Research On Association Rules Recommendation Algorithm Based On User Context

Internet technology has brought a large amount of information and rich resources,which bring convenience to users.At the same time,users have no choice but to face a lot of meaningless information and services,that is,information overload.Personalization recommendation system can predict user information based on the users' historical behavior data or other constraints.Association rules recommendation is a common recommendation algorithm in personalization recommendation system.In this paper,data source is web logs which are the user browse information.The Apriori algorithm of association rule mining algorithm has two shortages: multi scanning database and not considering user context information.In order to improve the association rules recommendation algorithm,this paper introduces user context.Firstly,recommendation algorithm is improved based on the context pre-filtering.The recommendation algorithm uses single dimensional and multidimensional context information to classify original data.As a result that the model dimension of recommendation algorithm is reduced and that the amount of scanning data is also reduced.So,the improved recommendation algorithm can enhance efficiency.At the same time,the improved recommendation algorithm can effectively perceive needs of different users because of user context information.So,the improved recommendation algorithm can increase accuracy.Secondly,in order to make better use of user context information,the algorithm constructs a multi-dimensional association rules model on basis of the improvement of context pre-filtering,that is context modeling.The model combines user context information and recommendation items.Improved algorithm has a pruning before scanning data information through a theory.The theory is that two values of user context information cannot appearsimultaneously in an item set.The improved algorithm reduces the number of scanning databases.So,the improved recommendation algorithm can enhance efficiency.At the same time,the improved recommendation algorithm can increase accuracy to a certain extent because of the integration of user context information and recommendation items.Finally,according to huge amounts of web logs and the characteristics of layer by layer search iterative method of Apriori algorithm,this paper chooses Spark cluster as the experiments environment.Experiments results show that compared with the standard association rules recommendation algorithm,the improved algorithm based on context pre-filtering has high efficiency,accuracy and recall rate;in addition,compared with the improved algorithm based on context pre-filtering,the improved recommendation algorithm based on context modeling and context pre-filtering has higher efficiency,accuracy and recall rate.