a matching matrix,the algorithm make recommendations based on the most matched user and improved the choosing strategy of candidate items,the dissertation proves the efficiency of the proposed recommendation algorithm to traditional algorithms by test analysis.